Mass and number size distributions of particulate matter components : comparison of an industrial site and an urban background site

Size-resolved composition of particulate matter (PM) sampled in the industrial town of Port Talbot (PT), UK was determined in comparison to a typical urban background site in Birmingham (EROS). A Micro-Orifice Uniform Deposit Impactor (MOUDI) sampler was deployed for two separate sampling campaigns with the addition of a Grimm optical spectrometer at the PT site. MOUDI samples were analysed for water-soluble anions (Cl - , NO 3- and SO 42- ) and cations (Na + , NH 4+ , K + , Mg 2+ and Ca 2+ ) and trace metals (Al, V, Cr, Mn, Fe, Cu, Zn, Sb, Ba and Pb). The PM mass distribution showed a predominance of fine particle (PM 2.5 ) mass at EROS whereas the PT samples were dominated by the coarse fraction (PM 2.5-10 ). SO 42- , Cl - , NH 4+ , Na + , NO 3- , and Ca 2+ were the predominant ionic species at both sites while Al and Fe were the metals with highest concentrations at both sites. Mean concentrations of Cl - , Na + , K + , Ca 2+ , Mg 2+ , Cr, Mn, Fe and Zn were higher at PT than EROS due to industrial and marine influences. The contribution of regional pollution by sulphate, ammonium and nitrate was greater at EROS relative to PT. The traffic signatures of Cu, Sb, Ba and Pb were particularly prominent at EROS. Overall, PM at EROS was dominated by secondary aerosol and traffic-related particles while PT was heavily influenced by industrial activities and marine aerosol. Profound influences of wind direction are seen in the 72-hour data, especially in relation to the PT local sources. Measurements of particle number in 14 separate size bins plotted as a function of wind direction and speed are highly indicative of contributing sources, with local traffic dominant below 0.5 µm, steelworks emissions from 0.5-15 µm, and marine aerosol above 15 µm.


INTRODUCTION
Particulate matter is commonly classified in three modes namely ultrafine (nucleation and Aitken mode, diameter less than 0.1 µm), fine (mainly accumulation mode, aerodynamic diameter between 0-2.5 µm) and coarse (aerodynamic diameter between 2.5 and 10 µm). Generally, fine and ultrafine PM are formed from high temperature processes such as vehicular exhaust, oil and coal combustion, biomass burning, industrial processes, and chemical reactions in the atmosphere (Harrison et al., 2003a;Samara et al., 2003). Coarse particles are generally evolved from attrition processes including mechanical abrasion of crustal material and re-suspension of road and soil dusts, sea spray, volcanic eruptions and brake and tyre wear from vehicles (Allen et al., 2001).
Atmospheric PM is made up of diverse chemical substances including water soluble ions, trace metals and organic compounds. Water-soluble ions constitute a significant portion of PM mass (Yin and Harrison, 2008), and therefore play an important role in aerosol chemistry. Sulphate and nitrate are formed mainly from oxidation of SO 2 and NO x . Sodium, magnesium and chloride are the main components of sea spray; potassium arises from biomass burning or soil and Ca from construction, soil and steelworks emissions (Oravisjavi et al., 2003;Pandolfi et al., 2011). A number of anthropogenic, geogenic and biogenic activities are responsible for emissions of trace metals into the atmospheric environment and hence play important roles in determining size distributions (Allen et al., 2001). Each has a characteristic size distribution reflective of its source.
The Micro-Orifice Uniform Deposit Impactor (MOUDI) has been widely used for particle size measurement in both indoor and outdoor pollution studies. These studies have reported PM size distributions for water soluble and trace metal components (Allen et al., 2001;Cabada et al., 2004;Harrison et al, 2003a;Chang et al., 2008;Dall'Osto et al., 2008;Liu et al., 2008;Zhao and Gao, 2008;Gietl et al., 2010;Ny and Lee, 2011). Na, Cl, Ca and Al typically show modes in the coarse fraction while Cd, Zn, Mn, Ni and Cu have modes in the fine fraction (Ny and Lee, 2011). Allen et al. (2001) reported MOUDI data for trace metals from three urban background sites in the UK.
Dall 'Osto et al. (2008) employed a MOUDI for particle size-resolved measurements at a steel industry site in the UK. Water-soluble ions in particulate matter, from nanoparticles to the coarse mode have been determined with a MOUDI and nano-MOUDI in Taiwan by Chang et al. (2008). In London, particle size-segregated aerosol has been measured at roadside and background sites by Gietl et al. (2010). PM size-segregation and associated metallic elements in an industrialized city in Korea have been reported (Ny and Lee, 2011). None of these studies have compared the size distribution of PM and its components collected at urban background and industrial (especially steelworks) locations for both water soluble ions and trace metals. In this study, particle size distributions of both ionic species and trace metals at typical background and industrial sites were studied, offering an opportunity for identifying source signatures of components contributing to PM in the atmosphere of the two study areas. The results add to a relatively small international database of mass size distributions for specific chemical components, and highlight the very different sources of particulate matter at the two sites. Additionally, particle number spectra from 0.3 µm to > 15 µm are reported and analysed according to wind speed and direction.

