Two main EC particle types, representing together more than 50 % of detected particles (58.9 % at the UB, 53.8 % at the RS), were identified at both sites. For reasons outlined below, they are named EC_Aged_R (regional) and EC_Aged_L (local). Both EC particle types presented a fine aerosol size distribution mode at both sites (about 300–500 nm; see Fig. S2 in the Supplement). Figure 2a shows the positive mass spectrum of particle type EC_Aged_R. It is dominated by EC peaks at m/z 12 [C]+, 36 [C3]+, 48 [C4]+, and 60 [C5]+. Cluster EC_Aged_L (see Fig. 2b) also shows a similar EC positive mass spectrum pattern, although a strong signal at m/z 39 dominates the positive mass spectrum. This peak is often associated with potassium [K]+, although there may also be a contribution from the organic ion [C3H3]+ (Dall'Osto et al., 2009). The EC signals present in the positive mass spectra, and the near-total absence of peaks in the negative mass spectra suggests these EC particles are not freshly emitted. Indeed, Giorio et al. (2012) reported for a regional background site that strong EC peaks in the negative ion mass spectrum are more indicative of fresh emissions while strong EC peaks in the positive mode represent aged EC. This is consistent with observations which indicate that particle composition affects the ionization and fragmentation pattern of EC (Reinard and Johnson, 2008). The two EC-rich particle types were mainly detected in stagnant air masses. As shown in Fig. S1a, an enhancement of these particle types was observed during sulfate-rich air masses (7–10 October 2010) and nitrate-rich stagnant regional air masses (13–17 October 2010), as described in Dall'Osto et al. (2013a). The diurnal profiles of these two aged EC particle types (Fig. 3a) can also help to classify their origin. Whilst EC_Aged_R shows little diurnal variation, particle type EC_Aged_L shows a gradual increase during the day, peaks in the afternoon (15:00 local time, LT), and drops to a minimum during night-time. The absence of sharp morning peaks indicates that these EC particle types are not associated with primary emission from local traffic. Similarly, the absence of a sharp evening peak also excludes the association with biomass burning, which is not expected at this time of the year and is generally not important in Barcelona (Dall'Osto et al., 2013a). Whilst the flat diurnal trend of EC_Aged_R is typical of regional aerosol, particle type EC_Aged_L suggests a local influence or an enhancement during the warmer part of the day. This assignment is also supported by the observed correlation of the EC types with some secondary inorganic particle types associated with local and regional aerosol sources, described in the next section. Similar conclusions were recently reported by Decesari et al. (2014), where most of the aged EC particles detected with the ATOFMS were related to aged anthropogenic aerosols accumulating in the lower layers of the Po Valley (Italy) overnight. Overall, we did not find major differences in the diurnal profiles of similar particle types detected simultaneously at the UB site and RS site. In other words, there were not major differences that could have been affected by wind speed and direction considering that the RS site is close to a large street canyon (Dall'Osto et al., 2013a).

The application of the ART-2a neural network algorithm to the ATOFMS data apportioned two main distinct nitrate particle types, already previously reported (Dall'Osto et al., 2009; Harrison et al., 2012; Decesari et al., 2014). The first (local nitrate (Loc-NIT); 4.2–7.3 % of particles by number) appears to be locally produced in urban locations during night-time, whilst the second (long-range transport – nitrate (LRT-NIT); 4.2–7.3 % of particles by number) is regionally transported within the Iberian Peninsula and the rest of Europe. Briefly, particle type Loc-NIT is characteristic of nitrate aerosol in small particles (Da at about 300–500 nm; Fig. S2). The average mass spectrum (Fig. 2c) shows a peak at m/z 39 that can be due to potassium, although previous studies (Dall'Osto et al., 2009) suggested that an organic contribution may be also present depending on the m/z 39 / 41 ratio. The ATOFMS nitrate particle type appears to be associated with local formation processes and occurred in the main at times outside of the long-range transport episode (Fig. S2b). By contrast, a second nitrate particle type (LRT-NIT) is regionally transported. The average mass spectrum (Fig. 2d) shows that nitrate (m/z -62) is internally mixed with sulfate (m/z -97), ammonium (m/z 18) and both elemental (m/z 36, 48, 60) and organic carbon (m/z 37, 39, 43). As explained in Decesari et al. (2014) these aerosols likely originate from the night-time condensation of nitric acid on EC-containing primary particles (e.g. Shiraiwa et al., 2007). The LRT-NIT particle type is volatile, and partially evaporates during the day leaving a core of about 300 nm mainly composed also of sulfate, elemental and organic carbon (Dall'Osto et al., 2009). Indeed, this can be seen in the LRT-SUL (long-range transport – sulfate) particle type, whose average mass spectrum for this (Fig. 2e) shows peaks due to nitrate (m/z -62), sulfate (m/z -97), elemental carbon (m/z -36, 48, 60), and organic carbon (m/z -39, 43). The LRT-SUL and LRT-NIT diurnal trends are anti-correlated, as previously described elsewhere (Dall'Osto et al., 2009; Decesari et al., 2014), with LRT-SUL concentrations peaking in the afternoon hours. Such behaviour is attributed to the effect of the diel cycle of nitric acid and ammonia condensation/evaporation on the same particle type: during night-time this regional particle type is seen with nitrate, which evaporates during daytime, leaving a smaller aerosol core composed of EC and sulfate (Decesari et al., 2014). These two particle types were more abundant at the RS site, likely because of the improved focussing of smaller particles provided by the aerodynamic lens inlet. However, this study shows a novel aspect not found in the London local–regional nitrate study (Dall'Osto et al., 2009) or in the Po Valley (Decesari et al., 2014). During the period 12–14 October, under rainy conditions and with air masses arriving from Europe (Dall'Osto et al., 2013a), the transport of LRT-SUL was detected without the nitrate component (LRT-NIT, Fig. S1b). This could be due to a different source of LRT-SUL not linked with LRT-NIT and previously unobserved.

