Particles size and composition in Mediterranean countries: geographical variability and short-term health effects MED-PARTICLES Project 2011-2013 Under the Grant Agreement EU LIFE+ ENV/IT/327 Particles size and composition in Mediterranean countries: geographical variability and short-term health effects MED-PARTICLES ACTION 8. Report on: geographical comparison of different PM sizes and components across cities and countries, and according to Saharan dust/no dust days and forest fires days. Summary: Report on the methodology adopted in each city and the results of the speciation process. 1 PM chemical components 1 Introduction Atmospheric aerosols are generally composed of variable amounts of sulphate (SO42-), ammonium (NH4+), nitrate (NO3-), sodium (Na), chloride (Cl), trace metals (e.g: Ni, V, Cu, Zn, Mn) crustal elements (Al, Si, Ca, Fe, Mg, K) water and carbonaceous material. The sulphate component is derived predominantly from the atmospheric oxidation of anthropogenic and natural sulphur-containing compounds such as sulphur dioxide (SO2). Nitrate is formed mainly from the oxidation of atmospheric nitrogen dioxide (NO2). Sulphate and nitrate are initially formed as sulphuric (H2SO4) and nitric acids (HNO3), but are progressively neutralised by atmospheric ammonia forming the corresponding ammonium salts. Chlorides also enter atmospheric particles as a result of ammonia neutralisation of hydrochloric acid (HCl) vapour, which is emitted from sources such as incinerators and power stations. But the main source of chlorides is sea spray even at locations hundreds of miles from the coast. Crustal materials include soil dust and windblown minerals. They vary in composition according to local geology and surface conditions and reside mainly in the coarse particle fraction. The carbonaceous fraction of the aerosols consists of both elemental and organic carbon. Elemental carbon (EC), that might also called black carbon (BC), is emitted directly into the atmosphere, predominantly from combustion processes. 2 Sampling sites Barcelona, Spain: The site is called Palau Reial, 41°23'14"N, 2°6'56" E, 77 a.s.l.). It is an urban background site, but highly influenced by emissions from the Diagonal Avenue located at around 300m with a traffic flow of around 100000 cars/working day. The Barcelona metropolitan area is characterized by high road traffic density and by a wide range of industrial activities. Facilities comprise ferrous and non-ferrous smelters, cement and asphalt production industries, which are spread between the two river basins in the North and South of the Metropolitan area. Furthermore, two power 2 stations and two city waste incinerators are also based in the considered area. Traffic density is a consequence of the high population concentration in the city of Barcelona (101 km2 with 1.6 million inhabitants, 4.5 million in the greater metropolitan area) leading to one of the highest car densities in Europe. Jointly, the urban architecture, characterized by square-blocks with narrow streets, reduces the dispersion of pollutants and the scarce precipitation favors the accumulation of air pollutants. Madrid, Spain: The site is called Escuelas Aguirre (40º25’32’’N, 03º40’52’’W, 672 m a.s.l.), situated in an urban location in the city centre, significantly influenced by vehicular traffic. Madrid (40º25’N, 03º42’W) is a South-European city with specific characteristics from the point of view of air pollution and climate. The population of the city of is approximately 3.2 million inhabitants (6 million in the metropolitan area), with a car fleet of nearly 1.9 million vehicles in the city (4.5 in the metropolitan area). More than a half (53 %) of the vehicle fleet in 2012 is diesel powered, confirming an increasing trend towards a dieselization in the last ten years. Since industrial activity in the surrounding area consists essentially of light factories, Madrid urban area is characterised as the typical urban plume, fed by road traffic emissions and heating devices in the winter months. These devices are mainly natural gas fuelled, and in a minor extent fuel-oil, although there are a number of coal combustion devices still operating in the city. Huelva, Spain: The site is called Campus del Carmen, 37° 15' 57"N, 6° 55'26" W, 10 a.s.l), an urban background site influenced by industrial emissions. The city of Huelva (160,000 inhabitants) lies on the Atlantic coast of southwestern Spain. An industrial state with several chemical factories (a copper smelter among them) was built in the 1960s, located 1.5 km to the south of the city. Montelibretti is a peri-urban site, located inside the CNR Research Area RM1, about 25 km from the centre of Rome. The nearest village, Monterotondo Scalo, is at about 5 km, but the site is also influenced by the traffic caused by the CNR staff and the Salaria National road, at about 500 meters. Rome, Italy: The site is called Villa Ada, is an urban background station located inside the main green area in the city of Rome (Villa Ada Park). The sampling point is about 200 m from the nearest street. Bologna, Italy: an urban background site located in Bologna in the Po Valley. PM2.5 samples were collected on a daily basis during the period 2011-2012. 3 3 Instrumentation and chemical analysis In the three Spanish cities, Huelva, Barcelona and Madrid high volume samplers (MCV, and Andersen samplers) were used for PM sampling with a flow rate of 30 m3 h−1 equipped with PM10 and PM2.5 inlets for the determination of PM10 and PM2.5 mass concentrations. PM sampling had 24h duration. Approximately 2-3 samples were collected every week using quartz-fiber filters ( 15 cm), previously baked at 200 °C. The filters were conditioned at constant temperature and humidity before and after sampling. They were then weighed at least three times (in 24h) to obtain constant values. PM 10 and PM2.5 concentrations were determined gravimetrically. Subsequently, the collected filters were analyzed for water-soluble ions (NH4+, Cl-, SO42-, NO3-), major elements (Al, Ca, K, Mg, Fe, Na) and 46 trace elements by inductively coupled plasma atomic emission, ICP-AES and mass spectrometry, ICP-MS respectively following the procedure described by Querol et al. (2001). Briefly, ½ of each filter was acid digested (HF:HNO3:HClO4, with a mixture of 2.5:1.25:1.25 ml, kept at 90°C in a Teflon reactor during 6h, driven to dryness and re-dissolved with 1.25 ml HNO3 and then diluted with water up to 25 ml) for the chemical analysis using ICP-AES and ICP-MS. To assure the quality of the analytical procedure a small amount (5 mg) of the NIST-1633b (fly ash) reference material loaded on a ¼ quartz micro-fibre filter was also analysed. Detection limit and accuracy of the techniques were estimated as 0.18 ng/m3 and 1-3% respectively for ICP-AES, and 0.007 ng/m3 and 0-7% respectively for ICP-MS. Another ¼ of each filter was water leached (6h at 60ºC, preceded by incubation in an ultrasound bath for 10 minutes, in 50 ml sealed PVC bottles) for the determination of soluble ion concentrations (Cl-, SO42-, NO3-) by ion chromatography and ion selective electrode (NH4+), allowing an average detection limit for the analysed components of 25-30 ng/m3. In the remaining ¼ of each filter total carbon was determined by means of the ECOSC-144DR instrument. In Barcelona, for the samples collected in the period 2007-2010 a (1.5 cm2) portion of each filter was also used for the determination of organic and elemental carbon (OC and EC, respectively) by a thermal–optical transmission technique (Birch and Cary 1996) using the Sunset Laboratory OCEC Analyser and the NIOSH protocol. The sum of OC+EC was expressed as total carbon, Ctotal. PM10 and PM2.5 samples were collected in Barcelona, during the period 2003-2010. A total of 680 and 737 valid samples of PM10 and PM2.5 respectively, were used for source apportionment analysis. PM sampling was conducted in Huelva, Campus El Carmen during 2003-2010 and a total of 4 412 and 405 valid samples of PM10 and PM2.5 respectively, were used for further analysis. In Madrid, PM10 and PM2.5 samples were collected during the period 2007-2008. A total of 94 and 92 valid samples of PM10 and PM2.5 respectively, were used for source apportionment analysis. In Montelibretti site and at the Villa Ada site PM2.5 and PM10 sampling was conducted by both the beta attenuation monitor SM200 (Opsis AB Furulund-S) operating at flow