Port Talbot (PT)
PT is a coastal industrial town with a population of approximately 35,000 and located on the M4 corridor in South Wales (51° 34' N and 3° 46' W). The Tata steel complex located in Port Talbot town is the main industry in the study area and a major source of PM emissions (AQEG, 2011). The site covers approximately 28 km 2 , comprises of ~50 km of roads, 100 km of railway, and 25,000 vehicle movements per day. The production capacity is around 5 m tonnes per year with the main processes in the steelworks being iron-making (sintering, blast furnace and raw materials), steelmaking (basic oxygen steel-making (BOS) and coking) and rolling mills (hot and cold mills) (Moreno et al., 2004;Dall'Osto et al., 2008). The location of the sampling site (Fire Station) is shown in Figure 1. (population of over one million and part of a conurbation of 2.5 million population (Yin et al., 2010). The EROS study site is shown in Figure 2. The nearest roads are lightly trafficked and the nearby railway line carries mainly electric trains.

Sample Digestion and Analysis
Exposed filters were cut into two equal halves. One half was analysed for water-soluble ions (Na + , Mg 2+ , Ca 2+ , K + , NH 4 + , Cl -, NO 3 -, SO 4 2-) while the second half was analysed for trace metals (Al, V, Cr, Mn, Fe, Cu, Zn, Sb, Ba and Pb). Blank filters were also analysed and subtracted from the exposed filter data.
Filter samples for ionic species analysis were extracted with 7.5 ml distilled de-ionized water (DDW, conductivity of 18.2 MΩ) with agitation on a mechanical shaker (Model IKA Labortechnik KS 250 basic) set at 240 rpm for a period of 40 minutes. Prior to addition of DDW and shaking, PTFE filters were first treated with 0.3 mL propan-2-ol to wet the surface. The extracts were analysed with a Dionex DX 500 and ICS-2000 Ion Chromatography (IC) system for cations and anions respectively. Eluants used were methane sulphonic acid (MSA) and KOH for cation and anion IC instruments, respectively. Detailed procedures for the Dionex analysis have been described by Yin et al. (2010).
The second half of the exposed filters was extracted with dilute reverse aqua regia (2.23 M HCl and 1.03 M HNO 3 ) following the procedure of Harrison et al. (2003b). The mixed acid extractant (2 mL) was added to the filters placed inside 4 mL narrow neck bottles and heated at 100 o C for 30 minutes in a water bath and then placed in a ultrasonic bath at 50 o C for another 30 minutes. This cycle was repeated and the resulting digests transferred into 15 mL narrow neck bottles and rinsed with 10 mL distilled deionised water. The extracts were then analysed by Inductively Coupled Plasma Mass Spectrometry (ICPMS, Agilent 7500 Ce). Mixed standards (from stock 1000 mg/L VWR standard solutions) were prepared in the series 0, 1, 5, 10, 20, 50 and 100 ppb. Internal standards employed by the ICPMS instrument were Sc, Ge, Y, In and Bi.
In earlier work, this procedure has been evaluated by analysis of NIST SRM 1684a (Allen et al., 2001) and more recent trials have shown high efficiency (< 85%) for all elements analysed except Al (ca 50% efficiency, also by comparison with XRF) and Cr (ca 50% efficiency). Since extraction efficiency may vary according to particle size, this may introduce bias into the size distributions reported for these elements.