The K-CN particle type was a minor one, representing only 1.3–2.4 % of the total particles analysed. Figure 2f shows the average mass spectrum, which features a strong peak at m/z 39 [K]+in the positive mode, as well as peaks at m/z 113 [K2Cl]+ and m/z -35 [Cl]-, suggesting a biomass-burning source (Pastor et al., 2003; Dall'Osto and Harrison, 2006). K-rich particles similar to K-CN have previously been attributed to biomass burning (Silva et al., 1999; Guazzotti et al., 2003) and were found to correlate with gas-phase measurements of acetonitrile, a good biomass-burning tracer. The negative ion mass spectrum shows strong peaks at m/z - 26 [CN]- and m/z -42 [CNO]-, indicating that the potassium and chloride are internally mixed with organonitrogen species. The size distribution, centred at about 350 nm (Fig. S2), also points to a combustion source. Overall, this particle type represented only about 2 % of the total particles sampled (Table 1) and its temporal variation is presented in Fig. S2d. Its diurnal trend (Fig. 3) tracks the anthropogenic activities of the city of Barcelona, suggesting general minor urban combustion processes. It should be noted that biogenic plant debris also has a similar single-particle mass spectrum, with strong signals at m/z 39, -26, and -42 (Silva and Prather, 2000). However, vegetative dust is usually internally mixed with sodium and phosphate and presents aerodynamic diameters above 1 µm (see Sect. 3.3).

Two sea salt particle types (fresh and aged) were detected at both sites, accounting for 9.4 and 18.3 % of the total particles sampled at RS and UB, respectively (Table 1). The higher percentage detected at the UB is likely due to the instrument configuration given that the nozzle inlet enables more efficient detection of coarser particles. The average mass spectrum (Fig. 2g) for the particle type assigned to fresh sea salt (labelled NaCl) shows peaks typical of sodium chloride clusters ([Na]+ (m/z 23), [K]+ (m/z 39), [Na2]+ (m/z 46), [Na2Cl]+ (m/z 81 and 83), [NaCl2]- (m/z 93, 95 and 97)), whilst aged sea salt (NaCl-NIT, Fig. 2h) also exhibited nitrate peaks (m/z -46 and -62) reflecting the reaction between NaCl and HNO3 and the replacement of chloride by nitrate (Gard et al., 1998). Both particle types had mean size distributions above 1 µm (Fig. S2c). NaCl was mainly detected during air masses that had travelled over Mediterranean regions (7–10 and 11–13 October), as shown in Fig. S2e. Interestingly, the diurnal variation of the two NaCl particle types is quite different (Fig. 3). Whilst NaCl-NIT does not show a clear trend, NaCl shows an enhancement in the afternoon, associated with the sea breeze, peaking at 15:00 (Dall'Osto et al., 2013c).

