Characterisation of the traffic sources of PM through size-segregated sampling, sequential leaching and ICP analysis

Atmospheric Environment 42 (2008) 8161–8175

Contents lists available at ScienceDirect

Atmospheric Environment journal homepage: www.elsevier.com/locate/atmosenv

Characterisation of the traffic sources of PM through size-segregated sampling, sequential leaching and ICP analysis Silvia Canepari a, *, Cinzia Perrino b, Fabio Olivieri a, Maria Luisa Astolfi a a b

Department of Chemistry, University of Rome ‘‘La Sapienza’’, P. le A. Moro 5, 00185 Rome, Italy CNR Institute of Atmospheric Pollution, Via Salaria Km. 29300 Monterotondo St., 00015 Rome, Italy

a r t i c l e i n f o

a b s t r a c t

Article history: Received 21 March 2008 Received in revised form 28 July 2008 Accepted 31 July 2008

A study of the elemental composition and size distribution of atmospheric particulate matter and of its spatial and temporal variability has been conducted at two traffic sites and one urban background site in the area of Rome, Italy. Chemical analysis included the fractionation of 22 elements (Al, As, Ba, Ca, Cd, Co, Cr, Cu, Fe, Mg, Mn, Na, Ni, Pb, S, Sb, Si, Sn, Sr, Ti, Tl, V) into a water-extractable and a residual fraction. Size distribution analysis included measurements of aerosols in twelve size classes in the range 0.03–10 mm. The simultaneous determination of PM10 and PM2.5 at three sites during a 2-week study allowed the necessary evaluation of space and time concentration variations. The application of a chemical fractionation procedure to size-segregated samples proved to be a valuable approach for the characterisation of PM and for discriminating different emission sources. Extractable and residual fractions of the elements showed in fact different size distributions: for almost all elements the extractable fraction was mainly distributed in the fine particle size, while the residual fraction was in general predominant in the coarse size range. For some elements (As, Cd, Sb, Sn, V) the dimensional separation between the extractable fraction, almost quantitatively present in the fine mode particles, and the residual fraction, mainly distributed in the coarse mode particles, was almost quantitative. Under these conditions, the application of the chemical fractionation procedure to PM10 samples allows a clear distinction between contributes originating from fine and coarse particle emission sources. The results related to PM(10–2.5) and PM2.5 daily samples confirmed that chemical fractionation analysis increases the selectivity of most elements as source tracers. Extractable and residual fractions of As, Mg, Ni, Pb, S, Sn, Tl, Sb, Cd and V showed different time patterns and different spatial and size distributions, clearly indicating that the two chemical fractions are provided by different emission sources. The extractable fractions of As, Pb, Sn, S, Tl and V, in agreement to their dimensional distribution, were almost completely associated to PM2.5 and showed the same time pattern at all three sites, revealing the presence of spatially homogeneous sources. On the other hand, for most elements the relevant increase of PM10 elemental concentration from background stations to traffic urban stations can be attributed almost completely to the residual fraction of coarse particles. Taking into account temporal and size variability, the residual fractions of all these elements showed a very consistent co-variation, which indicates a prevailing common source of coarse particles; this has been identified in nontailpipe traffic source (road dust re-suspension, brake and tyre ware). Ó 2008 Elsevier Ltd. All rights reserved.

Keywords: Airborne particulate matter Size distribution Chemical fractionation Metals Road dust

* Corresponding author. Tel.: þ39 06 49913742; fax: þ39 06 4451571. E-mail address: [email protected] (S. Canepari). 1352-2310/$ – see front matter Ó 2008 Elsevier Ltd. All rights reserved. doi:10.1016/j.atmosenv.2008.07.052

8162

S. Canepari et al. / Atmospheric Environment 42 (2008) 8161–8175

1. Introduction The identification of the sources of atmospheric particles constitutes one of the main objectives of the chemical and physical characterisation of PM. Any cost-effective air pollution control policy, in fact, cannot be planned without a robust knowledge of the main contributors to atmospheric aerosol concentration. The high number of possible sources and the fast variations of their relative contribution to the atmospheric aerosol make this goal attainable only if the highest possible number of information about particle dimension, shape and chemical composition, as well as of characteristics of the sampling sites and meteorological situation during the observation periods, are combined. During the last few years, the scientific community has paid much attention to the study of atmospheric particle size distribution, as many processes occurring in the atmosphere (from climate forcing to cloud formation, as well as health effects) are dependent, besides chemical composition, also on particle size. Studies of the size distribution have been carried out for both PM mass and individual chemical components (ions, metals), and the existing relationship between particle formation pathways and particle dimensions has allowed the use of the size fractionation and chemical analysis, in combination with multivariate statistical techniques such as factor analysis, for the identification of some PM sources (e.g. local emission vs. long range transport) (Allen et al., 2001; Pakkanen et al., 2001, 2003; Singh et al., 2002; Harrison et al., 2003; Salma et al., 2005; Samara and Voutsa, 2005, among others). From the point of view of the chemical analysis, the availability of multi-parametric analytical techniques for elements and the relative simplicity of sampling and pretreatment procedures have allowed a wide use of the elemental analysis for source apportionment studies. On the other hand, elements are scarcely selective tracers, since most of them are produced by a variety of processes and sources. The need for a deeper differentiation in the chemical analysis of elements has led to the application of sequential leaching methods to particulate matter samples. The real behaviour of a metal in the environment, its fate after mixing of the aerosol with water, its bio-availability, its geochemical cycle etc. depend, in fact, on its specific metallic form. The development and application of chemical fractionation procedures for the analysis of trace metals in atmospheric particles have yielded many interesting results (Heal et al., 2005; Smichowski et al., 2005; Al-Masri et al., 2006; Birmili et al., 2006; Karthikeyan et al., 2006; Qureshi et al., 2006; Vasconcellos et al., 2007; Sato et al., 2008). In our laboratory, the recent application of a two-step chemical fractionation procedure to atmospheric particle samples has showed that the chemical fractionation noticeably increases the selectivity of some tracers of specific pollution sources (Canepari et al., 2006a). During the first stage of this procedure the collected particles are chemically fractionated for their solubility in a pH-buffered extracting solution; then the residue is mineralised. Among the most interesting results, the identification of magnesium as a responsive tracer of natural events (desert dust