Number Size Distributions
Particle number size spectra in the range 0.3 to 20 µm were measured at the Port Talbot site using a Grimm model #1.108 optical particle spectrometer which measures the intensity of light scattered at 90ºC to a beam produced by a laser diode. Particles are classified into 14 size fractions between the smallest (0.3-0.4) and largest (15-20 µm) bins. Figure S1 represents the temporal variations of MOUDI data for mass and ionic components from the industrially influenced PT site. Bimodal peaks (0.5 and 2-6 µm) were observed for mass distribution for each MOUDI sample during the campaign with highest mass concentration observed on April 23-26 which coincided with the start of a PM 10 episode on April 26th reaching hourly concentrations in excess of 100 µg m -3 . Most of the water soluble components displayed two distinct modal peaks in the fine mode except Cl -(other than the May 11-14 sample which showed two modal peaks), NH 4 + and Ca 2+ . Common fine modal peaks of 0.5 µm and coarse modal peaks of 2-6 µm were observed for most of the ionic species except for K + which showed diverse modal peaks in various sampling periods, suggesting multiple sources of K + in Port Talbot. However, the common peak at 0.5 µm shown by K + in all the samples indicated a common emission source for which biomass combustion or steelworks activities are the major suspects. An episode of Ca 2+ , Cl -, and NO 3 was observed between May 2-5 all showing higher concentrations at 3.0 µm. This was also observed for ammonium and sulphate with a peak at 0.5 µm. Elevated Ca 2+ observed during this episode might have sources either from construction activities or steelworks emissions. In the blast furnace of the steelworks, limestone (CaCO 3 ) and dolomite (CaMg(CO 3 ) 2 ) are used as fluxing agents while in the basic oxygen furnace (BOF) section, lime (CaO) and fluorspar (CaF 2 ) are fluxing raw materials (Machemer, 2004). The MOUDI sample of April 20-23 showed higher concentrations of Cl -, Na + and Mg 2+ all peaking in the particle size range 3-5 µm. This typifies a sea spray episode. Figure S2 shows the trace metal size distribution of MOUDI samples during the one-month campaign. The concentration of Fe was the highest, followed by Al (may be underestimated as mentione dabove) and Zn. All the metal data show two or more modal peaks covering the fine and coarse particle ranges except Pb and Zn which have their modes in the fine fraction only (0.5 and 1-2 µm). Mn and Fe showed similar modes throughout the sampling periods suggesting common emission sources. In the fine mode, most of the trace metals peak at 0.5 µm except for the April 26-29 sample which showed exceptional elevated peaks for Cr and Cu at 0.2-0.3 µm. Combustion of copper chrome arsenate (CCA) treated wood could explain the unusual peaks of these trace metals on these days. CCA was used as a wood preservative before its ban in the UK in 2007 (Wood Protection Association, 2010). However, CCA treated wood may still remain in many structures.

Port Talbot
Despite the fact that Cu is often a good marker for traffic (brake dust), its appearance at 0.2-0.3 µm is not consistent with traffic as other traffic markers such as Sb and Ba were peaking at a different diameter (0.5 µm). Additionally, in the size range 1-2 µm significant amounts of Al and V appeared in the April 26-29 MOUDI sample. Apart from this date, V peaked at 0.5 µm throughout; but Al showed different fine modal peaks at 0.2-0.3, 0.5 and 1-2 µm. Al in the fine mode could arise from vehicular, coal combustion and metallurgical activities (Kleeman et al., 2000;Xia and Gao, 2010) while fine V normally arises from fuel/oil combustion and shipping emissions (Figueroa et al., 2006;Pandolfi et al., 2011). The appearance of irregular peaks on April 26-29 implies that Al and V could arise from the same sources that emitted Cu and Cr into the atmosphere. The wind rose for that sampling interval (see Figure S5 in Supplementary Information) indicates that the source is unlikely to be associated with the steelworks.