Two common ATOFMS particle types have average mass spectra that were dominated by metals. The first was rich in iron (type Fe, 1.4 % of the total particles) and has a spectrum (Fig. 2i) characterized by a strong signal at m/z 56 and weaker features at m/z 73 (iron oxide, [FeOH]+) and m/z 54 (isotope [54Fe]). The negative spectrum has strong features at m/z -46 and -62, indicating that the iron is internally mixed with nitrate. During SAPUSS, the Fe particle type was found to correlate (R2 = 0.75) with LRT-NIT, which is associated with long-range transport of pollutants. The small mode of the Fe particle type (Fig. S2) reflects the fact that only fine particles were likely to travel long distances relative to the coarser ones which were lost during transport. This is in line with previous ATOFMS field studies showing the transport of iron-containing particles internally mixed with nitrate from continental Europe (Harrison et al., 2012). The flat diurnal profile for the Fe particle type (Fig. 3) also suggests a regional origin. The long-range transport of fine Fe-containing particles (internally mixed with nitrate) is discussed further elsewhere (Dall'Osto et al., 2016).

The average mass spectrum of the second metal-rich particle type is shown in Fig. 2j. This particle type is labelled Pb since lead is one of the largest contributors in the positive mode, occurring at m/z +206, 207, and 208. Other peaks in the positive ion spectrum include m/z 23 [Na]+, m/z 56 [Fe]+, and m/z 39 and 113 ([K]+ and [K2Cl]+). In addition to nitrate (m/z -46 [NO2] and m/z -62 [NO3]-), chloride (m/z -35 ([Cl]-) was one of the most abundant species in the negative ion mode. Further information can be found in Dall'Osto et al. (2013b), where this ATOFMS particle type was found highly correlated with hourly elemental concentrations determined by PIXE analysis (Dall'Osto et al., 2013b), showing that this source of lead and chloride (Pb–Cl) is a major (82 %) source of fine Cl in the urban agglomerate of Barcelona. A similar particle type containing Pb–Cl was detected by ATOFMS in Mexico City and attributed to a waste incinerator source (Moffet et al., 2008), although other studies in the same area attributed lead-containing particles to multiple sources, including trash burning (Salcedo et al., 2010; Hodzic et al., 2012). It is worth noting that whilst the regional particle type Fe was mainly distributed in the fine mode (about 300–500 nm), a much larger mode (about 700–900 nm) was observed for this local ATOFMS Pb particle type (see Fig. S2d). Finally, the temporal trend (Fig. S1d) and the diurnal profile (Fig. 3) of the Pb particle type also suggests a local origin, likely related to emissions from urban combustion.

A particle type containing vanadium (m/z 51 [V]+ and m/z 67 [VO]+) Na (m/z 23) and Fe (m/z 56), along with minor peaks due to EC at m/z 36, 48, and 60, was observed at the UB site. A negative ion mass spectrum was not acquired for this particle type (Fig. 4a). It has been shown that the condensation of secondary material on soot particles (Moffet and Prather, 2009) and the consequent change in aerosol hygroscopicity (Spencer et al., 2006) can suppress the formation of negative ions in real-time laser desorption/ionization mass spectrometry, causing many aged EC-containing particles to lack negative mass spectra (Neubauer et al., 1998). The V particle type accounted for only 1 % of the particles characterized at the UB, with a unimodal distribution peaking at about 350 nm pointing to a fresh combustion emission source. Field observations have confirmed that ships produce significant amounts of soot, vanadium, nickel, and sulfate (Pattanaik et al., 2007). Previous ATOFMS field measurements also reported that single particles containing organic carbon, vanadium, and sulfate (OC-V-sulfate) resulted from residual fuel combustion (i.e. bunker fuel), whereas high quantities of fresh soot particles represented distinct markers for plumes from distillate fuel combustion (i.e. diesel fuel) from ships as well as trucks in the port area (Healy et al., 2009; Ault et al., 2010). Indeed, the V particle type presented a temporal variation showing a maximum on 3 October 2010, when strong wind was blowing from the port of Barcelona (North African tropical air masses from the west (NAF_W); Fig. S1e). This Vanadium particle type is therefore most likely associated with fuel oil and ship emissions from the port. However, it was not detected at the RS site, possibly due to local street canyon effects which restrict transport and mixing in the area (Solazzo et al., 2008).

The average ATOFMS mass spectrum of a dust particle type rich in calcium is shown in Fig. 4b, where peaks for calcium (m/z 40 [Ca]+, 56 [CaO]+, 57 [CaOH]+, and 96 [Ca2O]+) are present. Peaks associated with sodium (m/z 23), magnesium (m/z 24, 25), and potassium (m/z 39) can also be seen. The Soil-Ca particle class represented 0.9 % of the particles sampled at the UB site and was found to moderately correlate (R2=0.6) with the urban dust (Ca) factor obtained by PIXE-PMF analysis (Dall'Osto et al., 2013b). The large aerosol size mode above 1 µm (Fig. S1e) also points to a dust origin. Enhanced levels of the Ca ATOFMS particle type were observed during ATL air masses (Dall'Osto et al., 2013a), likely due to extra resuspension of dust caused by the high wind speeds during these periods. Urban areas in the south of Europe are known to have high dust loadings and three different types of dust were previously reported in the SAPUSS PIXE study (Dall'Osto et al., 2013b), representing ca. 25 % of the PM2.5 mass concentrations measured. It should be noted that this particle type did not possess a negative ion mass spectrum and was also not detected at the RS site, likely for the same reasons explained above (Sect. 3.3.1).