and sea-spray): a selective increase of the only mineralised residual fraction of Mg is observed in the case of desert dust, while during sea-spray events also the extractable fraction of Mg increases. In addition, it has been shown that the acetate extractable and the mineralised residual fraction of a group of elements (Mn, Cu, Fe, Ni, Pb) may have different emission sources and that in urban areas the insoluble fractions of these metals have a common source. These last findings point to interesting perspectives in the identification of non-emission traffic contribution (re-suspension, brake and tyre abrasion) to PM levels (Weckwerth, 2001; Sternbeck et al., 2002; Gomez et al., 2005; Almeida et al., 2006). The coupling of size-segregated sampling techniques and chemical speciation analysis constitutes a step forward in the direction of a sound identification of PM sources in urban areas (Dodd et al., 1991). The purpose of work reported in this paper is the study of the size distribution, spatial variability and temporal variability of two chemical fractions of particulate matter having different elemental solubility. The data were collected during a field study carried out at three urban locations in the city of Rome by using a 13-stage low-pressure impactor and PM10 and PM2.5 samplers. The analyses were carried out by ICP-OES and ICP-MS. The interpretation of the data was carried out in the light of their space and time variability and of the meteorological conditions during the study. 2. Experimental and method 2.1. Sampling sites and sampling equipment Sampling was conducted in 2006, from April 12th to 26th, at three locations in the urban area of Rome: Montezemolo (MZ), a traffic site, Villa Ada (VA), an urban background location sited inside the main green park in Rome about 200 m from two highly travelled roads, Chemistry Department (CD), an urban site located inside the area of the University of Rome ‘‘La Sapienza’’, about 50 m from the nearest urban road but affected by local traffic (parking sites). The first two stations belongs to the air quality network of the local environmental protection agency (ARPALazio); here daily PM10 and PM2.5 samples were collected on Teflon filters by means of four automatic sequential samplers operating at the flow rate of 2.3 m3 h1 (SWAM 5a, FAI Instruments, Fontenuova, RM, Italy); the mass concentration was determined by the beta accumulation method. The sampled filters were kindly supplied for this study by ARPALazio. At the Chemistry Department, daily PM10 and PM2.5 samples were collected on 47 mm diameter PTFE membranes, 1 mm pore size (PALL Corporation, U.S.A.) by means of a dual channel sampling unit (HYDRA Dual Sampler, FAI Instruments, Fontenuova, RM, Italy) equipped with one PM10 and one PM2.5 sampling head, both operating at the flow rate of 2.3 m3 h1. At the same site, a 13stage low-pressure impactor (DLPI, DEKATI Ltd., Tampere, Finland) was run during the whole period (15 days). The instrument operates at the flow rate of 10 l min1 and at the pressure of 100 mbar under the last impactor stage. The

S. Canepari et al. / Atmospheric Environment 42 (2008) 8161–8175

nominal values for the equivalent aerodynamic 50% cut-off diameters of the impactor stages are: 10, 6.8, 4.4, 2.5, 1.6, 1.0, 0.65, 0.40, 0.26, 0.17, 0.108, 0.060 and 0.030 mm. PTFE membranes, 25 mm diameter, (ALBET, Barcelona, Spain) were used as substrates on the collection plates of the impactor. Collection substrates of impactors are generally greased to reduce the bounce and blow-off of the particles, but for this study it was decided not to add any grease in order to improve the analytical quality of the analyses. PM mass concentration at the CD site (PM10, PM2.5 and stages of the impactor) was determined by gravimetry. All PTFE filters were conditioned for 48 h at 50% R.H. before weighing. They were weighed before and after sampling using a 1 mg analytical balance (Gibertini Elettronica, Novate Milanese, MI, Italy). 2.2. Meteorological conditions Meteorological conditions during the period of the study were characterised in different ways: – measuring weather parameters including temperature, relative humidity, atmospheric pressure, wind velocity, wind direction and global sun radiation; – evaluating the backward trajectories of the air masses as calculated by the HYSPLIT (HYbrid Single-Particle Lagrangian Integrated Trajectory) model run by NOAA (http://www.arl.noaa.gov) and by the FLEXTRA trajectory model run by NILU (http://www.nilu.no/ trajectories); – evaluating the results of the Dust Regional Atmospheric Model (DREAM), which predicts the atmospheric life cycle of the eroded desert dust (www.bsc.es/ projects/earthscience/DREAM); – estimating the dilution properties of the lower atmosphere through natural radioactivity measurement. The latter technique is based on the monitoring of natural radioactivity due to Radon short-lived decay products. Radon is supplied by the Earth’s crust and its emission flow into the atmosphere for a given geographical location and for the time scale of our observations (weeks) can be considered constant; thus the air concentration of Radon and 222Radon short-lived daughters can be assumed to depend only on the dilution factor (Porstendorfer et al., 1991; Shweikani et al., 1995). The monitoring of natural radioactivity due to Radon progeny attached to atmospheric particles yields a reliable picture of the dilution properties of the lower atmosphere; in many studies it has been used to identify periods characterised by atmospheric stability, which favours the build-up of pollution, and periods characterised by advection, which favours the dispersion of locally emitted pollutants (Perrino et al., 2001, 2008; Sesana et al., 2003; Vecchi et al., 2007). Weather parameters and natural radioactivity were measured at the peri-urban CNR station of Montelibretti, 25 km from downtown Rome. Hourly average natural radioactivity was measured by means of an automated stability monitor (PBL Mixing Monitor, FAI Instruments, Fontenuova, RM, Italy). 2.3. Analysis All the samples were analysed by following the twostep procedure described in Canepari et al., 2006b,c. Briefly,

8163

the sampled PTFE membranes were extracted in ultrasonic bath in acetate buffer solution at pH 4.5, then the extract were filtered on a cellulose nitrate filter and analysed (extractable fraction); after this first step, both the cellulose nitrate filter and the original PTFE filter with the residue from the first stage were acid-digested in microwave oven and analysed (mineralised residual fraction). Note that the acid-digestion with HNO3/H2O2, which allows a good recovery of most elements (Canepari et al., 2006b), leads to incomplete recovery for some elements, such as Al, Si and Cr. For these elements, the reported results represent only the amounts recovered under our specific operational conditions. However, due to the good repeatability of the recovery, Si and Cr results can be considered for comparison of samples. For Al (in both extractable and residual fractions) and Ca (in residual fraction), instead, only cascade impactor results are reliable, as the polyethylene ring supporting PM10 and PM2.5 membranes give rise to high blank values. Conversely, Ni results are reliable only in PM10 and PM2.5 samples, probably because of some release of Ni from the stainless steel body of the impactor. For this work the analyses were carried out by ICP-OES (Varian Vista MPX CCD Simultaneous ICP–OES, with US nebulizer U 5000 ATþ, Cetac Technologies) and, after dilution, by ICP-MS (Varian 810). Preliminary studies have shown that ICP-MS results are more reliable for As, Co, Ni, Sb, Sn and Tl, whose ambient concentration are close to the quantification limits of ICP-OES, while the latter instrument yields more reliable results for Al, Ca, Cr, Fe, S and Si. For the other elements (Ba, Cd, Cu, Mg, Mn, Na, Pb, Sr, Ti, V) the results of the analyses carried out by the two techniques are comparable, with differences below 4%. 3. Results and discussion 3.1. Meteorological conditions and back-trajectories A variety of meteorological conditions alternated during the period of the study. Atmospheric pressure increased from the beginning of the examined period until April 15; then it decreased to reach a minimum on April 18 and then increased again. High wind intensity was recorded on April 13 and 18–19, while the sea-breeze conditions (afternoon winds from SE) dominated during the last week of the period (Fig. 1a). Temperature and solar radiation data (Fig. 1b,c) show that April 17 and 18 were cloudy days with a warm night in between. This first analysis is well complemented by the study of natural radioactivity time pattern. The data (Fig. 1d) indicate advection during the first two days of the study (low values of natural radioactivity) then a 3-day period of night-time stability and daytime convective mixing (14th, 15th and, at a less extent, 16th), then one day characterised by high values of natural radioactivity during both the night and the daytime hours, which indicates atmospheric stability and poor pollutant dilution. These conditions (April 17) favour the build-up of particulate pollution. After this critical day, the data show a 2-day advection period (18th and 19th) and then a period of high pressure, with stability during the night and convective mixing during the daytime hours. According to this comprehensive picture of