EROS
Individual 72-h MOUDI size segregated mass and water soluble ion distributions from the EROS site are represented in Figure S3. The gravimetric mass showed modal peaks in the fine and coarse PM ranges at 0.5 µm and 2-4 µm. Among the water soluble ions, Cl -, Na + , Mg 2+ and Ca 2+ depicted a unimodal peak at 3-6 µm. Ammonium exhibited a single peak in the fine mode (0.5 µm). Other ionic species showed bimodal behaviour generally peaking at 0.5 and 3 µm except in the April 8-11 sample where NO 3 peaked at a single mode of 1-2 µm. A sulphate coarse mode occurred at around 5.0 µm. Among the ionic components of PM, sulphate constituted the highest concentration while K + <Mg 2+ <Ca 2+ were the ionic species of lowest abundance.
MOUDI trace metal size distributions at the EROS sampling site are shown in Figure S4. Unlike water soluble ions where the highest concentrations were observed in the March 30-April 2 sample, metal concentrations were higher on April 5-8. All of the metals displayed at least two peaks covering fine and coarse PM fractions except Pb, Zn, Sb where the two modes were both in the fine fraction, other than in the sample collected on April 8-11 which displayed a coarse mode peak at 2-3 µm for Sb. Multiple modes were observed for Al, V and Cr (0.5, 1-2, 3 and 10 µm). On March 30-April 2, V, Cr, Cu, Zn, Sb and Pb all showed similar peaks at fine modes of 0.5 and 1-2 µm indicating possible emissions from traffic and industrial combustion processes (Zanobetti et al., 2009). Zn measured on April 5-8 showed a coarse modal peak which might be linked with traffic emissions of brake or tyre dust (Harrison et al., 2012a). Different peaks exhibited by the measured metals during different sampling periods suggested diverse emission sources.

