Barcelona is located on the north-eastern coast of the Iberian Peninsula in the Western Mediterranean Basin. This region is a model case of environmental challenges in transition zones between temperate and arid climates. The WMB experiences intense solar radiation throughout the year, leading to high photochemical activity, even in winter. This solar activity interacts with the biogenic VOCs emitted by the region's abundant vegetation, resulting in the formation of large amounts of SOA (Ciccioli et al., 2023; van Drooge et al., 2018). Additionally, the PM mass and composition in the WMB are often altered by the transport of mineral dust from the arid regions of North Africa worsening the air quality in the basin (Querol et al., 2009; Schepanski et al., 2016).

The synoptic and regional meteorological scale affecting the WMB is mainly driven by the Azores high-pressure system, the Saharan and Iberian low-pressures, the topography, and the heat-moisture regulation by the Mediterranean Sea. The interactions between these factors lead to highly contrasting temperature, humidity and rainfall conditions among seasons. Barcelona is geographically limited by two rivers on the east and west sides, by the Mediterranean Sea in the south and by the Collserola hills in the north (512 ma.s.l.). Due to this orography, the city is typically influenced by sea breezes that flow inland during the day as solar heating enhances the vertical ascent of warmer air over the city creating a surface horizonal thermal gradient. At night, the opposite occurs as the cooler air above the city subsides. In the absence of large-scale meteorological phenomena, this can create a recirculation pattern, leading to distinct air masses at the surface and upper levels.

The city of Barcelona is home to 1.7 million inhabitants, while its metropolitan area hosts around 3.7 million people, involving one of the highest population densities in Europe (16 339 hab.km-2). The car density in the city is around 6000 km-2 with a daily vehicle flux of 350 000 in the city centre and up to 151 000 in the main ring roads (Ajuntament de Barcelona, 2024). Moreover, the activities in the nearby harbour and airport, the agricultural works in the Llobregat river delta and the emissions from several industrial zones and power plants in the metropolitan area are additional potential sources influencing the air quality of the city (Dall'Osto et al., 2013).

In the present study, PM10 samples were collected at two background stations at different altitudes (Fig. 1) to assess how interactions between specific meteorological conditions and multisource particle emissions influence the organic composition of airborne aerosols in the city. One station, located at 81 ma.s.l., represents the urban background site at IDAEA (city site), while the other, at 415 ma.s.l., is situated atop the Collserola hills overlooking the city (elevated site). The two sites are separated by a horizontal distance of 3.5 km. The urban background site is situated within the city airshed while the atmosphere in the elevated site can be located either within the urban airshed or above the atmospheric layer influenced by the city in the case of low planetary boundary layer height (PBLH) because of strong temperature inversions.

PM10 samples (n=52) were collected at 12 h intervals on 150 mm-diameter quartz filters (Pall Corporation, USA) by means of a high-volume sampler (DHA-80, Digitel, Switzerland) equipped with a PM10 inlet. Filters were pre-baked at 450 °C before use. The instrument air flow was 500 Lmin-1, providing a final sample volume of 360 m3. Sample collection was performed simultaneously at both sites during two one-week campaigns in the warm and cold periods (26 April–3 May 2022 and 31 January–6 February 2023, respectively). Daytime samples were taken from 07:00 to 19:00 UTC (09:00–21:00 and 08:00–20:00 LT during the warm and cold periods, respectively). The nighttime samples were collected from 19:00 to 07:00 UTC.

The analytical procedure to determine the concentration of organic compounds in the samples was performed as described in previous studies (Fontal et al., 2015; Jaén et al., 2021b, 2023). Briefly, two 45 mm-diameter punches of the whole filter were spiked with deuterated PAH (, , , , , , , , , , , ), and . Then, the filter fractions were extracted with a dichloromethane : methanol 1:1 v/v mixture by ultra-sonication (3mL×10 mL; 15 min). The obtained extracts were filtered with glass-microfibre discs and concentrated to 0.5 mL by means of roto-evaporators and with a nitrogen stream. A 25 µL aliquot of the extract was transferred to a conical vial, evaporated to dryness and treated with 25 µL of BSTFA and 10 µL of pyridine to obtain the trimethylsilyl derivatives of the compounds with hydroxyl groups. This aliquot was injected to a GC-MS (8290 GC with 5975 MSD, Agilent Technologies) in full scan mode to quantify the polar compounds (isoprene or α-pinene SOA products, dicarboxylic acids and saccharides). A liquid–liquid extraction was performed with the remaining extract with n-hexane (3mL×0.5 mL) to analyse the non-polar fraction (polycyclic aromatic hydrocarbons (PAHs), their methyl (mPAH) and oxygenated (oxyPAH) derivatives, alkanes and hopanes). The hexane extract was concentrated with a nitrogen stream to a final volume of 25 µL prior to its injection in a Q-Exactive GC Orbitrap MS (Agilent Technologies) in full scan mode. The chromatographic separation was obtained with a HP-5MS 60 m capillary column for the polar compounds and with a HP-5MS 30 m column for the non-polar, in order to achieve optimum separation compounds of each group.