8164

S. Canepari et al. / Atmospheric Environment 42 (2008) 8161–8175

b

SOLAR RADIATION

c

TEMPERATURE

30 27 24 21 18 15 12 9 6 3 0

1000 900 800 700 600 500 400 300 200 100 0

W m-2

WIND INTENSITY

8 7 6 5 4 3 2 1 0

C°

m s-1

a

12 13 14 15 16 17 18 19 20 21 22 23 24 25 26

12 13 14 15 16 17 18 19 20 21 22 23 24 25 26

d

12 13 14 15 16 17 18 19 20 21 22 23 24 25 26

NATURAL RADIOACTIVITY 1800

counts / minute

1500

1200

900

600

300

0

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

APRIL 2006 Fig. 1. Time patterns of meteorological parameters during the study: a: wind intensity, b: atmospheric temperature, c: solar radiation, d: dilution properties of the lower atmosphere as traced by natural radioactivity.

the dilution properties of the lower atmosphere, a fast increase of PM concentration is expected on April 17 and a slower increase can be envisaged for the period 20th– 25th. Low PM values are instead expected for April 12–13 and 18–19. The results of both the NOAA HYSPLIT and the NILU FLEXTRA models indicate that the air masses reaching Rome on April 16 and 17 had a marine origin (see the comparison between the HYSPLIT back-trajectories of 16th and 17th, reported in Fig. 2a, and those of the following three days, Fig. 2b). For the rest of the periods, the models indicate air masses roughly from North during April 12–15 and from North-East during April 18–25. The DREAM model forecasted the intrusion of African dust (air from South) during April 15–18, with a maximum intensity on 16th; a second episode was forecasted to start on April 25–26. 3.2. Consistency of impactor results and PM10, PM2.5 results The soundness of the sampling and analysis procedures and the overall quality of the results at the CD site were verified by comparing, for overall PM and for each element, the average concentration of the 15 daily PM10 samples with the concentration calculated by adding the amounts on the 13 impactor stages, and the average concentration of the 15 daily PM2.5 samples with the concentration obtained by adding the amounts on the lower 10 impactor

stages. For the evaluation of the results, reported in Table 1, it must be taken into account that, for each element, accuracy and uncertainty of the results on PM10–PM2.5 filters and on impactor filters may be different because of different operative conditions and collecting media. Moreover, for many elements (e.g. Cr, Ti in the extractable fraction, As, Tl in the residual fraction) the values detected on the lower impactor stages were below the quantification limits and were discarded, causing an underestimation of the results. The data in Table 1 show that for PM10 the agreement of the two data series is generally satisfactory, and that the differences are in agreement with the uncertainty and error propagation typical of these analytical determinations (Canepari et al., 2006b). As a consequence, the possibility of a deviation from the theoretical flow rate of the impactor, which would lead to a variation in the cut size of the impactor stages, can be ruled out. For PM2.5, instead, the values are generally in agreement for elements which are mainly in the size fraction below 2.5 mm (refer to Figs. 4 and 5); for elements which are mainly distributed in the coarse fraction, instead, the amounts in the impactor stages are higher than the single PM2.5 values. These results could be explained by particle bounce. It is known that newly collected particles can rebound on the impaction surface or on previously collected particles, being carried away by the air stream and collected on subsequent stages. The extent

S. Canepari et al. / Atmospheric Environment 42 (2008) 8161–8175

8165

Fig. 2. Back-trajectories of the air masses reaching the area of Rome according to the NOAA HYSPLIT model: a: February 16th and 17th; b: February 18th, 19th and 20th.

of this artefact depends on filter type, particle loading and use of grease or oil to coat the impaction substrates (Marjama¨ki et al., 2000). In our study, the absence of grease on the impaction plates, chosen in order to improve the analytical quality of the results, and the use of PTFE filters may have shifted the mode diameter downward, causing an overestimation of the mass amount on the smaller impaction plates. In spite of this artefact, the results reported in Table 1 show that the size distributions can be considered reliable for the following qualitative evaluation of the results.

3.3. Size distribution of the extractable and residual fraction 3.3.1. Cascade impactor Fig. 3 shows the size distribution (13 dimensional classes) of the total particulate mass collected during the fifteen days of the study. The data indicate a trimodal distribution, with half of the mass accumulated in a large coarse mode having a peak between 4.4 and 6.8 mm. About 40% of the mass constitutes a second mode in the accumulation range, with peak between 0.65 and 1 mm and only 10% of the mass constitutes the third, smaller mode, with peak between 0.108 and 0.17 mm. This distribution is quite typical of urban sites, where re-suspended and frictionallygenerated dust in the coarse mode, as well as local combustion products in the ultra-fine mode, adds to

particles in the accumulation mode, which are generally favoured by least efficient removal processes. The results from the elemental analysis of the sizesegregated samples obtained with the cascade impactor are reported in Figs. 4 and 5. For these analyses, particles retained on the last two stages of the impactor, having cut size of 0.060 and 0.030 mm, have been analysed together in order to improve the reliability of the results. The data show that the solubility distribution is typical of the each element, in agreement with previous results (Heal et al., 2005; Birmili et al., 2006). For example, Al (96%), Cr (96%), Fe (96%), Sn (97%), Si (94%) and Ti (99%) are mostly in the form of insoluble species and thus almost totally in the residual fractions, while their extractable fractions are practically negligible. On the other side, elements such as Na (90%), S (89%) and Tl (89%) are almost exclusively in the extractable fraction. The remaining elements are distributed between the two chemical fractions, with characteristic solubility distribution: Co (64%) and Pb (67%) are mostly in the residual fraction, while As (64%), Ca (62%), Cd (64%), Sr (66%) and V (62%) are predominantly in the extractable fraction. A one-to-one distribution is observed for Ba, Cu, Mg, Mn and Sb (all of them in the range 42–58%). The size distribution of the total elemental content (sum of the extractable and residual fraction) is also characteristic of each element and the results are in agreement with the findings of previous studies (Pakkanen et al., 2003;

8166

S. Canepari et al. / Atmospheric Environment 42 (2008) 8161–8175

Table 1 Comparison of element concentrations in both the extractable and the residual fraction calculated as average of the 15 daily PM10 or PM2.5 samples and as sum of the lower 12 (PM10) or 9 (PM2.5) stages of the impactor Extractable fraction (ng m3)

Residual fraction (ng m3) PM2.5

PM10

Al As Ba Ca Cd Co Cr Cu Fe Mg Mn Na Ni Pb S Sb Si Sn Sr Ti Tl

PM10

PM2.5

Daily

Impactor

Daily

Impactor

Daily

Impactor

Daily

Impactor

– 0.28 7.25 1007 0.26 0.10 0.37 22.7 27.0 114 5.38 799 1.62 2.21 994 3.96 34.6 0.13 4.11 0.15 0.10

31.8 0.28 7.71 1000 0.21 0.09 0.32 21.3 23.6 111 5.32 810 1.50 3.20 863 4.11 37.3 0.14 4.07 0.10 0.09

– 0.28 1.82 221 0.18 0.06 0.22 5.55 13.2 18.1 1.64 139 1.31 2.73 864 3.18 6.74 0.13 0.85 0.03 0.07

16.7 0.27 2.47 230 0.19 0.05 0.24 7.21 16.6 24.4 2.13 152 1.28 2.97 798 3.13 6.93 0.14 1.00 <0.01 0.07

– 0.16 10.10 – 0.13 0.15 7.01 24.2 661 75.1 6.72 81.7 2.37 8.95 103 5.03 573 4.57 1.86 13.1 0.01

705 0.16 9.07 622 0.12 0.16 7.98 23.2 632 89.2 6.62 85.3 – 6.40 108 4.79 560 4.22 2.06 12.8 0.01

– 0.03 1.62 – 0.06 0.07 3.41 5.26 108 22.1 1.63 48.8 1.52 3.52 34.4 1.53 99.7 1.45 0.57 2.33 <0.01

342 0.04 1.77 162 0.06 0.06 3.63 6.80 138 30.5 2.20 46.4 – 3.64 49.4 1.78 109 1.75 0.59 3.03 <0.01

Total mass concentration: PM10 daily: 32.0 mg m3; PM10 impactor: 27.9 mg m3; PM2.5 daily: 13.6 mg m3; PM2.5 impactor: 15.8 mg m3.

