Atmospheric Environment 44 (2010) 1308e1320
Contents lists available at ScienceDirect
Atmospheric Environment journal homepage: www.elsevier.com/locate/atmosenv
A European aerosol phenomenology e 3: Physical and chemical characteristics of particulate matter from 60 rural, urban, and kerbside sites across Europe J.-P. Putaud a, *, R. Van Dingenen a, A. Alastuey b, H. Bauer c, W. Birmili d, J. Cyrys e, H. Flentje f, S. Fuzzi g, R. Gehrig h, H.C. Hansson i, R.M. Harrison j, H. Herrmann d, R. Hitzenberger k, C. Hüglin h, A.M. Jones j, A. Kasper-Giebl c, G. Kiss l, A. Kousa m, T.A.J. Kuhlbusch n, G. Löschau o, W. Maenhaut p, A. Molnar l, T. Moreno b, J. Pekkanen q, C. Perrino r, M. Pitz e, s, H. Puxbaum c, X. Querol b, S. Rodriguez t, I. Salma u, J. Schwarz v, J. Smolik v, J. Schneider w, G. Spindler d, H. ten Brink x, J. Tursic y, M. Viana b, A. Wiedensohler d, F. Raes a a
European Commission, JRC, Institute for Environment & Sustainability, I-21027 Ispra (VA), Italy Consejo Superior de Investigaciones Científicas, Institute for Environmental Assessment and Water Research, 08028 Barcelona, Spain Technical University Vienna, A-1060 Vienna, Austria d Leibniz Institute for Tropospheric Research, D-04318 Leipzig, Germany e Helmholtz Zentrum München, German Research Center for Environmental Health, D-85764 Neuherberg, Germany f Deutscher Wetterdienst, Meteorologisches Observatorium, D-82383 Hohenpeissenberg, Germany g Institute of Atmospheric Sciences and Climate (ISAC), I-40129 Bologna, Italy h EMPA, Swiss Federal Laboratories for Materials Testing and Research, Air Pollution/Environmental Technology, CH-8600 Dübendorf, Switzerland i Stockholm University, Department of Applied Environmental Science (ITM), SE-106 91 Stockholm, Sweden j School of Geography, Earth & Environ. Sciences, University of Birmingham, Birmingham B15 2TT, UK k Faculty of Physics, University of Vienna, Boltzmanngasse 5, A-1090 Vienna, Austria l Hungarian Academy of Sciences, University of Veszprém, H-8201 Veszprém, Hungary m YTV, Helsinki Metropolitan Area Council, Opastinsilta 6 A, FIN-00520 Helsinki, Finland n IUTA e.V., Air Quality & Sustainable Nanotechnology Unit, D-47229 Duisburg, Germany o Saxonian Office for the Environment, Agriculture and Geology, D-01311 Dresden, Germany p Institute for Nuclear Sciences, Ghent University, B-9000 Gent, Belgium q National Institute for Health and Welfare, PO Box 95, FIN-70701 Kuopio, Finland r National Research Council, Institute for Atmospheric Pollution, I-00016 Monterotondo Stazione (Roma), Italy s University of Augsburg, Environmental Science Center, D-86159 Augsburg, Germany t University of Huelva, Associated Unit to CSIC on ‘‘Air Pollution’’, E-21071, Huelva, Spain u Eotvos University, Institute of Chemistry, H-1518 Budapest, Hungary v Institute of Chemical Process Fundamentals ASCR, 16502 Prague 6, Czech Republic w Umweltbundesamt GmbH, A-1090 Wien, Austria x Energy Research Centre of The Netherlands (ECN), 1755 ZG Petten, The Netherlands y Environmental Agency of the Republic of Slovenia, Ministry for Environment and Spatial Planning, SI-1000 Ljubljana, Slovenia b c
a r t i c l e i n f o
a b s t r a c t
Article history: Received 14 July 2009 Received in revised form 27 November 2009 Accepted 4 December 2009
This paper synthesizes data on aerosol (particulate matter, PM) physical and chemical characteristics, which were obtained over the past decade in aerosol research and monitoring activities at more than 60 natural background, rural, near-city, urban, and kerbside sites across Europe. The data include simultaneously measured PM10 and/or PM2.5 mass on the one hand, and aerosol particle number concentrations or PM chemistry on the other hand. The aerosol data presented in our previous works (Van Dingenen et al., 2004; Putaud et al., 2004) were updated and merged to those collected in the framework of the EU supported European Cooperation in the field of Scientific and Technical action COST633 (Particulate matter: Properties related to health effects). A number of conclusions from our previous studies were confirmed. There is no single ratio between PM2.5 and PM10 mass concentrations valid for all sites, although fairly constant ratios ranging from 0.5 to 0.9 are observed at most individual sites. There is no general correlation between PM mass and particle number concentrations, although particle number concentrations increase with PM2.5 levels at most sites. The main constituents of both PM10 and PM2.5 are