rate of 1 m3 h-1 and teflon filters as a substrate and the MICRODUST sequential sampler (AQUARIA, Lacchiarella, MI-I) operating at a flow rate of 2 l min-1 equipped with quartz filters. The Teflon filters were subjected to energy-dispersion X-ray fluorescence analysis (X-Lab2000, Spectro Analytical Instruments, Kleve-D) for the determination of major and minor elements: Al, Si, Fe, K, Ca, As, Cr, Cu, Mn, Ni, Pb, Ti, V, Zn. Then the filters were extracted in deionized water and analyzed by ion chromatography (ICS90 Dionex Co) for the determination of Cl-, NO3-, SO4=, Na+, NH4+, K+, Mg++ and Ca++. Elemental EC, and organic carbon OC were analysed in the quartz filters by thermo-optical analysis (OCEC Sunset Analyser, Sunset Laboratory) using the NIOSH temperature protocol. PM10 samples were collected in Montelibretti and in Villa Ada sites on a daily basis during the period 2005-2010. A total of 1940 and 860 valid PM10 samples in Montelibretti and Villa Ada respectively were used for source apportionment analysis. In Bologna site, two Dual Channel Monitors (SWAM, Fai Instrument) were used to sample PM on quartz filters. After sampling EC and OC were determined by thermo-optical analysis (OCEC Sunset Analyser, Sunset Laboratory) using the EUSAAR_2 protocol. Ion chromatography (Dionex) was used for the determination of Cl-, NO3-, SO4=, Na+, NH4+, K+, Mg++ and Ca++. PM2.5 samples were collected in Bologna on a daily basis during the period 2012-2013. 4 Average concentrations of PM sizes and chemical components Tables 1-6 provide the average concentrations and the standard deviations of PM chemical components in the sampling sites. Only species with more than 75% of their concentration values > LOD are given in the tables. Some atmospheric components had a preference for the coarse sizes such as mineral matter, sea spray and nitrate. Other components such as caronaceous aerosol, ammonium and in less proportion sulphate are mostly found in the fine fraction. It is important to remark that the amount of coarse components (PM2.5-10) in the atmosphere of Barcelona was significant. In general the 5 concentrations were higher in Barcelona, Bologna and Madrid where the influence of traffic enhanced significantly the levels of all chemical components. In Figure 1 the long-term trends of four representative components in PM10 are shown. The cities of Barcelona and Huelva are presented as they cover longer periods of PM sampling than the other sites. The secondary aerosol components such as SO42-, NO3-, and typical traffic tracers Cu, Zn emitted from the abrasion of tyres and breaks show a decreasing trend from 2003 to 2010, especially evident after 2006. These trends are probably related to the mitigation strategies taken the last decade and a proportion of these reductions may be related to the economic crisis and favorable meteorology. Table 1. Average concentrations and standard deviation of PM10 and PM2.5 chemical components in Huelva, Spain Componen t µg/m3 Ctotal Al2O3 Ca K Na Mg Fe PO43SO42NO3Cl NH4+ ng/m3 Ti V Cr Mn Ni Cu Zn As Se Sb PM1 0 Mea n 5.42 1.66 1.17 0.42 1.10 0.30 0.67 0.29 3.78 2.67 1.15 1.04 56.83 5.40 2.73 12.89 3.46 47.95 50.98 6.80 1.78 0.87 PM2.5 SD Mean SD 3.43 1.72 0.97 0.33 0.89 0.23 0.58 0.37 3.26 1.81 1.61 1.11 3.91 0.48 0.33 0.21 0.34 0.08 0.18 0.11 3.31 1.15 0.31 1.15 2.26 0.70 0.52 0.20 0.47 0.09 0.20 0.12 2.74 1.21 0.43 1.22 86.33 4.39 3.80 24.04 3.45 78.23 83.67 10.18 3.14 1.74 19.55 3.83 1.54 3.50 2.42 29.02 45.56 5.70 1.44 0.69 38.69 3.41 3.00 4.29 2.98 33.29 83.94 9.31 4.38 0.17 6 Ba Pb 17.57 14.34 30.23 17.65 12.40 12.98 26.52 16.25 Table 2. Average concentrations and standard deviation (SD) of PM10 and PM2.5 chemical components in Barcelona, Spain Component µg/m3 Ctotal OC EC Al2O3 Ca K Na Mg Fe SO42NO3Cl NH4 PM 10 Me an 8.77 3.36 1.84 1.28 2.14 0.38 1.01 0.28 0.86 3.79 4.35 0.90 1.25 PM2.5 SD Mean SD 3.47 1.96 1.18 1.21 1.64 0.29 0.72 0.17 0.54 2.62 3.72 0.86 1.14 8.05 2.86 1.53 0.47 0.60 0.20 0.26 0.08 0.28 3.40 2.43 0.36 1.58 3.25 1.82 0.98 0.60 0.62 0.23 0.18 0.07 0.21 2.42 3.17 0.22 1.51 33.41 13.74 13.41 7.13 3.24 7.91 2.55 6.11 2.30 10.70 3.35 8.59 3.58 5.78 2.53 51.39 27.05 35.89 80.61 0.65 0.49 3.72 58.72 0.61 0.49 1.88 