Salma et al., 2005). Elements of natural origin, such as Ba, Ca, Fe, Mg, Mn, Na, Sn, Si, Sr, Ti, are mostly in the coarse fraction, with more than 40% of the total mass in the two impactor stages corresponding to the size range 4.4–10 mm (more than 50% for Ba and Fe, more than 60% for Ca, Si, Sr and Ti). Elements of anthropogenic or secondary origin, instead, show a relevant contribution in the fine fraction: more than 50% of the total mass of As, Cd, Pb, S, Tl and V is in the size range below 1.0 mm. Among these elements, it is interesting to note that for S and Tl only about 10% of the total mass is in the coarse size range, while As, Cd, Pb, Sb and V (Fig. 5) show a bimodal distribution, with one peak in the accumulation mode, between 0.26 and 1 mm, and one peak in the coarse mode, between 4.4 and 10 mm. A most interesting result of this work is the different dimensional distribution of the extractable and residual

5

mg/m3

4 3 2

fractions: many elements show remarkable differences between the size distributions of the two chemical fractions, reflecting a different solubility of the chemical species of the same element which derive from different sources. In particular (Fig. 5), As and V show an almost quantitative separation between the extractable fraction, which is mostly in the size range below 2.5 mm (97%), and the residual fraction, which is mainly (more than 65%) in the coarse size range. For these elements, the direct application of the chemical fractionation procedure to PM10 samples would allow a good discrimination of the element sources: without performing any size classification, the extractable amount can be attributed to the fine fraction and the residual fraction to the coarse fraction, giving precious indications about the emission sources. A similar pattern is shown by Cd and Sb, which also are predominantly in the fine size range for the extractable fraction (93% and 76%, respectively) and mostly in the coarse range for the residual fraction (49% and 62%). Also Sn show a quite different size distribution of the extractable and residual fractions, although the extractable fraction of this element is only 3% of the total mass, In the case of Pb, instead, the extractable and the residual fractions show a similar profile in the fine range, while the mass amount in the coarse range is only due to the residual fraction.

1

10

6.8

4.4

2.5

1.6

1.0

0.65

0.40

0.26

0.17

0.060

0.108

0

mm Fig. 3. Size distribution of the total particulate mass collected by the cascade impactor.

3.3.2. Fine and coarse fractions The results of the daily samplings of the fine (PM2.5) and coarse (PM10 – PM2.5) fractions of particulate matter are shown in Table 2. The data are reported as average values of the 15-day sampling period for the extractable and the residual fractions of elements at the three sampling sites (MZ, VA and CD).

S. Canepari et al. / Atmospheric Environment 42 (2008) 8161–8175

Al 3 dm/dlogDp (ng/m )

1200 1000 800

Ba 25

TOTAL extractable residual

20

2500

15

2000 1500

10 400

1000 5

200

Co dm/dlogDp (ng/m3)

500

0

0.4

0.3

0

Cu

Cr 15

60

12

50 40

9 0.2

30 6 20

0.1

3

10

0

0.0

0

Mg

Fe 3 dm/dlogDp (ng/m )

Ca 3000

600

0

Mn

1200

300

1000

250

800

200

600

150

400

100

200

50

4

0

0

0

20 16 12 8

Na dm/dlogDp (ng/m3)

8167

S

1200

1500

1000

1200

1200 1000

800 900

800

600 600

600

400

400

300

200

200 0

0

Sr dm/dlogDp (ng/m3)

Si 1400

0

Ti

Tl

12

30

10

25

8

20

0.06

6

15

0.05 0.04 0.03

4

10

2

5

0

0

0.09 0.08 0.07

0.02 0.01 0.00

Fig. 4. Size distribution of the extractable (grey lines, squares) and residual fractions (dark lines, triangles) and of the total amounts (bars) of 15 elements.

In agreement with the results of the cascade impactor, for almost all elements the extractable fraction predominates in the fine particle size while the residual fraction is predominant in the coarse size range. The data also show that the fine particle elemental concentrations at the three sampling sites are quite similar, in agreement with their lower removal rate and consequent more homogeneous space distribution, while the elemental concentration of coarse PM is much more variable among the different sites. On average, the relative standard deviation between the three sites is of the order of 30% for elements in PM2.5 and higher than 100% for elements in the coarse fraction.

PM10 concentration at the urban background site of VA is, on average, one third lower than at the traffic site of MZ. Also the total elemental concentration is about 30% lower. This difference can be attributed almost completely to the residual fraction of coarse particles. Taking the case of Fe, its average concentration in PM10 was 1027 ng m3 at MZ and 393 ng m3 at VA. The difference, 634 ng m3, is consistent with the value of the difference between the two residual fractions in coarse PM, 583 ng m3. The differences between VA and the site of CD, inside the urban area but in a parking lot about 50 m from the nearest road, are smaller than in the case of VA and MZ. The

8168

S. Canepari et al. / Atmospheric Environment 42 (2008) 8161–8175

As

V 3.0

TOTAL extractable residual

0.3

2.5 2.0

0.2

1.5 1.0

0.1 0.5

0.

Sb

8

10

6.

5

4. 4

6

2.

Cd 0.35 0.30

6

0.25 0.20

4

0.15 0.10

2

0.05

4 4.

6.

4. 4

6.

8 10

2. 5

6 1.

2. 5

6 1.

0 0 1.

65

m

1.

0.

0.

0.

8 10

8 10

4

6.

4.

6

0

2. 5

1.

1.

0. 4 0. 65

0. 17 0. 26

8 10 0.

4

0.00

0

0. 17 0. 26

m Pb

Sn 6

0.30

5

0.25

4

0.20

3

0.15

2

0.10

1

0.05

0

0.00

8

6

4

m

10

0. 65

0. 4

0. 17 0. 26

10

8

0 0.

10

8

4

6.

4.

0. 10 8 0. 17 0. 26 0. 4 0. 65 1. 0 1. 6 2. 5

2

8

dm/dlogDp (ng/m3)

0

m

8

dm/dlogDp (ng/m3)

1.

10 8 0. 17 0. 26

8

10

6.

8 0. 17 0. 26 0. 4 0. 65 1. 0 1. 6 2. 5 4. 4

10 0.

m

1.