Keywords: Aerosol Chemical composition Number concentration PM10 PM2.5
* Corresponding author. E-mail address:
[email protected] (J.-P. Putaud). 1352-2310/$ e see front matter Ó 2009 Elsevier Ltd. All rights reserved. doi:10.1016/j.atmosenv.2009.12.011
J.-P. Putaud et al. / Atmospheric Environment 44 (2010) 1308e1320
1309
generally organic matter, sulfate and nitrate. Mineral dust can also be a major constituent of PM10 at kerbside sites and in Southern Europe. There is a clear decreasing gradient in SO2 4 and NO3 contribution to PM10 when moving from rural to urban to kerbside sites. In contrast, the total carbon/PM10 ratio increases from rural to kerbside sites. Some new conclusions were also drawn from this work: the ratio between ultrafine particle and total particle number concentration decreases with PM2.5 concentration at all sites but one, and significant gradients in PM chemistry are observed when moving from Northwestern, to Southern to Central Europe. Compiling an even larger number of data sets would have further increased the significance of our conclusions, but collecting all the aerosol data sets obtained also through research projects remains a tedious task. Ó 2009 Elsevier Ltd. All rights reserved.
1. Introduction
2. Compilation of European aerosol data
Because aerosols affect human health, ecosystems (acidification, eutrophication) and visibility, emission abatement measures have been implemented over the last decades to reduce particle matter (PM10 and PM2.5) concentrations. As the legislation on particulate emission and air pollution is based on PM mass all over the world and is becoming more and more stringent (WHO, 2005), PM10 and PM2.5 concentrations are expected to further significantly decrease over the next decades. But can we assert that future PM concentration reductions will show positive effects on human health and ecosystems? Can we forecast the impact PM abatement measures on climate global and regional radiative forcing? Aerosol impacts are indeed linked to different characteristics that do not necessarily co-vary. Climate effects are related to its optical properties (scattering, absorption) throughout the atmospheric column, and cloud condensation formation potential, which in turn depends on particle size distribution and chemical composition. Impact on ecosystems depends on the deposition flux of specific species (acids, oxidized and reduced nitrogen, base cations, etc.). And we do not know yet which PM properties cause the health effects, although several epidemiological studies highlighted a link between short or long term exposure to PM mass concentration and human health (Zanobetti and Schwartz, 2005; Pope et al., 2008; Boldo et al., 2006; Pope, 2007; Miller et al., 2007). Therefore, abatement policies might be unnecessarily costly, inefficient, and even counter-productive if they address PM mass concentrations only. A thorough knowledge of the aerosol would help in designing a better legislation, using potential synergies and anticipating possible trade-offs between measures aiming at limiting air pollution and climate change. But PM10 and more recently also PM2.5 mass concentrations are still the only aerosol metrics measured systematically in national and international air pollution monitoring networks. Our goal is to show how different (or similar) other aerosol characteristics can be among various sites located in different regions across Europe, even where equal PM mass concentrations are observed. These characteristics are generally not available from regulatory monitoring networks (see e.g. EMEP and AirBase data banks), but rather from research projects. We have already pursued this process in two previous works (Van Dingenen et al., 2004; Putaud et al., 2004). The present study revises our previous conclusions in the light of data collected from more than 30 new sites. In view of the current European directive on particulate air pollution (2008/50/EC), we stratified the data according to PM mass concentration bins to address the following questions: (1) can PM2.5 and PMcoarse mass concentrations be inferred from PM10 and PM2.5 mass concentrations, respectively, (2) which PM constituents are responsible for PM10 high mass concentrations, and (3) can submm and ultrafine particle number concentrations be estimated based on PM2.5 mass concentration measurements. Answers are modulated according to the type of sites, from natural background to kerbside sites, and their location in Northwestern, Southern or Central Europe.