67.65 0.41 0.38 1.50 31.46 15.76 34.17 22.06 13.53 21.08 ng/m3 Ti V Cr Mn Ni Cu Zn As Se Sb Ba Pb 40.0 6 10.4 4 5.21 17.5 5 4.81 56.5 1 85.4 8 0.84 0.70 4.94 25.9 2 17.4 9 Table 3. Average concentrations and standard deviation (SD) of PM10 and PM2.5 chemical components in Madrid, Spain 7 Componen t µg/m3 PM10 Mean PM2.5 SD Mean SD Ctotal 10.44 5.59 7.59 3.86 Al2O3 0.72 0.59 0.36 0.29 Ca 2.09 1.35 0.26 0.15 K 0.33 0.22 0.20 0.41 Na 0.38 0.33 0.16 0.10 Mg 0.22 0.15 0.08 0.08 Fe 1.67 0.87 0.20 0.15 SO42NO3- 3.15 1.40 2.35 1.13 2.46 2.09 1.49 1.72 Cl 0.75 0.46 0.44 0.31 NH4+ 1.51 0.83 1.48 0.47 Ti 39.92 30.91 8.27 13.97 V 1.86 1.25 0.89 0.67 Cr 6.73 3.85 4.79 4.73 Mn 20.40 11.42 5.40 3.02 Ni 58.98 43.54 3.91 2.89 Cu 1.72 1.00 18.51 9.98 Zn 112.75 68.86 50.16 26.22 As 1.74 ng/m3 1.97 1.76 1.75 Se 0.43 0.25 0.33 0.22 Sb 14.98 8.82 4.80 14.52 Ba 38.25 35.82 43.69 32.56 Pb 15.09 11.01 9.75 7.90 Table 4. Average concentrations and standard deviation (SD) of PM10 chemical components in Montelibretti, Italy Component µg/m3 Ctotal OC EC Si Al PM10 Mean 8.58 7.37 1.21 0.89 0.25 SD 4.40 3.77 0.64 1.08 0.28 8 Ca K Na Mg Fe Ca2+ Mg2+ SO42NO3Cl NH4+ ng/m3 Ti V Cr Mn Ni Cu Zn As Pb 1.35 0.38 0.52 0.09 0.24 1.10 0.09 2.89 1.73 0.35 0.91 1.09 0.24 0.62 0.09 0.22 0.72 0.09 1.95 1.35 0.73 0.61 17.00 3.66 5.08 7.27 3.50 17.24 23.64 1.23 4.18 27.42 2.98 4.34 4.92 3.22 16.03 25.28 1.01 5.39 Table 5. Average concentrations and standard deviation (SD) of PM10 chemical components in Villa Ada, Rome, Italy Component PM10 µg/m3 Mean Ctotal 10.72 OC 8.47 EC 2.25 Si 1.00 Al 0.26 Ca 1.26 K 0.38 Na 0.54 Mg 0.10 Fe 0.46 2+ Ca 1.15 2+ Mg 0.10 K+ 0.29 2SO4 3.37 NO3 1.99 Cl 0.36 SD 5.52 4.25 1.37 0.91 0.24 0.71 0.32 0.57 0.10 0.28 0.71 0.10 0.29 2.07 1.40 0.70 9 NH4+ ng/m3 Ti V Cr Mn Ni Cu Zn As Pb 0.87 0.63 22.92 5.56 9.93 12.10 4.64 40.37 38.04 3.11 10.40 24.74 3.77 6.03 7.73 2.69 19.70 34.97 2.13 8.81 Table 6. Average concentrations and standard deviation (SD) of PM2.5 chemical components in Bologna, Italy Component PM2.5 µg/m3 OC 1.6 1.5 SiO2 2.4 2.2 SO42- 5.1 2.1 NO3- 0.1 0.1 Fe 3.2 3.5 Mn 2.4 2.6 Ni 6.2 5.9 Ti 10.8 11.7 ng/m3 10 Figure 1. Long term trends of SO42-, NO3-, Cu and Zn in PM10 in Barcelona and Huelva 11 5 Average concentrations of PM sizes and chemical components by presence of dust or forest fires events Barcelona Table 1. Descriptive characteristics of the PM10 components by presence of dust at the ground level, Barcelona, Spain, (study period 2003-2010) PM components PM10 Ctotal SiO2 Al2O3 Ca K Na Mg Fe SO42NO3Cl NH4+ P Ti V Cr Mn Ni Cu Zn As Se Sr Cd Sn Sb Ba La Pb Mean #No Dust days # Dust days 557 557 557 557 557 557 557 557 557 557 557 557 557 557 557 557 557 557 557 557 557 557 557 557 557 557 557 557 557 557 109 109 109 109 109 109 109 109 109 109 109 109 109 109 109 109 109 109 109 109 109 109 109 109 109 109 109 109 109 109 No Dust Dust 37.512 6.664 3.098 1.033 1.972 0.346 0.924 0.244 0.785 3.267 4.145 0.859 1.188 0.025 0.033 0.009 0.005 0.016 0.004 0.055 0.086 0.001 0.001 0.005 0.000 0.006 0.005 0.024 0.000 0.017 52.097 7.793 6.840 2.280 2.843 0.542 1.298 0.429 1.176 6.023 4.526 0.958 1.438 0.035 0.070 0.014 0.006 0.022 0.006 0.060 0.082 0.001 0.001 0.008 0.000 0.007 0.006 0.034 0.001 0.020 12 Table 2. Descriptive characteristics of the PM10 components by presence of forest fires events, Barcelona, Spain, ( study period 2003-2010) PM components PM10 Ctotal SiO2 Al2O3 Ca K Na Mg Fe SO42NO3Cl NH4+ P Ti V Cr Mn Ni Cu Zn As Se Sr Cd Sn Sb Ba La Pb #No Fire days # Fire days 663 663 663 663 663 663 663 663 663 663 663 663 663 663 663 663 663 663 663 663 663 663 663 663 663 663 663 663 663 663 16 16 16 16 16 16 16 16 16 16 16 16 16 16 16 16 16 16 16 16 16 16 16 16 16 16 16 16 16 16 Mean No Fire Fire 39.885 6.845 3.773 1.258 2.122 0.377 0.997 0.275 0.848 3.718 4.203 0.868 1.212 0.027 0.039 0.010 0.005 0.017 0.005 0.057 0.085 0.001 0.001 0.005 0.000 0.006 0.005 0.026 0.000 0.017 55.454 7.941 6.157 2.052 2.639 0.546 1.083 0.415 1.123 5.916 6.525 1.161 1.981 0.032 0.066 0.013 0.006 0.024 0.006 0.051 0.103 0.001 0.001 0.008 0.000 0.006 0.005 0.036 0.001 0.019 13 Rome- Villa Ada Table 3. Descriptive characteristics of the PM10 components in Villa Ada monitoring station by