0.0

0.0

0. 4 0. 65

dm/dlogDp (ng/m3)

0.4

m

Fig. 5. Size distribution of the extractable (grey lines, squares) and residual fractions (dark lines, triangles) and of the total amounts (bars) of 6 most interesting elements.

difference for Fe in PM10 is 339 ng m3. Again, however, this value is consistent with the value of the difference between the two residual fractions in coarse PM, 302 ng m3. A similar behaviour, also shown by Ba, Ca, Cu, Mg, Mn, Si, Sr and Ti, indicates that the differences in the elemental content of PM between traffic and urban background stations are due to local contributions from traffic sources, pointing out that most of the elemental pollution in urban areas can be attributed to dust re-suspension (Sternbeck et al., 2002; Almeida et al., 2006). The case of lead constitutes a particularly interesting anomaly. The data in Table 2 show that the extractable

fraction is higher in PM2.5 than in PM10, a behaviour that was already recorded during previous monitoring campaigns. The source of this bias must be necessarily searched in the extraction procedure, as no sampling artefact may reasonably lead to this result. A possible reason could be the presence in the coarse fraction of an ion 3 2  (e.g. CO2 3 , S , PO4 , Cl ) able to form slightly soluble salts 2þ with Pb . In this case, soluble Pb2þ, which is mostly in the fine fraction, in the presence of coarse particle could precipitate, causing the observed reduction of the extractable Pb in PM10. The hypothesis is confirmed looking at the data in Table 1. In PM2.5, the amount of lead in the impactor

S. Canepari et al. / Atmospheric Environment 42 (2008) 8161–8175

8169

Table 2 Average of the daily concentration of the extractable and residual fractions in fine and coarse particles at the three sampling stations PM2.5 (ng m3)

PM10–2.5 (ng m3)

MZ

CD

VA

MZ

CD

VA

Ext.

Res.

Ext.

Res.

Ext.

Res.

Ext.

Res.

Ext.

Res.

Ext.

Res.

Al As Ba Ca Cd Co Cr Cu Fe Mg Mn Na Ni Pb S Sb Si Sn Sr Ti Tl V

30.1 0.28 2.50 197 0.17 0.06 0.34 5.54 18.7 21.8 1.76 185 1.54 2.73 966 2.84 7.19 0.16 0.81 0.04 0.06 2.82

– 0.19 1.68 – 0.05 0.08 3.98 6.95 132 16.0 1.38 57.2 1.67 4.13 39.6 1.56 91.1 1.48 0.43 2.86 0.02 0.54

33.5 0.28 1.82 221 0.18 0.06 0.22 5.55 13.2 18.1 1.64 139 1.31 2.73 864 3.18 6.74 0.13 0.85 0.03 0.08 2.01

– 0.03 1.62 – 0.06 0.07 3.41 5.26 108 22.1 1.63 48.8 1.52 3.52 34.4 1.53 99.7 1.45 0.57 2.33 0.01 0.43

25.8 0.31 1.73 185 0.26 0.07 0.31 4.78 16.2 18.6 1.57 169 1.40 3.30 1006 9.13 5.67 0.16 0.70 0.03 0.09 2.60

– 0.18 1.21 – 0.07 0.10 3.30 4.55 91 20.9 1.34 56.8 1.73 4.41 39.2 2.44 71.8 1.00 0.61 1.83 0.02 0.50

25.4 <0.01 7.28 879 0.02 0.05 0.14 24.3 22.2 89.5 4.21 568 0.378 0.97 146 0.93 33.1 0.03 3.36 0.15 <0.01 0.15

– 0.212 12.80 – 0.09 0.12 7.12 35.6 855 72.0 7.15 28.1 1.31 5.91 77.8 6.44 605 5.98 1.90 11.6 0.01 1.44

26.3 <0.01 5.43 786 0.08 0.04 0.15 17.1 13.8 95.9 3.74 660 0.312 0.52 130 0.78 27.9 0.00 3.26 0.11 0.03 0.17

– 0.128 8.48 – 0.07 0.08 3.60 18.9 553 53.0 5.09 32.9 0.853 5.43 68.6 3.50 473 3.12 1.29 10.8 0.01 1.00

14.5 0.026 3.42 591 0.01 0.04 0.13 7.58 14.8 73.0 2.98 470 0.179 0.91 115 0.14 17.3 0.03 2.29 0.07 <0.01 0.13

– 0.101 3.74 – 0.02 0.07 1.90 7.87 272 19.3 2.46 7.8 0.360 1.98 5.5 1.54 364 1.55 0.59 7.76 0.01 0.69

Mass mg m3

20.4

13.6

13.5

15.7

is close to the amount in the daily filters, both in the extractable and the residual fractions. Conversely, in PM10, extractable Pb in the impactor is higher than in daily samples, while the opposite holds for the residual fraction. In other words, this artefact is observable only when the extracting solution contains coarse and fine particles together.

18.4

9.8

increase of PM values occurred on April 17. The lowest values were always recorded at the urban background site VA, while values at MZ and CD were comparable. Further information can be obtained by studying the temporal pattern of single elements in the extractable and the residual fraction of PM10 and PM2.5 collected at the three sites. We report in Fig. 7 the temporal patterns of the extractable (upper figure) and residual (lower figure) fractions of vanadium. As far as the soluble fraction is concerned, it is manifest that sample of the two size fractions collected at the three locations show a pattern of co-variation. This observation confirms that V is almost quantitatively in the fine fraction, and shows that the sources of this elemental fraction are spatially

3.4. Temporal pattern Fig. 6 shows the time pattern of daily PM10 mass concentration at the three sites. As expected from the meteorological situation described above, the lowest concentrations were recorded on April 13 and 18–19 and an

50 MZ CD VA

45 40

mg/m3

35 30 25 20 15 10 5 0 12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

APRIL 2006 Fig. 6. Time pattern of daily PM10 mass concentration at the three sites: VA (urban background), CD and MZ (traffic).

8170

S. Canepari et al. / Atmospheric Environment 42 (2008) 8161–8175

EXTRACTABLE VANADIUM 9 8

VA PM2.5 VA PM10

7

ng m-3

CD PM2.5 6

CD PM10

5

MZ PM2.5 MZ PM10

4 3 2 1 0 12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

APRIL 2006 RESIDUAL VANADIUM 4,5 4,0 3,5

ng m-3

3,0 2,5 2,0 1,5 1,0 0,5 0,0 12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

APRIL 2006 Fig. 7. Temporal patterns of the extractable (upper figure) and residual (lower figure) fractions of vanadium in two size fractions (PM10, PM2.5) at three sites (VA, CD, MZ).

homogeneous. Conversely, the graphs of the residual fractions (lower figure) show that most of this fraction is in the coarse size (PM10 [ PM2.5) and that the temporal patterns are not consistent, with clear differences among sites. This finding suggests a likely multiplicity of local sources contributing to vanadium in the residual fraction. The extractable fractions of Cd and Sb, reported in Fig. 8 (upper and lower graph, respectively), are, again, almost quantitatively present in the fine fraction, but show concentration peaks at specific locations. For example, on April 14th Cd concentration in both the fine and the coarse PM fraction sharply increased at CD; the increase was also recorded at VA, in the only fine fraction (Cd contents in PM10 and in PM2.5 coincide). On April 17th, instead, Cd increase was observed mainly at VA, again in the fine PM fraction, and, at a less extent, in the fine and coarse fraction at CD. On April 20th a ten-fold increase of Sb in the only fine fraction and at the only VA station was detected (Fig. 8, lower graph). These results show the action of specific sources of these elements, which may contribute relevantly to its average concentration in the urban area. For Cd this is

also a relevant point for the observance of target values stated in the EU Directive 2004/107/CE. As far as the residual fractions of Cd and Sb are concerned, as in the case of V, they are mostly in the coarse size and their time patterns at the three sites are inconsistent. In summary, the extractable and the residual fractions of V, Sb and Cd are completely different in temporal pattern, spatial and size distribution, and thus their sources must also be different. Similar behaviour is also observed for As, Ni, Pb, Sn, S, and Tl. Finally, Fig. 9 shows that at each station, the temporal patterns of Fe (upper graph) and Mn (lower graph) residual fractions are very similar. Alike patterns are also shown by other elements in the residual fractions. A general picture of correlation between the extractable and residual fractions of elements is reported in Table 3. The correlation was calculated on a data set that includes PM10 and PM2.5 at all three stations; the anomalous episodes reported in Fig. 8 were deleted. Very good correlations, with Pearson coefficient in most cases higher than 0.75, are shown in the residual fraction