To achieve our goal, we selected data sets including PM mass concentrations on the one hand, and particle number and/or aerosol chemical composition data on the other hand, which were representative for a site during at least a season (i.e. minimum 6 weeks of continuous measurements). The selected sites and analytical methods employed are listed in Annexes 1 and 2. Our previous articles (Van Dingenen et al., 2004; Putaud et al., 2004) were fruits of a voluntary collaboration among scientists for synthesizing the results they obtained from advanced monitoring or atmospheric research in single documents. These works compiled aerosol physical and/or chemical characteristics from 34 sites in 11 countries. Few data sets had been collected for Eastern Europe and the semi-arid Mediterranean area though. An attempt to fill these gaps was conducted in the framework of the COST (European Cooperation in the field of Scientific and Technical Research) Action 633 “Particulate matter: Properties related to health effects” (http://cost633.dmu.dk). Further observations of aerosol physical characteristics and aerosol chemical characteristics were collected from 28 and 29 additional sites, respectively, mostly located in Southern and Central Europe. New data sets were also provided by some of the co-authors of the first two “European aerosol phenomenologies”. By merging these data, we can present and discuss here the physical and chemical characteristics observed over a network of more than 60 sites across Europe (Fig. 1). As previously, we have categorized the sampling sites using criteria proposed by the European Environment Agency (Larssen et al., 1999):
- Natural Background - Rural Background - Near-City Background - Urban Background To which we have added the following - Industrial - Kerbside
- distance from large pollution sources > 50 km - distance from large pollution sources 10e50 km - distance from large pollution sources 3e10 km - less than 2500 vehicles/day within a radius of 50 m - located within industrial areas - located by traffic lanes
For highlighting regional similarities and differences in aerosol characteristics, these sites have further been split in 3 large regions: Northwestern (NW), Southern and Central Europe (Fig. 1). Our scope is to describe intensive properties of the aerosol, i.e. not directly dependent on pollution dispersion. We therefore compile data sets in which particle number concentrations or size segregated (PM10, PM2.5, PMcoarse ¼ PM10 PM2.5) aerosol chemical composition can be related to PM10 and/or PM2.5 mass concentrations.
1310
J.-P. Putaud et al. / Atmospheric Environment 44 (2010) 1308e1320
Fig. 1. Location of the sampling sites. The pastel background colors delimit the 3 geographical sectors Northwestern, Southern, and Central Europe.