Two additional 45 mm-diameter punches of the filter were spiked with 9-nitroanthracene-d₉ and extracted with 7 mL of DCM by ultra-sonication during 20 min for the analyses of nitro derivatives of PAH (nitro-PAH). The extract was concentrated with a nitrogen stream and then filtered with anhydrous sodium sulphate contained in a Pasteur pipette with glass wool. The filtered extract was further evaporated to 50 µL with a nitrogen stream, 50 µL of nonane were added to the extract and the concentration continued to evaporate the remaining DCM. The extracts were analyzed by GC-MS using an Agilent 5973N Inert EI/CI Mass Spec Selective Detector with a Agilent 6890N Gas Chromatograph) operated in SIM and in negative ion chemical ionisation (NICI) mode equipped with a Restek Rxi-5Sil MS 60 m column (2022 campaing) or by GC–MS/MS using an Agilent 7000 Series Triple Quad) equipped with a HP-5MS 60 m capillary column (2023 campaign). In this last case, the instrument operated under electron impact ionization and multiple reaction monitoring (MRM) mode was used for acquisition. These instrumental methodologies provided equivalent nitro-PAH results.

Sixty-eight compounds were identified and quantified based on their chromatographic retention times, mass spectra and calibration with analytical internal standards and 5-point calibration curves. Field blanks of each sampling campaign and site were collected and treated together with the sample filters. The average blank levels were subtracted from the sample concentrations. The method limit of detection (MLOD), which reflects the entire analytical procedure including sampling, extraction, and analysis, was determined as the average blank level plus three times the standard deviation of blanks (Table 1). In cases where blank values were below the instrumental limit of detection (ILOD), the ILOD was applied instead. ILOD refers specifically to the instrumental performance and was set based on a minimum signal-to-noise ratio of 3 calculated from the lowest point of the calibration curve. To process the data, concentrations below MLOD were replaced by half of this value.

NO, NO2, O3 and PM10 hourly values were obtained from the two air quality stations of the Atmospheric Pollution Monitoring and Forecasting Network situated in the same sites as the PM10 samplers: Palau Reial, IDAEA-CSIC, for the city site and Observatori Fabra for the elevated sub-urban site (Generalitat de Catalunya, 2024a) (Fig. S1 in the Supplement).

Meteorological parameters, including temperature, pressure, relative humidity, precipitation, solar radiation, mean wind velocity (10 m), mean wind direction (10 m) and solar radiation, were obtained every half an hour from two Automatic Weather Stations of the Catalan Meteorological Service (Generalitat de Catalunya, 2024b). These stations were located near the elevated and city sampling sites. The elevated station was adjacent to the corresponding sampling site, while the city station was situated approximately 1.2 km away from the PM sampling site (Fig. S1). This spatial separation may introduce slight differences between the recorded meteorological parameters and those directly affecting the sampled air masses. However, the station shares the same urban background classification as the sampling site and was considered representative of the meteorological conditions. Complementary meteorological data nearby the Barcelona harbour was obtained from a station located in this installation (Figs. S1 and S2 in the Supplement).

PBLHs were estimated from a CL31Vaisala Ceilometer set located at 98 ma.s.l. and at 340 m in horizontal distance from the city site (Fig. S1). The original data was processed using the Vaisala Boundary-Layer View software (BL-VIEW) Enhanced Gradient method (García-Dalmau et al., 2021; Vaisala, 2024) followed by a selection algorithm based on the methodology of Lotteraner and Piringer (2016) to obtain PBLH values every 10 min.