presence of dust at the ground level, Rome, Italy, ( study period 2005-2007) PM components PM10 Si Al Ca K Na Mg++ Fe SO42NO3Cl NH4+ Ti V Cr Mn Ni Cu Zn As Pb K+ Ca++ OC EC #No Dust days # Dust days 843 711 711 705 711 711 711 711 711 711 711 711 710 711 711 711 710 711 711 710 705 556 711 860 860 173 144 144 140 144 144 144 144 144 144 144 144 144 144 144 144 144 144 144 144 144 114 144 170 170 Mean No Dust Dust 29.022 0.773 0.205 1.150 0.366 0.544 0.093 0.424 3.094 1.928 0.391 0.812 15.428 4.874 9.952 11.374 4.574 40.589 38.316 2.402 9.703 0.294 1.015 8.557 2.194 34.346 1.997 0.532 1.816 0.471 0.533 0.108 0.620 4.538 2.220 0.242 1.075 46.477 8.386 9.735 15.260 4.666 39.667 33.481 2.450 11.715 0.302 1.639 8.200 2.340 14 Table 4. Descriptive characteristics of the PM10 components in Villa Ada monitoring station by presence of forest fires events, Rome, Italy, ( study period 2005-2007) PM components PM10 Si Al Ca K Na Mg++ Fe SO42NO3Cl NH4+ Ti V Cr Mn Ni Cu Zn As Pb K+ Ca++ OC EC #No Fire days # Fire days 993 834 834 824 834 834 834 834 834 834 834 834 833 834 834 834 833 834 834 833 828 644 834 1007 1007 40 37 37 37 37 37 37 37 37 37 37 37 37 37 37 37 37 37 37 37 37 35 37 40 40 Mean No Fire 29.907 0.956 0.254 1.234 0.378 0.539 0.094 0.456 3.318 1.972 0.366 0.863 20.234 5.417 9.940 11.907 4.626 40.596 38.108 2.460 10.088 0.292 1.098 8.564 2.236 Fire 32.720 1.892 0.511 1.959 0.528 0.578 0.122 0.542 4.614 2.412 0.245 0.999 41.057 8.070 9.857 17.808 4.262 35.730 21.892 1.276 10.003 0.345 1.745 6.575 1.688 15 Rome - Montelibretti Table 5. Descriptive characteristics of the PM10 components in Montelibretti monitoring station by presence of dust at the ground level, Rome, Italy, ( study period 2005-2010) PM components PM10 Si Al Ca K Na Mg++ Fe SO42NO3Cl NH4+ Ti V Cr Mn Ni Cu Zn As Pb Ca++ OC EC #No Dust days # Dust days 1711 1594 1595 1595 1595 1610 1610 1594 1610 1610 1610 1610 1589 1589 1589 1589 1589 1589 1589 1590 1589 1610 1760 1760 353 331 331 331 331 334 334 331 334 334 334 334 331 331 331 331 331 331 331 331 331 334 351 351 Mean No Dust 27.297 0.636 0.179 1.178 0.360 0.507 0.087 0.189 2.660 1.719 0.361 0.882 10.580 3.123 4.936 6.359 3.409 17.013 23.908 1.123 4.079 1.003 7.538 1.187 Dust 36.159 2.056 0.548 2.148 0.496 0.606 0.111 0.462 3.941 1.836 0.331 1.029 45.955 5.905 5.596 11.405 3.754 17.952 22.103 1.279 4.102 1.616 6.919 1.309 16 Table 6. Descriptive characteristics of the PM10 components in Montelibretti monitoring station by presence of forest fires events, Rome, Italy, ( study period 2005-2010) PM components PM10 Si Al Ca K Na Mg++ Fe SO42NO3Cl NH4+ Ti V Cr Mn Ni Cu Zn As Pb Ca++ OC EC #No Fire days # Fire days 2032 1898 1899 1899 1899 1916 1916 1898 1916 1916 1916 1916 1893 1893 1893 1893 1893 1893 1893 1894 1893 1916 2080 2080 54 49 49 49 49 50 50 49 50 50 50 50 49 49 49 49 49 49 49 49 49 50 53 53 Mean No Fire 28.701 0.862 0.238 1.322 0.381 0.520 0.091 0.232 2.853 1.745 0.354 0.913 16.106 3.574 5.074 7.152 3.499 17.250 23.805 1.145 4.122 1.096 7.465 1.210 Fire 35.215 2.003 0.538 2.455 0.524 0.609 0.109 0.470 4.466 1.604 0.292 0.875 47.451 5.820 5.153 11.731 3.120 16.510 16.653 1.349 2.900 1.648 6.121 1.193 17 6 Source apportionment analysis Source apportionment analysis was conducted in the datasets of PM chemical components of Huelva, Barcelona, Madrid, Montelibretti and Villa Ada. The dataset from Bologna was not appropriate for such an analysis since only 8 components had 75% of the measured concentrations above the detection limit, LOD. The investigation of the emission sources was carried out by means of a Positive Matrix Factorization (PMF, Paatero and Tapper, 1994). PMF is a least squares factor analysis based on mass conservation principle (1) to assist in identification of sources and their contributions to observed pollutant loadings. In equation (1), xij is the jth species loading measured in the ith sample, gik is the contribution of the kth source to the ith sample, fkj is the loading of the jth species in the kth source and eij is the residual associated with the jth species loading measured in the ith sample. The number of elements to use within the PMF model and their uncertainties were determined taking into account the possible