Comparison of Average MOUDI Size Distributions from EROS and PT
A summary of the source assignments and modes in the size distributions appears in Table 1. Figure 3 shows the average size distribution data for mass, water soluble ions and trace metal concentrations at EROS and PT sites. Different patterns were observed for these particle constituents. At both sites, mass concentrations, Na + , K + , Fe, Cu and Ba exhibited two modal peaks generally occurring at 0.5-0.6 µm and at 2-6 µm. Average mass concentrations showed a slight elevation of fine particle mass at EROS while an elevated concentration of coarse particles was evident at PT relative to the other sites. Meteorological conditions, especially wind speed and temperature are known to influence particle concentration and size distributions (Charron and Harrison, 2005;Jones et al., 2010). During dry and windy conditions, elevated amounts of coarse particles could be favoured due to re-suspension of soil, road and industrial dusts.
Generally, the modes in the size distributions were similar at the two sites ( Figure 3). Notable differences in the modes were as follows: • calcium showed a fine mode at PT which was not evident at EROS. This may be an emission from the steel industry.
• aluminium showed a pronounced mode > 10 µm at EROS which has no parallel at PT, and probably arises from resuspension of road dust (Harrison et al., 2012a).
• chromium shows multi-modal behaviour at both sites, but the smallest mode at PT is at around 0.2 µm, whereas at EROS it is close to 0.5 µm. Either steelworks emissions or combustion of wood treated with copper chrome arsenate (CCA) preservative may be the source.
• manganese, for which PT shows a very pronounced coarse particle mode at ca 5 µm, while at EROS, the mode is much broader, peaking at 1-2 µm.
• iron, for which both sites show a bimodal distribution. The most obvious difference is the coarse mode, which centres on 5 µm at PT and 2 µm at EROS, and parallels the behaviour of manganese. The mode at 2 µm seems likely due to brake dust, as that for Cu and Ba at EROS are at a similar size, close to that reported by Gietl et al. (2010) from a site in central London. The large peak at ca 5 µm for both Fe and Mn is most probably associated with emissions from the steelworks, based upon the much elevated concentrations observed at the PT site (see Figure 3).
Components showing similar size modes, but different relative abundances were: • particle mass, for which the fine mode was very similar at both sites, but the coarse mode was much more prominent at PT due largely to increased marine aerosol (Cl -, Na + , Mg 2+ ) and metallic elements, most notably Fe and Mn.
• coarse Cland Na + from marine aerosol reflecting the proximity of the PT site to the sea.
• SO 4 2and NH 4 + which showed a greater abundance at the EROS site and are primarily associated with regional transport of secondary ammonium sulphate.
• NO 3 is markedly coarser than ammonium at both sites, and a little finer than Cland Na + , consistent with an association with aged marine aerosol (Ottley and Harrison, 1992).
Concentrations are higher at EROS, probably reflecting the east to west gradient in secondary nitrate and sulphate observed across the UK (AQEG, 2012), although the relatively short duration of sampling is insufficient to establish this beyond doubt.
• K + , which shows similar modes, but significantly higher concentrations at PT. Woodsmoke is usually the main source of fine K, while the coarse K may in part reflect marine aerosol, with also a likely contribution from steelworks emissions.
• coarse Ca 2+ is also highly elevated at PT, probably reflecting emissions from steelworks processes and stockpiles.
• Al and V show elevated relative concentrations at EROS, probably reflective of soil/road dust as a source of Al (although the fine mode is hard to explain) and fuel oil combustion for V.
• Mn and Fe both show a much enhanced coarse mode at similar sizes at PT, consistent with a common steelworks source.
• Cu, Sb and Ba are all elements associated with brake wear emissions (Gietl et al., 2010). All appear to show a brake dust component at both sites -the mode at 3 µm is characteristic (Gietl et al., 2010). In all cases, the data from EROS show higher concentrations. Published studies have reported Cu/Sb values of 4.6 and 5.3 for brake wear emissions (Sternbeck et al., 2002;Hjortenkrans et al., 2007). Studies on major roads have measured higher Cu/Sb values of 8.7-12 for fine and 10.5-13 for coarse particles (Song and Gao, 2011); and 9.1 (Gietl et al., 2010). In this study, these ratios were calculated to be 2.3, 7.2 and 3.7 (fine, coarse and PM 10 ) at EROS and 4.3, 5.4 and 4.9, respectively at PT. In the coarse mode the Cu/Sb ratio obtained at EROS was closer to Song and Gao (2011) and Gietl et al. (2010) consistent with a traffic influence. Cu, Sb, Pb and Zn all show a marked peak at 0.4-0.5 µm, probably associated with a higher temperature source, in the case of Cu, Sb and Pb most prominent at EROS.
• Zn shows modes at 0.4 µm and 2 µm which are more abundant at PT, especially the coarser mode. This is an element which as previously been associated with steelworks emissions (Dall'Osto et al., 2008  and 13.5±7.2 (PT). The far higher ratios of Ca/Al at PT are most likely attributable to an emission from the steelworks (see previous section), as the only major local construction activity related to earthworks associated with road building, particularly the Port Talbot Peripheral Distribution Road, which would be expected to cause an elevation in both elements.
Calculation of chloride depletion adopting the formula given by Zhao and Gao (2008)  SO 4 2+ and NH 4 + appearing in the fine mode at 0.5-0.6 µm suggests formation of (NH 4 ) 2 SO 4 or NH 4 HSO 4 in the droplet mode. The droplet phase reaction involving oxidation of SO 2 to sulphate is very important in the atmosphere (Khoder, 2002). The modal peaks of 0.5 µm and a small mode between 1-2 µm observed for SO 4 2at PT coinciding with those of Zn and Pb suggested internal mixing between SO 4 2and Pb, and Zn. Formation of PbSO 4 and ZnSO 4 from the reactions involving Pb, Zn and SO 4 at the two modal sizes might have evolved from separate emission sources, which might be linked to the steel works units or other anthropogenic activities. Previous studies at the steelworks have identified Pb and Zn as major emissions from the sinter and basic oxygen steel making sectors (Oravisjavi et al., 2003;Dall'Osto et al., 2008).
At the PT site, Dall'Osto et al. (2008) found two peaks of Pb occurring at fine modes of 0.4-0.5 µm and 1-2 µm, which is similar to the modal peaks of Pb obtained in this study. Also, Dall'Osto et al. (2008) reported Zn to have a lone peak occurring at 1-2 µm relating to that obtained in this present study. However, another larger mode was found for Zn at 0.5 µm in this study. A small peak of . This may be an emission from the sinter process. Dall'Osto et al. (2008) reported different coarse modes for Fe dependent on the emission source. Such behaviour is also reflected in the size distributions for Fe seen in Figure S4.