3. PM10 and PM2.5 mass concentrations While PM10 mass concentrations have been monitored by national and international air pollution monitoring networks for two decades (see e.g. EMEP and AirBase data banks), PM2.5 measurements are still being started up at many sites, because the new European Directive 2008/50/EC establishes target values for PM2.5 concentration and exposure to be met by 2010. Fig. 2a and b show 24-h averaged PM10 and PM2.5 yearly statistics observed at 48 sites across Europe (see also Annexes 3 and 4). The methods used to measure PM mass concentrations (Annex 2) and the uncertainties arising from comparing data produced by different institutes using different techniques were described by Van Dingenen et al. (2004), and are further discussed in Annex 5. When another technique than the European reference method was used to obtain PM mass concentrations, correction were applied by the data providers to make them equivalent to the reference method when necessary. PM mass data are therefore comparable within 15% (Annex 5). However, the data plotted in Fig. 2 were acquired over a decade (1996e2007), over which PM mass concentrations also decreased by up to probably 30% at some of the sites (Van Dingenen et al., 2004), which is nevertheless not sufficient to affect the spatial gradients in PM mass we describe here. A large range of PM10 concentrations (5e54 mg m3 annual average) is observed across the network (Fig. 2a). In each of the NW, Southern and Central sectors, an increasing gradient in PM10
concentration is generally observed when moving from naturalerural background to kerbside sites. Exceptions are however observed in all three sectors (e.g. lower or similar PM10 at urban or kerbside sites compared to rural sites). As there are no reasons why impact of local to urban sources should be significantly smaller at these specific urban or kerbside sites, we believe that these exceptions are due to differences in regional PM10 background at a smaller scale. Annual and 24-h limit values are exceeded in all three sectors of Europe, generally (but not solely) at urban and kerbside sites. Urban background PM10 annual mean and median values are significantly larger in southern Europe (median ¼ 36 mg m3) compared to NW and Central Europe (medians ¼ 24 and 26 mg m3, respectively). The range in PM2.5 concentrations observed across the network (3e35 mg m3 annual average) is similar to that of PM10 (Fig. 2b). An increasing gradient in PM2.5 is generally observed when moving from natural to urban background sites in Northwestern and Southern Europe. In Central Europe, PM2.5 can be as large at rural sites as at urban background sites, and concentrations at kerbside sites do not appear to be particularly high compared to urban background. The 2010 EU target value for annual PM2.5 mean concentration (25 mg m3) is exceeded at more urban background sites and less kerbside sites than the PM10 annual limit value (40 mg m3) is. Therefore, the PM2.5 directive is more stringent for urban background sites, while PM10 directives were more stringent for kerbside sites. The largest PM2.5 concentrations at near-city, urban background, and kerbside sites are observed in Southern Europe.
J.-P. Putaud et al. / Atmospheric Environment 44 (2010) 1308e1320
Northwestern 120
Central
average median EU 24-hr limit value 2005 EU annual standard
100
PM10 (µg/m3)
Southern
1311
80
60
40
Linz (AT) Graz (AT) Wien W (AT) Praha (CZ) Debrecen (HU) Bern (CH) Dresden (DE) Wien S (AT)
Hortobágy (HU) Melpitz (DE) Basel (CH) Zuerich (CH) Augsburg (DE)
Sevettijarvi (FI) Skreadalen (NO) Birkenes (NO)
0
Harwell (UK) Speulderbos (NL) Waasmunster (BE) Glasgow (UK) Belfast (UK) Manchester (UK) Birmingham (UK) London N (UK) London B (UK) Gent (BE) Uccle (BE) Duisburg (DE) Port Talbot (UK) Helsinki V (FI) London M (UK) Monagrega (ES) Montelibretti (IT) Iskrba (SI) Ispra (IT) Los Barrios (ES) Las Palmas (ES) Llodio (ES) Tarragona (ES) Huelva (ES) Algeciras (ES) Barcelona S (ES) Bologna (IT) Maribor T (SI) Madrid (ES) Barcelona H (ES) Maribor C (SI) Chaumont (CH) Hohenpeissenb. (DE) Streithofen (AT) Illmitz (AT) Kosetice (CZ)
20
120
average median
100
PM2.5 (µg/m³)
2010 EU annual target value 80
60
40
Wien S (AT)
K-Puszta (HU) Hortobágy (HU) Melpitz (DE) Basel (CH) Zuerich (CH) Augsburg (DE) Leipzig (DE) Linz (AT) Graz (AT) Wien W (AT) Praha (CZ) Debrecen (HU) Bern (CH)
Chaumont (CH) Hohenpeisenb. (DE) Streithofen (AT) Illmitz (AT)
Madrid (ES) Barcelona H (ES)
Las Palmas (ES) Llodio (ES) Tarragona (ES) Huelva (ES) Algeciras (ES) Barcelona S (ES) Bologna (IT)
Ispra (IT)
Montelibretti (IT)
Glasgow (UK) Belfast (UK) Manchester (UK) Birmingham (UK) London N (UK) London B (UK) Gent (BE) Uccle (BE) Duisburg (DE) Port Talbot (UK) Helsinki V (FI) London M (UK)
0
Sevettijarvi (FI) Skreadalen (NO) Birkenes (NO) Aspvreten (SE) Harwell (UK)
20
Fig. 2. Annual averages of PM10 and PM2.5 mass concentrations, including the 5, 25, 50 (median), 75, and 95% percentiles of their 24-h integrated concentrations. Symbol colors indicate the type of site (blue: natural background, green: rural background, yellow: near-city, red: urban background, grey: industrial, black: kerbside).