sources of uncertainty as the method analytical uncertainty, the error introduced and the blank values Karanasiou et al., (2009). The calculated uncertainty of these components was in the range of 5-25% of the measured concentrations. The robust mode was used to reduce the influence of extreme values. After several trials with different Fpeak values (-2, -1.5, -1, -0.5, 0, 0.5, 1, 1.5, 2) the Fpeak value was set to zero. To determine the number of sources, different numbers of factors were tested. Preliminary runs of PMF were performed to investigate if the resulting factors make sense with respect to known source profiles provided in the literature. Key parameters given in the PMF analysis output, such as the Explained Variation of the matrix F were used to select the number of factors and assisted in interpreting the solution. The global optima of the PMF solution was also tested using multiple starting values. Finally, we selected the number of factors that both adequately fit the data and provided the most physically meaningful results. The resolved solution of PMF produced a good fit to the data as indicated by the Q-value and the scaled residuals. Furthermore, the comparison of the reconstructed PM10 mass contributions from all sources with the observed PM10 mass concentrations provided a squared 18 correlation coefficient of >0.75 (Figure 2) with the only exception PM2.5 in Huelva where the correlation between measured and predicted mass was lower (0.54). This indicates that the resolved sources effectively account for most of the variation in the PM mass concentration. PM10 determined gravimetrically 100 Huelva 90 80 70 60 y = 0.96x - 0.23 R² = 0.75 50 40 30 20 10 0 0 20 40 60 80 100 PM10 predicted 50 45 PM2.5 determined gravimetrically y = 0.74x + 3.03 R² = 0.54 Huelva 40 35 30 25 20 15 10 5 0 0 10 20 30 40 50 60 PM2.5 predicted 19 PM10 detrmined gravimetrically 140 y = 0.99x + 0.339 R² = 0.98 Madrid 120 100 80 60 40 20 0 0 20 40 60 80 100 120 140 PM10 predicted PM2.5 determined gravimetrically 60 Madrid 50 40 y = 0.99x - 0.37 R² = 0.89 30 20 10 0 0 10 20 30 40 50 60 PM2.5 predicted 20 PM10 detrmined gravimetrically 140 y = 0.97x + 0.10 R² = 0.85 Barcelona 120 100 80 60 40 20 0 0 20 40 60 80 100 120 140 PM2.5 detrmined gravimetrically PM10 predicted 100 90 80 70 60 50 40 30 20 10 0 y = 0.95x - 0.05 R² = 0.82 Barcelona 0 20 40 60 80 100 PM2.5 predicted PM10 determined gravimetrically 140 Montelibretti 120 y = 0.83x + 4.75 R² = 0.83 100 80 60 40 20 0 0 50 100 150 200 PM10 predicted 21 PM10 determined gravimetrically 100 Villa Ada 90 80 70 60 50 y = 0.88x + 3.78 R² = 0.89 40 30 20 10 0 0 20 40 60 80 100 PM10 predicted Figure 2. Regression of the PM mass determined gravimetrically and the PM mass predicted by PMF model for the four cities. 6.1 Emission sources and mass contribution in Huelva, Spain The chemical profiles of the emission sources in Huelva determined are given in Figure 3 with the most reasonable results obtained when 6 factors were selected. The first source is secondary aerosol as it mainly consists of the secondary components, sulphate, nitrate and ammonium. The phosphate production source was recognized by the high contribution of phosphate ion while the mineral source was made up of typical soil components (Al, Ca, Mn, Fe, K, and Mg). A mixed source related to traffic was characterized by elements emitted by vehicles exhaust (carbonaceous components, Ctotal) and road dust elements like Fe, Cu, Zn produced during tyre and break abrasion. The marine aerosol source was consisted of the typical marine species Cl, Na, and a proportion of Mg. Finally, a source called metallurgical industry was identified, as it explained the most of Cu variance, probably characterizing the emissions of the Cu plant situated in the proximity of Huelva city. Figure 4 provides the mass contribution of the emission sources in PM10 and PM2.5. The highest contributors to PM were the mineral source (28% for PM10 and 21% PM2.5), the traffic source (30% and 22% of PM10 and PM2.5 mass, respectively) and the secondary aerosol (19% of the PM10 and 23% of the PM2.5 mass). The two industrial sources when summed contributed around 13% of PM levels. 