S. Canepari et al. / Atmospheric Environment 42 (2008) 8161–8175

8171

EXTRACTABLE CADMIUM 1,4 VA PM2.5 VA PM10 CD PM2.5 CD PM10 MZ PM2.5 MZ PM10

1,2

ng m-3

1,0 0,8 0,6 0,4 0,2 0,0 12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

22

23

24

25

26

APRIL 2006 EXTRACTABLE ANTIMONY 90 80 70

ng m-3

60 50 40 30 20 10 0 12

13

14

15

16

17

18

19

20

21

APRIL 2006 Fig. 8. Temporal patterns of the extractable fractions of Cd (upper figure) and Sb (lower figure), in two size fractions (PM10, PM2.5) at three sites (VA, CD, MZ).

(lower part of Table 3) by most of the elements. This covariation of the residual fractions, which takes into account both temporal and size variability, indicates a prevailing, common particle source. As it was shown that the residual fraction is mostly in the coarse size, we can deduce that this common source produces mechanically generated, coarse particles. It is worth noting that the covariance was observed not only for anthropic elements (e.g. Cu, Fe, Sb), which are known to be generated by frictional processes (break and tyre wear), but also for elements which are generally of natural origin (e.g. Mg, Si, Ti). This is clearly suggestive of road dust re-suspension as a relevant source. These two major contributes (frictional processes and natural dust re-suspension), both due to vehicular traffic, are almost totally responsible for the concentration increase observed when comparing PM10 elemental concentrations at the traffic stations and at the urban background station. Pearson coefficients related to the extractable fraction (upper part of Table 3) show, in general, a lower co-variation. Coefficients higher than 0.75 are obtained only by a group of elements (Ba, Ca, Cu, Fe, Mn, Si, Sr), all characterised (see Fig. 4) by a very similar dimensional distribution of the extractable and residual fractions. Therefore, these elements seem to be contributed, also in the

extractable fraction, by non-tailpipe traffic sources. A good correlation is observed also for soluble Na and Mg (R2 ¼ 0.97), due to the sea-spray events occurred on April 16 and 17, for Pb and Cd, whose origin will need further investigation, and for Ni, V and S, whose soluble species probably derive from combustive processes. 3.5. Break and tyre wear, road dust With the purpose to evaluate qualitatively the possible role of non-tailpipe traffic sources, four types of samples were gathered: dust collected on the road surface in the vicinity of the CD station (three sampling points), dust obtained mechanically from brake pads (three types), from tyres (three types) and from asphalt (three samples). All samples were sieved to obtain particles lower than 63 mm in size, and analysed for their extractable and residual elemental fractions. The average results are summarised, in logarithmic scale, in Fig. 10. The three samples of road dust, tyres and asphalt showed good repeatability (RSD lower than 40%). Brake pads of the three brands, instead, showed very different composition, particularly as far as Cu and Sb are concerned, with a quantitative ratio ranging from 2 to 2000. It is worth noting that the ratio Cu/Sb is often

8172

S. Canepari et al. / Atmospheric Environment 42 (2008) 8161–8175

RESIDUAL IRON 1800 VA PM2.5

1600

VA PM10

1400

CD PM2.5

1200

CD PM10

ng m-3

MZ PM2.5

1000

MZ PM10

800 600 400 200 0

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

APRIL 2006 RESIDUAL MANGANESE 16 14 12

ng m-3

10 8 6 4 2 0

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

APRIL 2006 Fig. 9. Temporal patterns of the residual fractions Fe (upper graph) and Mn (lower graph) in two size fractions (PM10, PM2.5) at three sites (VA, CD, MZ).

reported in the literature as a characteristic index of brake wear, even though the values indicated by different authors in different urban environments range from about 4 to about 12 (Weckwerth, 2001; Sternbeck et al., 2002; Pakkanen et al., 2003; Furuta et al., 2005; Gomez et al., 2005; Maenhaut et al., 2005; Salma et al., 2005). This ratio has to be taken with caution, as it seems to be dependent on the brake brand most frequently used in the area. Our data show that in the area of Rome the value of Cu/Sb is on average 4.9  1.5 and that it ranges from 3.9  1.7 at the urban background station to higher values at the two traffic stations (4.7  1.2 at CD and 5.8  0.9 at MZ). This difference is justified by the fact that Sb content in the fine fraction, presumably not provided by brake wear, is not negligible and constitutes a higher percentage at the urban background station than at the traffic stations. Referring more selectively to the contribution of brake wear, and thus considering only the mineralised fraction and the coarse dimensional range (PM10–PM2.5), we obtain an average value of 5.2  0.7 and much more similar values at the three sites: 5.3  0.5 at VA, 5.2  0.7 at CD and 5.2  0.6 at MZ. Although the data in Fig. 10 come from a small number of samples and are to be considered as only qualitative, they suggest some interesting remarks. Road dust contains

all elements at quite high concentration. Asphalt and tyres particularly contribute to road dust with some elements: As, Ti and V in asphalt and Si and Tl in tyres. Among the three materials, brake pads show the highest elemental concentrations and a profile most similar to road dust. It follows that a major contribution to the elemental concentration in road dust come from this source, particularly as far as Ba, Fe, Sb and Sr are concerned. It is worth noting that particles produced by brake wear, of high density, are much smaller than expected on the basis of their collection stage and thus of their aerodynamic diameter. This observation, confirmed by preliminary Scanning Electron Microscope images, suggests that these particles may reach deep into our lower respiratory system and thus their toxicity has to be considered with much more attention (Uexkull et al., 2005). The solubility distribution shows that in all the examined materials most of the elements are found in the residual fraction. An interesting exception is constituted by Tl, whose residual fraction seems to be contributed mainly by tyres, while the extractable fraction comes mainly from asphalt. Also interesting is the case of sulphur, which is present as insoluble species at relevant concentrations in both tyres and brakes. In atmospheric

Table 3 Pearson coefficients for element concentrations in the extractable (upper values are in italic) and the residual fraction (lower values) of daily PM10 and PM2.5 samples at the three sampling sites As

Ba

Ca

Cd

Co

Cr

Cu

Fe

Mg

Mn

Na

Ni

Pb

S

Sb

Si

Sn

Sr

Ti

Tl

V

1 0.68 – 0.90 0.76 0.84 0.82 0.84 0.83 0.89 0.66 0.79 0.84 0.78 0.85 0.86 0.84 0.82 0.85 0.73 0.82

0.02 1 – 0.73 0.59 0.68 0.78 0.78 0.69 0.76 0.77 0.55 0.64 0.67 0.78 0.75 0.76 0.75 0.69 0.45 0.64

0.09 0.91 1 – – – – – – – – – – – – – – – – – –

0.03 0.19 0.22 1 0.74 0.82 0.88 0.89 0.83 0.92 0.73 0.76 0.96 0.85 0.91 0.86 0.88 0.88 0.84 0.70 0.86

0.02 0.73 0.68 0.36 1 0.69 0.60 0.64 0.63 0.70 0.59 0.71 0.67 0.62 0.61 0.68 0.61 0.67 0.66 0.66 0.67