Considering that the European directive 2008/50/EC sets limits for PM2.5 concentration and exposure, one might wish to estimate what levels of PM2.5 can be anticipated at sites where PM10 measurements have been carried out. For each site, days were categorized according to the PM10 daily mean mass concentrations in 4 bins: 0 PM10 < 20, 20 PM10 < 50, 50 PM10 < 70, and 70 PM10 100 mg m3. Fig. 3a shows averages of simultaneous
measured PM2.5 and PM10 over these bins for 34 sites of the network. Although a significant correlation is observed (R2 ¼ 0.89, n ¼ 118), PM2.5 can obviously not be predicted from PM10 concentrations, since e.g. for 50 PM10 < 70 mg m3, PM2.5 mean concentrations range from 20 to 49 mg m3 across the network. However, PM10 and PM2.5 concentrations are well correlated (R2 0.99) at 26 sites. Sitespecific PM2.5/PM10 ratios (slopes) range from 0.44 to 0.90, without
1312
J.-P. Putaud et al. / Atmospheric Environment 44 (2010) 1308e1320
a
80
Sevettijarvi (FI) y = 0.90x R2 = 1.00
70
Skreadalen (NO) Birkenes (NO) Harwell (UK)
PM2.5 (µg/m3)
60
Waasmunster (BE) 50
Belfast (UK) London N (UK)
40
Gent (BE) y = 0.51x R2 = 0.98
30
Uccle (BE) Duisburg (DE)
20
Port Talbot Helsinki V (FI)
10
London M (UK) 0 0
10
20
30
40
50
PM10
60
70
80
90
100
(µg/m3)
Ispra (IT) Bologna (IT) Las Palmas (ES)
b
Llodio (ES)
200
Tarragona (ES) Huelva (ES)
190
Algeciras (ES) 180
y = 0.52x + 0.87 R2 = 1.00
PMcoarse (µg/m3)
70
y = 0.32x - 0.95 R2 = 0.92
60
y = 0.10x + 5.57 R2 = 0.98
Barcelona S (ES) Barcelona H (ES) Madrid (ES) Chaumont (CH) Streithofen (AT)
50 y = 0.14x + 2.39 R2 = 0.93 40
y = 0.06x + 6.17 R2 = 0.91
30
Hortobagy (HU) Basel (CH) Zuerich (CH) Linz (AT)
20
Graz (AT) Wien W (AT)
10
Praha (CZ) 0 0
10
20
30
40
50
60
PM2.5 (µg/m3)
70
80
90
100
Debrecen (HU) Bern (CH)
Fig. 3. Averaged (a) PM2.5 vs. PM10 and (b) PMcoarse vs. PM2.5 mass concentrations. Mass concentrations were averaged over days on which daily mean (a) PM10 were in the bins [0e20], [20e50[, [50e70[, [70e100[ mg m3 and (b) PM2.5 mass concentrations were in the bins [0e15[, [15e25[, [25e40[, [40e70[, and [70e100[ mg m3, respectively. Regression lines highlight (a) the lowest and highest site-specific regression slopes, and (b) examples of site-specific regressions between PMcoarse and PM2.5.