A rarely observed particle type with peaks due to titanium at m/z 48 and 64 ([Ti]+ and [TiO]+, respectively) was found to represent 0.9 % of the particles sampled at the UB site (Fig. 4c). Additional peaks are associated with other metals including Na (m/z 23), Al (m/z 27), Ca (m/z 40, 56, 96), and Fe (m/z 56), as well as minor peaks of silicate at m/z -60 [SiO2]- and -76 [SiO3]-. These particles presented an aerosol size coarse mode (1–3 µm; Fig. S2) and were detected mainly during the period 8–10 October, when Barcelona experienced air masses originating in the North African Saharan region (Dall'Osto et al., 2013a). It is likely that the Art-2a algorithm did not apportion this minor particle type of regional origin in the RS site. With regard to secondary species, it is interesting to note that the dust was internally mixed only with nitrate and not with sulfate (Fig. 4c). Previous studies (Dall'Osto et al., 2010) showed that Saharan dust particles collected near the Cape Verde Islands contained internally mixed nitrate but no sulfate, whilst Saharan dust particles collected on the coast of Ireland showed a very high degree of internally mixed secondary species including nitrate, sulfate, and methanesulfonate (Dall'Osto et al., 2004).

The ATOFMS has already proven to be a good tool for identifying and separating dust (mainly Ca-rich or Al-Si-rich) and biological particles (Fergenson et al., 2004). A particle type dominated by K (m/z 39) and phosphate (m/z -63 and -79) in the average mass spectrum (Fig. 4d) accounted for 1.7 % of the total particles sampled at the UB site. Strong signals at m/z 26 and 42, due to [CN]- and [CNO]-, are also present, as well as peaks due to Na (m/z 23) and Fe (m/z 56). A coarse aerosol size mode of about 3 µm (Fig. S2) is likely again the reason why this particle type was not detected by the ATOFMS at the RS site. The size and chemical composition suggest a source such as vegetative debris and the particle type is thus labelled Veg K-P. A previous ATOFMS study reported a very similar particle type from samples of leaves collected from a roadside (Schofield, 2004). It is interesting to note that an increase of this vegetative debris particle type was detected under NAF_W air masses in concomitance with Saharan dust particles. Previous atmospheric measurements have shown that the concentration of bacteria over the sea may be much lower than over land but that higher concentrations of aerosolized microorganisms are generated during dust events compared to clean background marine conditions (Kellog and Griffin, 2006; Prospero et al., 2005).

Amines are ubiquitous in the atmospheric environment and have been detected in marine, urban, and rural atmosphere in the gas and particle phases as well as aqueous fog and rain water (Ge et al., 2011). The ATOFMS is a particularly good aerosol instrument for studying amines because the LDI laser wavelength (266 nm) ionizes them very efficiently (Angelino et al., 2001; Healy et al., 2015). During the SAPUSS measurement campaign, four amine particle types were detected at the RS. To the best of our knowledge, this is the first time such a variety of organonitrogen particle types has been detected at the single-particle level in real time in urban air. a.

Amine (POA 58): Fig. 5a shows the ATOFMS mass spectra of a particle type called amine-58, which accounted for 0.8 % of the total particles classified at RS. The strongest peak in the positive mass spectra at m/z 58 is likely due to [C2H5NH = CH2]+. In general, the most important primary fragmentation process occurring for aliphatic amines involves the removal of one of the electrons from the lone pair on N and cleavage of the C–C bond to the nitrogen, with loss of the heavier alkyl group favoured. This cleavage explains the presence of the fragments at m/z 58 [C2H5NH = CH2]+, 72 [C3H7NH = CH2]+, 86 [(C2H5)2N = CH2]+, and 114 [(C3H7)2N = CH2]+. The minor peak at m/z 86 was found to correlate (R2=0.8) with m/z 58. m/z 86 was suggested as amine species associated with fresh mobile emissions (Angelino et al., 2001). The negative mass spectrum shows nitrate and sulfate internally mixed with this particle type, although the former has a stronger signal than the latter. A size distribution peaking at about 300–400 nm was associated with this particle type. The diurnal variation of amine (POA 58) (Fig. 6a) shows a peak about 1 h later than morning and evening rush hour traffic and thus points to relatively fresh vehicular source. In conclusion, both the mass spectra (similar to previous traffic emissions; Angelino et al., 2001) and the temporal trend of this particle type point to traffic emission as a main source.

b.