0.12 0.56 0.41 0.12 0.61 1 0.90 0.86 0.75 0.90 0.55 0.82 0.71 0.75 0.85 0.76 0.90 0.77 0.76 0.55 0.75

0.02 0.97 0.88 0.20 0.74 0.54 1 0.95 0.81 0.96 0.65 0.74 0.81 0.81 0.97 0.84 0.97 0.86 0.83 0.50 0.79

0.06 0.75 0.66 0.29 0.87 0.64 0.79 1 0.89 0.97 0.69 0.71 0.82 0.82 0.94 0.92 0.95 0.90 0.90 0.49 0.86

0.16 0.62 0.70 0.20 0.40 0.13 0.63 0.46 1 0.91 0.78 0.65 0.81 0.83 0.82 0.96 0.81 0.94 0.96 0.55 0.84

0.01 0.94 0.94 0.21 0.78 0.60 0.92 0.82 0.62 1 0.72 0.79 0.85 0.86 0.95 0.92 0.95 0.93 0.91 0.59 0.85

0.17 0.44 0.53 0.17 0.27 0.03 0.46 0.32 0.97 0.44 1 0.52 0.66 0.80 0.68 0.77 0.64 0.82 0.77 0.54 0.61

0.28 0.43 0.30 0.30 0.70 0.60 0.44 0.69 0.14 0.47 0.07 1 0.72 0.72 0.76 0.65 0.77 0.77 0.64 0.72 0.63

0.23 0.07 0.12 0.75 0.21 0.25 0.03 0.20 0.03 0.02 0.01 0.48 1 0.75 0.90 0.84 0.80 0.81 0.83 0.66 0.81

0.23 0.46 0.32 0.34 0.77 0.69 0.48 0.73 0.24 0.52 0.17 0.90 0.53 1 0.83 0.76 0.84 0.89 0.75 0.52 0.77

0.04 0.06 0.06 0.20 0.27 0.06 0.08 0.12 0.02 0.08 0.00 0.15 0.20 0.18 1 0.84 0.96 0.86 0.83 0.53 0.84

0.06 0.89 0.95 0.15 0.59 0.34 0.86 0.59 0.67 0.90 0.51 0.25 0.17 0.26 0.05 1 0.83 0.91 0.99 0.57 0.86

0.02 0.37 0.31 0.26 0.61 0.52 0.37 0.57 0.17 0.42 0.11 0.55 0.30 0.63 0.32 0.22 1 0.84 0.81 0.46 0.84

0.09 0.91 0.97 0.22 0.66 0.37 0.88 0.68 0.80 0.93 0.66 0.29 0.11 0.34 0.04 0.95 0.30 1 0.89 0.63 0.80

0.31 0.54 0.50 0.32 0.70 0.59 0.54 0.79 0.40 0.69 0.30 0.65 0.33 0.77 0.17 0.44 0.56 0.54 1 0.57 0.85

0.10 0.29 0.38 0.01 0.10 0.12 0.27 0.03 0.11 0.32 0.23 0.14 0.20 0.20 0.02 0.37 0.08 0.27 0.06 1 0.46

0.03 0.10 0.02 0.33 0.48 0.35 0.12 0.46 0.04 0.13 0.04 0.82 0.54 0.82 0.14 0.06 0.47 0.02 0.43 0.44 1

S. Canepari et al. / Atmospheric Environment 42 (2008) 8161–8175

As Ba Ca Cd Co Cr Cu Fe Mg Mn Na Ni Pb S Sb Si Sn Sr Ti Tl V

Values higher than 0.75 are reported in bold.

8173

8174

S. Canepari et al. / Atmospheric Environment 42 (2008) 8161–8175

RESIDUAL

EXTRACTABLE

ROAD DUST

BRAKE PADS V Tl Ti Sr Sn Si Sb S Pb Ni Na Mn Mg Fe Cu Cr Co Cd Ca Ba As

V Tl Ti Sr Sn Si Sb S Pb Ni Na Mn Mg Fe Cu Cr Co Cd Ca Ba As 1

10

100

1000

10000 100000

1

10

ASPHALT

100

1000

10000 100000

TYRES V Tl Ti Sr Sn Si Sb S Pb Ni Na Mn Mg Fe Cu Cr Co Cd Ca Ba As

V Tl Ti Sr Sn Si Sb S Pb Ni Na Mn Mg Fe Cu Cr Co Cd Ca Ba As 1

10

100

1000

10000 100000

1

10

100

1000

10000 100000

Fig. 10. Elemental composition (extractable and residual fractions) of road dust, brake pads, tyres and asphalt.

PM, most of the S-compounds comes from the secondary formation pathway and are found in the soluble fraction, in the form of sulphate. A low amount of sulphur, anyway, is found in the insoluble form, and can possibly be produced by the two examined sources. 4. Conclusions Size distribution of the water-extractable and residual fractions of 21 elements in airborne PM was measured at two traffic and one urban background sites in the Rome area during a 2-week study. The joined use of chemical and dimensional fractionation was of great value in discriminating between local vs. spatially homogeneous sources. Extractable and residual elemental fractions showed very different temporal patterns and different size distribution: the former predominated in the fine particle size, while the latter constituted most of the coarse fraction. Coarse, residual fractions were mainly responsible for the increase in element concentration from the urban background station to the traffic stations. A co-variation was detected for the residual fraction of almost all elements, whose concentrations show the same

temporal and dimensional pattern at all three sites. The source for these insoluble species can be identified in nontailpipe traffic emission. Similarities and dissimilarities among the individual elemental size distributions in the two chemical fractions were studied. The soluble fractions of As, Mg, Ni, Pb, S, Sn, Tl and V showed the same temporal pattern at the three sites suggesting spatially homogeneous sources for the soluble species of these elements. All these, apart from Mg, were almost totally in the fine fraction. Also the soluble fractions of Cd and Sb were almost totally in the fine fraction, but these elements were also influenced by localized sources. For all these elements, the soluble and residual fractions show different temporal patterns and very different spatial and size distributions, indicating that chemical fractionation makes these elements much more selective source tracers. It is worth noting that this behaviour is shown by all the four elements in the EU Directive 2004/107/CE. Given the relationship between elemental solubility and bio-availability, the application of the chemical fractionation procedure may be of great help in improving risk assessment and source apportionment studies.