any clear relationship between the value of the PM2.5/PM10 ratio and the type of site or its location in Europe. Correlations between PM2.5 and PM10 concentrations indicate that PM10 and PM2.5 main source strengths co-vary, and/or that both concentrations are driven by a common phenomenon: pollution dispersion, depending on meteorology. A closer examination reveals that PM2.5/PM10 ratios slightly increase with PM10 levels at 12 sites, ranging from natural to urban background sites, which indicates that pollution periods are predominantly due to increases in PM2.5 mass concentration at these locations. Among noticeable exceptions, the kerbside site in Bern (CH), the urban background site in Debrecen (HU), and 2 urban background sites (Las Palmas and Llodio) in Spain (ES), where PM2.5/ PM10 ratios slightly to sharply decrease with increasing PM10 levels.
Significant contributions to high PM10 concentrations of coarse mineral dust suspended from the ground or transported from African deserts could explain this observation (see Section 5). Focusing PM monitoring on PM2.5 has sometimes been discussed in Europe. PM10 Fig. 3b shows that there is obviously no relationship between PM2.5 and PMcoarse which is valid across the whole network. This means that PMcoarse concentrations can generally not be inferred from PM2.5 measurements, and that PM2.5 and PMcoarse can be controlled by different sources and/or processes. However, PM2.5 and PMcoarse are well correlated at several locations ranging from rural to kerbside sites, and from Northwestern to Southern to Central Europe (Fig. 3b). At those sites, it would be possible to estimate average PMcoarse concentrations from PM2.5 mass measurements,
J.-P. Putaud et al. / Atmospheric Environment 44 (2010) 1308e1320
concentrations were PM2.5 < 10 mg m3, 10 PM2.5 < 20 mg m3, etc.. All data sets cover 1 year or more except those from Helsinki V (no summer data), Milano B (winter and summer data only), and Sagres, Mt. Foia, and Marseille V (summer data only). Total particle number concentrations as low as 2000 cm3 can be observed in all 3 sectors of Europe, and not only at sites impacted by clean maritime air. In contrast, total particle concentrations can exceed 40 000 cm3 at urban and kerbside sites e.g. in Northwestern and Central Europe. No big difference in particle number appears at rural sites among the three large Northwestern, Central, and Southern sectors of Europe. Clear increasing gradients in particle number concentrations generally occur when moving from natural background or rural sites to urban background or kerbside sites in all 3 sectors, excluding Helsinki V (FI), a site located 50 m away from a traffic lane with 14 000 vehicles a day, where the total particle concentration is very low compared to other kerbside sites. An increase in total particle concentration with increasing PM2.5 levels is clear at numerous
but of course not to detect sporadic and seldom occurrence of large PMcoarse concentrations. 4. Particle number concentrations We collected aerosol particle data from 12 more sites compared to our previous work (Van Dingenen et al., 2004), mostly located in the Northwestern sector of Europe. Total (Dp > 10 nm) and ultrafine (10 < Dp < 100 nm) particle concentrations rather than detailed size distributions were generally available from these new sites. The size distributions measured with Scanning or Differential Mobility Sizing systems (S/DMPSs) were integrated from 10 nm upwards to obtain comparable data. The intercomparability of particle number concentrations obtained from S/DMPSs and Condensation Particle Counters (CPCs) is estimated to 15%, as further discussed in Annex 5. Fig. 4a and b shows total and ultrafine particle number concentrations averaged over days on which daily mean PM2.5 mass