Influence of wind conditions upon size distributions
Wind roses for the respective periods appear in Figure S5.

Port Talbot
The periods showing the greatest mass concentrations in the coarse fraction ( Figure S3) are April 23-26, May 8-11 and May 5-8. All of these show a significant component to the wind rose in the southerly to westerly sector consistent with atmospheric transport from the steelworks. These periods also correspond to a major elevation in the concentrations of iron and manganese ( Figure   S4). The period April 23-26 was also associated with high concentrations of zinc, barium, coarse vanadium, coarse calcium and coarse potassium, all of which have potential sources within the steelworks. The periods with the highest concentrations of coarse sodium and chloride was April 20-23 consistent with strong winds in the westerly sector transporting marine aerosol. The period with the greatest association with the easterly wind sector were April 26-29, April 29-May 2 and May 2-5. The latter period was associated with high concentrations of nitrate, sulphate and ammonium, the secondary constituents whose concentrations are typically elevated in easterly air masses in the UK. Also elevated in this sample were concentrations of calcium and chloride with a mode at around 3 µm consistent with that of nitrate. This would appear to be due to transport of crustal calcareous material from the land masses to the east of Port Talbot which had collected nitrate and chloride from the vapour phase during their transport. Concentrations of iron and manganese were very low in these easterly samples.
Coarse particle modes for barium, antimony and copper, all constituents associated with automotive brake wear were notably elevated in the sample of April 17-20 consistent with a major automotive contribution on winds from the north-west. This sample was also elevated in fine potassium suggesting a possible wood burning source also within this sector.

EROS site
The directions associated with the sample were predominantly in the sector from north-west to northeast corresponding to the Birmingham city centre and these constituents appear to be associated largely with non-tailpipe emissions from road traffic. The sample collected on April 8-11 associated with a predominant westerly wind component showed the lowest concentrations of nitrate, sulphate, ammonium, potassium and calcium suggesting relatively clean marine air on this westerly sector, although concentrations of the marine indicator elements chloride and sodium were not particularly elevated. An interesting feature of the EROS dataset is that the ranking of concentrations between the samples for fine fraction potassium was the same as that for sulphate.
Since the latter is well known as a pollutant subject to regional transport, this suggests that fine potassium, whose main source is believed to be woodsmoke (Harrison et al., 2012b), is also subject to long-range transport in a broadly similar manner. Other constituents which showed identical or similar rankings of concentration to one another (but not to sulphate and potassium) were vanadium, manganese, iron, antimony, barium and lead suggesting a common source for most elements, most probably in non-tailpipe emissions from road traffic.

Coarse Particle Increment at Port Talbot
The measured data are insufficient to conduct a mass closure on the aerosol composition, but it is nonetheless instructive to examine the differences in coarse particle mass and composition between PT and EROS (see Table S1). The difference in coarse particle (PM 2.5-10 ) mass between the sites is 6.1 µg m -3 . Marine aerosol (taken as Na + , Cland Mg 2+ ) accounts for just 1.0 µg m -3 of this.
Elements thought to be associated with steelworks (Fe and Ca 2+ ) together amount to 1.6 µg m -3 , which may approximately double to 3.2 µg m -3 when associated elements are added (e.g. Fe 2 O 3 and CaCO 3 ). This suggests that under the conditions sampled, steelworks-related emissions account for at least half of the elevation in PM 2.5-10 . The mass unaccounted for of ca 2 µg m -3 may be attributable to carbonaceous emissions, for example from the coke ovens, but data were not collected allowing this to be tested.