22 6.2 Emission sources and mass contribution in Barcelona, Spain The chemical profiles of emission sources in Barcelona are given in Figure 5 with the most reasonable results obtained when seven factors were selected. The first source is vehicles emissions representing the emissions from the vehicle engine and exhaust. It explains most of total carbon variation. The second source is nitrate due to the high contribution of nitrate. The two natural sources, marine aerosol and the mineral source were resolved by the presence of typical sea salt and crustal components. The secondary sulphate source is the result of the formation of secondary sulfate in atmosphere from the photochemical oxidation of sulfur oxides initially emitted as gases from local emissions (power plants, port) and from long range transport. The fact that this source explains the variation of V and Ni, two typical fuel oil tracers indicates the presence of shipping emissions in this factor. The industrial factor, with high concentrations of TC, Zn, Pb, Fe, NH4+, Cl- and Mn was related to the mixed influence of industrial activities located in the area such as smelters and cement kilns. The road dust source was traced by the high contribution of TC, NO3-, Ca, Fe, Cu, components produced mechanically from the abrasion of the road, tyres and brakes. These species are deposited on the road surface and subsequently are resuspended by vehicles circulation. In Barcelona urban area the highest contributors to PM were the road dust source (21% for PM10 and 17% PM2.5), the vehicles emissions (16% in both PM10 and PM2.5) and secondary sulphate (14% of the PM10 and 20% of the PM2.5), Figure 6. 6.3 Emission sources and mass contribution in Madrid, Spain Figure 7 shows the chemical profile of the sources resolved in Madrid. Six sources were identified including coal combustion, road dust, mineral, marine aerosol, sulhate and nitrate/vehicles emissions. Coal combustion was identified by the high contribution of total carbon, secondary components and arsenic. Coal combustion is still used for heating purposes in Madrid urban although in a small proportion (Artiñano et al., 2003). The road dust source was recognized by the presence of typical road dust tracers (TC, Fe, Cu) and was clearly separated by the mineral source. The vehicles emissions source was characterized by elements emitted by vehicles (Ctotal) and nitrate, a secondary component which is formed by the primary emissions of NOx by the vehicle exhaust. A sulphate source was also identified. Finally the marine aerosol source was consisted of the typical marine species Cl, Na, and a 23 proportion of Mg. Figure 8 provides the mass source contribution of the emission sources in PM10 and PM2.5. The mineral source was the highest contributors to PM10 (31%) while the nitrate/vehicles source had significant contribution in each size fraction, 18% for PM10 and 29% PM2.5, respectively. 6.4 Emission sources and mass contribution in Montelibretti, Italy Figure 9 shows the chemical profile of the sources resolved in Montelibretti. Six sources were identified: organic aerosol/biomass burning, marine aerosol, mineral, sulhate, nitrate and vehicles emissions. The first source is a mixed source representing the organic aerosol (probably secondary organic aerosol, SOA) since organic carbon has the highest levels in this profile and also biomass burning due to the presence of potassium. Again, this source had a strong seasonal variation with the highest concentrations during the cold period when biomass burning is used for residential heating. As it was expected, the mineral source was also resolved. A marine aerosol was identified by the presence of typical sea salt components. Two secondary aerosol sources were resolved, named sulphate and nitrate. Finally, the vehicle emissions source was recognized by the presence of carbonaceous components, OC, EC and some typical road dust tracers (Ca, Fe). As Montelibretti site is a regional background site it is not directly influenced by the primary vehicles emissions but rather by the mixture of both primary vehicle emissions and resuspened dust. The organic aerosol/biomass burning and the mineral source were the highest contributors to PM10 mass levles 35% and 24% respectively, Figure 10. 