Amine (ETS 84): Fig. 5b shows the average mass spectrum of a particle type associated with environment tobacco smoke (ETS) called amine-84, which represented 0.5 % of the total particles classified at RS. In the positive mass spectrum, the peak at m/z 161 is assigned to nicotine, with the main fragment being at m/z 84 (methyl-pyrrole fragment). The negative mass spectrum shows nitrate and sulfate internally mixed with this particle type. The amine-84 mass spectrum is very similar to that previously reported in Athens (R2 = 0.9; Dall'Osto et al., 2007) and also attributed to tobacco smoke. This particle type generally followed human activity within the city (Fig. 6a), with a main peak in the morning (10:00) and a broader one during the evening (18:00–22:00 LT). It should be noted that an afternoon peak – potentially related to lunch activities – was not noticed. The ATOFMS cluster amine (ETS 84) attributed to tobacco smoke was compared to the measured nicotine value described in Alier et al. (2013). Briefly, much higher nicotine concentrations (58 ng m-3) were observed at the RS site than at UB site (7 ng m-3). This 8-fold difference pointed to a very high level of outdoor cigarette consumption near the RS site, which is situated next to a busy street and an exit of an underground metro station (Alier et al., 2013). Moreover, nicotine concentrations were more affected by daytime working activities during weekdays rather than during the weekends. ATOFMS cluster amine (ETS 84) was temporally correlated with nicotine concentrations available at 12 h resolution, and an R2 of 0.65 was found, confirming the tobacco smoke source suggested. Spain ranks among the countries with the highest levels of cigarette consumption in the European Union and in the World (WHO, 2004; Alier et al., 2013). High outdoor gas-phase nicotine concentrations (0.5 and 1.5 µg m-3) have been previously measured in Barcelona in summer 2010 (Sureda et al., 2012).

Overall, the four amine-containing particle types were found distributed mainly in the sub-micron mode. Amine (POA 58) and amine (ETS 84), which are attributed to primary emissions from traffic and ETS respectively, have similar size distributions which peak around 400–500 nm (Fig. S2). Amine (SOA 59) and amine (SOA 114) were also found in the same size range, but presented a broader shape, suggesting partial condensation of SOA material on existing particles. Previous work of Angelino et al. (2001) used the peak at m/z 118 to track a marker for the oxidation products of alkylamines observed in the condensed phase of aerosol particles, specifically attributed to the oxidation of triethylamine (m/z 118; [(C2H5)2NCH2]+). The analysis of the ATOFMS dataset for the RS site shows that the majority (about 80 %) of amine particle types containing a peak at m/z 118 were found in the tobacco-related ATOFMS amine (ETS 84) class, whereas the remaining 20 % was associated with amine (POA 58). This suggests – at least in our study – that m/z 118 cannot be used as a marker for amines produced by secondary processes. Other possible identification of higher mass ions detected and associated with organic nitrate include m/z 74 ([C2H4NO2]+), 88 ([C3H6NO2]+ or (([C2H5)2NO]+), 104 ([C3H6NO3]+), and 191([C6H11N2O5]+). Overall, m/z 104 was found to be to be linked to amine (POA 58), whereas m/z 74 and 191 were related to amine (SOA 114). Finally, m/z 88 was found associated with both amine (POA 58) and amine (SOA 114). In summary, organic nitrate peaks were associated with both primary and secondary amine particle types.

The high efficiency of the ATOFMS equipped with the aerodynamic lens and deployed at the RS allowed us to characterize four particle types rich in organic compounds. a.

Org (Lub. Oil): Fig. 7a shows the average mass spectrum of a particle type named “Lub. Oil” (1.8 %). The positive ion mass spectrum shows strong peaks for [Ca]+ (m/z 40) and minor ones for OC (m/z 27, 43, 57) and EC (m/z 12, 24, 36, 48, 60). The negative mass spectrum shows the presence of nitrate (m/z -46, -62) and EC (-36, -48, -60). This particle type was only detected in the lower size range of the ATOFMS (ca. 150–400 nm), was found to spike during traffic rush hour times (Fig. 6b), and is attributed to lubricating oil from traffic emissions (Spencer et al., 2006; Drewnick et al., 2008).