sub-µm particle number concentration (#/cm3)
ultrafine particle number concentration (#/cm3)
Northwestern 100000
100000
1313
Southern
Central PM2.5 bins (µg/m3) [0-10[
80000
[10-20[ [20-30[ [30-40[
60000
[40-50[ [50-60[ [60-70[
40000
[70-100]
20000
Leipzig L (DE)
Bern (CH)
Leipzig (DE)
Zuerich (CH)
Augsburg (DE)
Basel (CH)
Melpitz (DE)
Falkenberg (DE)
Hohenpeissen. (DE)
Barcelona (ES)
Chaumont (CH)
Milano B (IT)
Marseille V (FR)
Ispra (IT)
Sagres (PT)
Mt. Foia (PT)
Helsinki V (FI)
London M (UK)
Port Talbot (UK)
Amsterdam (NL)
London B (UK)
London N (UK)
Manchester (UK)
Birmingham (UK)
Belfast (UK)
Glasgow (UK)
Harwell (UK)
Aspvreten (SE)
Hyytiala (FI)
0
PM2.5 bins (µg/m3) [0-10[
80000
[10-20[ [20-30[ [30-40[
60000
[40-50[ [50-60[ [60-70[
40000
[70-100]
20000
Leipzig L (DE)
Leipzig (DE)
Zuerich (CH)
Augsburg (DE)
Melpitz (DE)
Falkenberg (DE)
Hohenpeissen. (DE)
Barcelona (ES)
Milano B (IT)
Marseille V (FR)
Ispra (IT)
Sagres (PT)
Mt. Foia (PT)
Helsinki V (FI)
Amsterdam (NL)
London B (UK)
Harwell (UK)
Aspvreten (SE)
Hyytiala (FI)
0
Fig. 4. Particle number concentration averaged over 10 mg m3 wide PM2.5 bins at the different sites of the network, split according to the sectors shown in Fig. 1. Pastel background colors indicate the site types (blue: natural background, green: rural background, yellow: near-city, rose: urban background, light grey: industrial, grey: kerbside).
1314
J.-P. Putaud et al. / Atmospheric Environment 44 (2010) 1308e1320
rural, urban or kerbside sites, but nowhere does the particle number increase proportionally to PM2.5: the ratio particle number/PM mass rather decreases with increasing PM2.5. This results in a lack of correlation (R2 ¼ 0.16, n ¼ 132) between particle number and PM2.5 mass concentrations, already observed at the US-EPA PM supersites (Solomon et al., 2008). This is due to the facts that most particles are generally smaller that 100 nm (ultrafine/total particle median ratio equals 0.76, and ranges 0.4e0.9), and that the ultrafine particle number concentration does generally not increase with PM2.5 levels (Fig. 4b). In contrast, increasing PM2.5 mass are related to increasing number of particles larger than Dp ¼ 100 nm. As a consequence, the ratio between ultrafine particle and total particle number concentrations decreases at all sites but one (Helsinki V, FI) when PM2.5 mass concentration increases (Table 1). The explanation of these observations is that different mechanisms are responsible for the production of ultrafine particles, which generally control the total particle number, and for the production of larger particles, which control PM mass. The production rates of these mechanisms do not necessarily co-vary with space and time (see e.g. Harrison and Jones, 2005). 5. Particulate matter chemical composition Fig. 5aec shows annual average chemical composition of PM10, PM2.5 (the fine fraction) and PMcoarse (¼ PM10 e PM2.5), respectively, at 39 sites, among which 23 are new compared to Putaud et al. (2004). Only sites for which at least one whole year of data was available are included in Fig. 5. A table with atmospheric concentrations of the main PM constituents is available in Annex 4. þ Elemental carbon (EC), nitrate (NO 3 ), and ammonium (NH4 ) concentrations derive directly from measurements. Organic matter (OM) and carbonaceous matter (CM) concentrations were calculated from organic carbon (OC) and total carbon data to account for the non-C atoms contained in particulate OM. Although sea-spray was clearly identified as a component of the aerosol at sites located