S. Canepari et al. / Atmospheric Environment 42 (2008) 8161–8175

Acknowledgements The authors are indebted to ARPALazio (Dr. Franco Troiano and his staff) for having supplied PM samples. The authors gratefully acknowledge the NOAA Air Resources Laboratory (ARL) for the provision of the HYSPLIT transport and dispersion model and READY website (http://www.arl.noaa.gov/ready.html) used in this publication. NILU is acknowledged for providing the FLEXTRA trajectories (www.nilu.no/trajectories). We gratefully thank Dr. Slobodan Nickovic, who developed the DREAM model. This project has been carried out in the framework of the network of excellence ACCENT. References Allen, A.G., Nemitz, E., Shi, J.P., Harrison, R.M., Greenwood, J.C., 2001. Size distributions of trace metals in atmospheric aerosols in the United Kingdom. Atmospheric Environment 35, 4581–4591. Al-Masri, M.S., Al-Kharfan, K., Al-Shamali, K., 2006. Speciation of Pb, Cu and Zn determined by sequential extraction for identification of air pollution sources in Syria. Atmospheric Environment 40, 753–761. Almeida, S.M., Pio, C.A., Freitas, M.C., Reis, M.A., Trancoso, M.A., 2006. Source apportionment of atmospheric urban aerosol based on weekdays/weekend variability: evaluation of road re-suspended dust contribution. Atmospheric Environment 40, 2058–2067. Birmili, W., Allen, A.G., Bary, F., Harrison, R.M., 2006. Trace metal concentrations and water solubility in size-fractionated atmospheric particles and influence of road traffic. Environmental Science and Technology 40, 1144–1153. Canepari, S., Cardarelli, E., Perrino, C., Catrambone, M., Pietrodangelo, A., Strincone, M., 2006a. A new 2-step chemical fractionation method for the analysis of elements and non-volatile inorganic ions in PM10 samples: application to real samples collected in Rome (Italy). Atmospheric Environment 40, 7908–7923. Canepari, S., Cardarelli, E., Pietrodangelo, A., Giuliano, A., 2006b. Determination of metals, metalloids and non-volatile ions in airborne particulate matter by a new two-step sequential leaching procedure. Part A: experimental design and optimization. Talanta 69, 581–587. Canepari, S., Cardarelli, E., Pietrodangelo, A., Strincone, M., 2006c. Determination of metals, metalloids and non-volatile ions in airborne particulate matter by a new two-step sequential leaching procedure. Part B: validation on real equivalent samples. Talanta 69, 588–595. Dodd, J.A., Ondov, J.M., Tuncel, G., Dzubay, T.G., Stevens, R.K., 1991. Multimodal size spectra of submicrometer particles bearing various elements in rural air. Environmental Science and Technology 25, 890–903. Furuta, N., Iijima, A., Kambe, A., Sakai, K., Sato, K., 2005. Concentrations, enrichment and predominant sources of Sb and other trace elements in size classified airborne particulate matter collected in Tokyo from 1995 to 2004. Journal of Environmental Monitoring 7, 1155–1161. Gomez, D.R., Gine`, M.F., Sanchez Bellato, A.C., Smichowski, P., 2005. Antimony: a traffic-related element in the atmosphere of Buenos Aires, Argentina. Journal of Environmental Monitoring 7, 1162–1168. Heal, M.R., Hibbs, L.R., Agius, R.M., Beverland, I.J., 2005. Total and watersoluble trace metal content of urban background PM10, PM2.5 and black smoke in Edinburgh, UK. Atmospheric Environment 39, 1417– 1430. Harrison, R.M., Tilling, R., Calle´n Romero, M.S., Harrad, S., Jarvis, K., 2003. A study of trace metals and polycyclic aromatic hydrocarbons in the roadside environment. Atmospheric Environment 37, 2391–2402. Karthikeyan, S., Joshi, U.M., Balasubramanian, R., 2006. Microwave assisted sample preparation for determining water-soluble fraction of trace elements in urban airborne particulate matter: evaluation of bioavailability. Analytica Chimica Acta 576, 23–30.

8175

Maenhaut, W., Raes, N., Chi, X., Cafmeyer, J., Wang, W., Salma, I., 2005. Chemical composition and mass closure for fine and coarse aerosols at a Kerbisite in Budapest, Hungary, in spring 2002. X-Ray Spectrometry 34, 290–296. Marjama¨ki, M., Keskinen, J., Chen, D.-R., Pui, D.Y.H., 2000. Performance evaluation of the electrical low-pressure impactor (ELPI). Journal of Aerosol Science 31 (2), 249–261. Pakkanen, T.A., Kerminen, V.-M., Korhonrn, C.H., Hillamo, R.E., Aarnio, P., Koskentalo, T., Maenhaut, W., 2001. Use of atmospheric elemental size distributions in estimating aerosol sources in the Helsinki area. Atmospheric Environment 35, 5537–5551. Pakkanen, T.A., Kerminen, V.-M., Loukkola, K., Hillamo, R.E., Aarnio, P., Koskentalo, T., Maenhaut, W., 2003. Size distribution of mass and chemical components in street-level and rooftop PM1 particles in Helsinki. Atmospheric Environment 37, 1673–1690. Perrino, C., Pietrodangelo, A., Febo, A., 2001. An atmospheric stability index based on radon progeny measurements for the evaluation of primary urban pollution. Atmospheric Environment 35, 5235– 5244. Perrino, C., Catrambrone, M., Pietrodangelo, A.,( 2008). Influence of atmospheric stability on the mass concentration and chemical composition of atmospheric particles: a case study in Rome, Italy. Environment International 34, 621–628. Porstendorfer, J., Butterweck, G., Reineking, A., 1991. Diurnal variation of the concentrations of radon and its short-lived daughters in the atmosphere near the ground. Atmospheric Environment 25, 709–713. Qureshi, S., Dutkiewicz, V.A., Khan, A.R., Swami, K., Yang, K.X., Husain, L., Schwab, J.J., Demerjian, K.L., 2006. Elemental composition of PM2.5 aerosols in Queens, New York: solubility and temporal trends. Atmospheric Environment 40, 238–251. Sato, K., Tamura, T., Furuta, N., 2008. Partitioning between soluble and insoluble fractions of major and trace elements in size-classified airborne particulate matter collected in Tokyo. Journal of Environmental Monitoring 10, 211–218. Salma, I., Ocskay, R., Raes, N., Maenhaut, W., 2005. Fine structure of mass size distributions in an urban environment. Atmospheric Environment 39, 5363–5374. Samara, C., Voutsa, D., 2005. Size distribution of airborne particulate matter and associated heavy metals in the roadside environment. Chemosphere 59, 1197–1206. Sesana, L., Caprioli, E., Marcazzan, G.M., 2003. Long period study of outdoor radon concentration in Milan and correlation between its temporal variations and dispersion properties of atmosphere. Journal of Environmental Radioactivity 65, 147–160. Shweikani, R., Giadutt, T.G., Durrani, S.A., 1995. The effect of soil parameters on radon concentration values in the environment. Radiation Measurements 25, 581–584. Singh, M., Jaques, P.A., Sioutas, C., 2002. Size distribution and diurnal characteristics of particle-bound metals in source and receptor sites of the Los Angeles basin. Atmospheric Environment 36, 1675–1689. Smichowski, P., Polla, G., Go`mez, D., 2005. Metal fractionation of atmospheric aerosols via sequential chemical extraction: a review. Analytical and Bioanalytical Chemistry 381, 302–316. Sternbeck, J., Sjo¨din, Å., Andre´asson, K., 2002. Metal emissions from road traffic and the influence of resuspension – results from two tunnel studies. Atmospheric Environment 36, 4735–4744. Uexkull, O., Skerfving, S., Doyle, R., Braungart, M., 2005. Antimony in brake pads – a carcinogenic component? Journal of Cleaner Production 13, 19–31. Vasconcellos, P.C., Balasubramanian, R., Bruns, R.E., Sanchez-Ccoyllo, O., Andrade, M.F., Flues, M., 2007. Water-soluble ions and trace metals in airborne particles over urban areas of the state of Sa˜o Paulo, Brazil: influences of local sources and long range transport. Water Air, and Soil Pollution 186 (1–4), 63–73. Vecchi, R., Marcazzan, G., Valli, G., 2007. A study on nighttime–daytime PM10 concentration and elemental composition in relation to atmospheric dispersion in the urban area of Milan (Italy). Atmospheric Environment 41, 2136–2144. Weckwerth, G., 2001. Verification of traffic emitted aerosol components in the ambient air of Cologne (Germany). Atmospheric Environment 35, 5525–5536.