Analysis of Particle Size Spectra Collected with the Grimm Optical Spectrometer
The Grimm instrument has a response time of seconds, and in order to obtain statistically significant counts, the data were averaged over hourly periods and were then plotted as polar plots as a function of wind direction and wind speed. The data are collected in 14 separate size bins and polar plots for each size bin appear in Figure S6 in the Supplementary Information. Figure 4 shows the polar plots for four of the size classes which give a good representation of the overall behaviour observed.
In the case of PM 0.4-0.5 (see Figure 4)  It is instructive to examine the sources of PNC in response to meteorology on the days of episodes shown as April 26, May 4, 8 and 11 ( Figure S6). On April 26, between 12 pm and 2 am, the prevailing wind was between 200 o and ca 300 o where the ironmaking section of the steelworks was located. At the windspeed of 8 m s -1 , it appeared that PNC of all the particle size bins are favoured by the elevated windspeed due to effective transport from the point source (the steelworks) to receptor site. As the windspeed increases to about 10 m s -1 , there was a drastic reduction in particle number concentrations due to a dilution effect (Charron and Harrison, 2005). There was an increase in concentration of PM 0.3-0.4 when the windspeed reduced at around 6 am. Over this period, the prevailing wind direction has shifted to the south-easterly where the M4 road linking South Wales to London was located. Elevated particle concentrations of size bins 0.3 to 0.5 µm at 6 am may therefore be linked to traffic emission favoured by low windspeed and/or low dispersion.
On May 4, the major particle increment was in the smaller size bins and probably due to an elevation in secondary particle concentrations. Elevated particle numbers were favoured at a moderate windspeed 4-5 m s -1 . When the windspeed began to increase, PNC also reduced with a change of wind direction. The situation on May 8 and 11 looks similar to that of April 26, with the elevated values of PNC (sizes >0.65 µm) from the steelworks under a moderate windspeed.
However, with a change of wind direction and speed; particles < 0.65 µm increased between 6 and 10 am probably due to traffic emissions.

CONCLUSIONS
This study has revealed distribution patterns of size-segregated particles at a typical urban background (EROS) and an industrial setting (Port Talbot). Individual sample and average MOUDI data presented were diverse and varied between the two sites. EROS was dominated by fine particles while PT showed an elevated coarse particle concentration reflected in a higher ratio of PM 2.5 /PM 10 obtained at EROS. The influence of secondary aerosol was more evident at the urban background than the industrial site. Port Talbot showed elevated concentrations of marine (Na + , Mg 2+ and Cl -) and steelworks emissions (Fe, Mn and Ca). Higher concentrations of trace metals such as V, Al and Pb were observed at EROS. Strong similarities in mass size distributions were observed between Mn and Fe at PT (this was not so at EROS) suggesting a common emission source from the steel industry. Trace metals associated with brake wear (Cu, Sb, Ba) were clearly observed at both sites.
The analysis of size distributions from individual 72-hour sampling intervals confirmed the inferences derived from the average data, but allowed some appreciation of the episodicity of contributions both from the steelworks and secondary particles. Both showed the expected associations with, in the former case local winds from the direction of the steelworks, and easterly sector winds associated with regional transport in the latter. Although containing no chemical information, the polar plots of particle number size spectra revealed much source-related information. Local emissions, probably from road traffic dominated the smaller size bins (0.3-0.5 µm), while steelworks emissions dominated the range 0.5-15 µm, and for particles > 15 µm marine aerosol appeared dominant. Although there appeared to be contributions from more than a single source within the steelworks, the wind direction-dependence suggested the sinter plant and/or blast furnaces as the major contributor.