6.5 Emission sources in Villa Ada, Rome, Italy The chemical profiles of emission sources in Villa Ada, Rome are given in Figure 11 with the most reasonable results obtained when 6 factors were selected. Similarly to all other sites the two natural sources, marine aerosol and the mineral source were easily identified. Also the secondary components sulphate and nitrate were resolved in two separate factors. The source related to vehicles emissions was characterized by high contribution of OC but also EC a typical tracer for diesel emissions. The road dust source was characterized by high concentrations of the carbonaceous components and the metals Fe, Ca, K and Cu typical tracers of brake and tyre wear. Vehicles emissions contributed 33% 24 to PM10 mass levels, followed by the sulphate contribution 18% of PM10 mass concentrations, Figure 12. Figure 3. Chemical profile of emission sources in Huelva resolved by PMF 25 Figure 4. Mass contribution (%) of emission sources in Huelva for PM10 and PM2.5 26 Figure 5. Chemical profile of emission sources in Barcelona resolved by PMF 27 Figure 6. Mass contribution (%) of emission sources in Barcelona for PM10 and PM2.5 28 Figure 7. Chemical profile of emission sources in Madrid resolved by PMF 29 Figure 8. Mass contribution (%) of emission sources in Madrid for PM10 and PM2.5 30 1 Organic aerosol/biomass burning 0.1 0.01 0.001 1 Al Si Fe K Ca As Cr Cu Mn Ni Pb Ti V Zn Cl- NO3- SO4 Na+ NH4+ Mg OC EC Marine 0.1 0.01 0.001 1 Al Si Fe K Ca As Cr Cu Mn Ni Pb Ti V Zn Cl- NO3- SO4 Na+ NH4+ Mg OC EC Mineral 0.1 0.01 0.001 1 Al Si Fe K Ca As Cr Cu Mn Ni Pb Ti V Zn Cl- NO3- SO4 Na+ NH4+ Mg OC EC Mg OC EC Sulfate 0.1 0.01 0.001 1 Al Si Fe K Ca As Cr Cu Mn Ni Pb Ti V Zn Cl- NO3- SO4 Na+ NH4+ Nitrate 0.1 0.01 0.001 1 Al Si Fe K Ca As Cr Cu Mn Ni Pb Ti V Zn Cl- NO3- SO4 Na+ NH4+ Mg OC EC NH4+ Mg OC EC Vehicles 0.1 0.01 0.001 Al Si Fe K Ca As Cr Cu Mn Ni Pb Ti V Zn Cl- NO3- SO4 Na+ Figure 9. Chemical profile of emission sources in Montelibretti resolved by PMF 31 Figure 10. Mass contribution (%) of emission sources in Montelibretti for PM10 32 Figure 11. Chemical profile of emission sources in Villa Ada resolved by PMF 33 Figure 12. Mass contribution (%) of emission sources in Villa Ada for PM10 Conclusions: All the six sampling site where described with reference to their location and type. The instrumentation used and the chemical analyses are different among cities. The average concentration of PM sizes and chemical components are provided for all the location sampled and only species with more than 75% of their concentration values are visualized. The trend of four representative components shows that the secondary aerosol components such as SO42-, NO3-, and typical traffic tracers Cu, Zn emitted from the abrasion of tyres and breaks have a decreasing trend from 2003 to 2010, especially evident after 2006. The cities of Barcelona and Huelva are presented as they cover longer periods of PM sampling than the other sites. The source apportionment analysis identifies the vehicles emissions as the first source of PM10 in Barcelona, Madrid and Rome Villa Ada. In Huelva the first source is secondary aerosol, while in Madrid the mineral source and in Montelibretti organic aerosol, respectively, are the highest contributors to PM10. 34 References: 1. Artíñano B., Salvador P., Alonso D.G., Alastuey A., Querol X., 2003. Characterization of PM10 and PM2.5 in Madrid (Spain): An analysis of the main sources and high concentration episodes. Environmental Pollution 125, 453-465. 2. Querol X., Alastuey A., Rodrıguez S., Plana F., Ruiz C.R., Cots, N., Massague G., Puig O., 2001. PM10 and PM2.5 source apportionment in the Barcelona Metropolitan area, Catalonia, Spain. Atmospheric Environment 35, 6407–6419. 3. Paatero P., Tapper U., 1994. Positive matrix factorization: a non-negative factor model with optimal utilization of error estimates of data values. Environmetrics 5, 111–126. 4. Karanasiou A.A., Siskos P.A, Eleftheriadis K., 2009. Assessment of source apportionment by positive matrix factorization analysis on fine and coarse urban aerosol size fraction, Atmospheric Environment 43, 3385-3395. 35