Backward air trajectories for 96 h were computed with the NOAA HYSPLIT model (Rolph et al., 2017; Stein et al., 2015) provided with NCEP/NCAR Global Reanalysis Data (Kalnay et al., 1996) with a horizontal resolution of 2.5° as meteorological files. The trajectories were computed hourly for the two sampling campaigns at the city site at 100 m above ground level (a.g.l.), at the elevated site at 400 ma.g.l., and at the middle point between both sites at 200 ma.g.l. as starting locations. Moreover, a 48 h-cluster analysis joining the two sets of trajectories (the two sampling periods) was performed to elucidate the air mass sources reaching Barcelona at a mesoscale to synoptic meteorological scale. This analysis provided the origins of air masses during the sampling periods. The clustering algorithm groups similar trajectories in clusters and represent them by their mean trajectory. The optimal number of clusters was 5 considering the percentage change in total spatial variance through a step-wise reduction in cluster numbers (92 % from 5 to 4 clusters in the last case). Moreover, the geographical interpretation of the clusters was also taken into consideration. Given the proximity between sampling sites, a lack of significant differences between different altitudes of ending point back-trajectories was observed (Fig. S3 in the Supplement). Therefore, the middle point (Fig. S3b) was used for the work description.

Multivariate Curve Resolution–Alternating Least Square (MCR-ALS 2.0) was applied to the dataset (67 variables and 52 samples) using MATLAB (Jaumot et al., 2005, 2015; Tauler, 1995). The MCR-ALS method is based on the D=CST+E matrix equation to decompose the initial matrix D (normalized dataset) into a reduced number of components. The output gives: C (a matrix with sample scores for each component), ST (a matrix with compound loadings for each component; profiles) and E (a matrix with residual non-explained data) The decomposition was performed under non-negativity constraints, which provide physically interpretable results and are more realistic for environmental data. The number of components was selected based on the interpretability of their chemical profiles in terms of emission sources and atmospheric processes.

Compared with other source apportionment approaches, such as Chemical Mass Balance (CMB), Principal Component Analysis (PCA), or Positive Matrix Factorization (PMF), MCR-ALS offers several advantages. CMB requires predefined source profiles and cannot resolve unknown sources, while PCA enforces orthogonality, which limits environmental interpretability. Both MCR-ALS and PMF apply non-negativity constraints and yield comparable results, but they differ in their optimization algorithms and normalization procedures (Tauler et al., 2009). Importantly, MCR-ALS does not impose orthogonality between components, allowing for overlapping explained variance that better reflects the reality of atmospheric sources, which are rarely independent. Previous studies have demonstrated the robustness of this method for source apportionment of air pollutants and organic aerosols (Jaén et al., 2021b, 2023; van Drooge et al., 2022).

The temporal evolution of potential temperature (θ) provides insight into the vertical mixing and stability of the atmosphere at the sampling sites. During both campaigns, θ exhibited similar trends at both altitudes (Fig. S4 in the Supplement). During the day, temperature and relative humidity (RH) remained comparable at the two sites, indicating that they were within the convective mixing layer of the urban airshed. Temperature peaks at the elevated site were slightly delayed relative to the city site, reflecting the progressive growth of the mixing layer in the hours following sunrise. At night, more pronounced variantion in temperature and RH were observed between the two sites, suggesting that the elevated site become partially decoupled from the urban airshed. Generally, θ was higher at the elevated site at night, consistent with the development of stable atmospheric conditions and temperature inversions. This day-night dynamic was also reflected in variations of wind speed and direction, showing similar patterns during the day but becoming stronger and originating from a different direction at the elevated site during the night compared to the city site (Fig. S4).

Regarding the air quality parameters, ozone exhibited a clear day-night contrast between sites. During the day, O3 concentrations were all well correlated between the city site and the elevated site (r2=0.7) (Figs. S4 and S5 in the Supplement), reflecting a well-mixed urban atmosphere. At night, this correlation weakened substantially, indicating a decoupling of the elevated site from the urban airshed. O3 concentrations remained consistently high at the elevated background site throughout both day and night (78±15 µgm-3), while at the city site, O3 exhibited a diurnal oscillation peaking during sunlight hours (47±19 µgm-3). These observations are consistent with previous studies in the same sampling sites across different seasons (Dall'Osto et al., 2013) and suggest that differences in oxidant distribution along the vertical column may influence the organic composition at the studied altitudes. The nitrogen oxides also showed substantial correlations between sites during the day, although always higher in the city, but they were uncorrelated at night (Figs. S4 and S5). Their diurnal trends at the city site were strongly linked to traffic emissions, with pronounced peaks in the early morning and late afternoon (up to 107 µgm-3) but those high concentrations reached the elevated site only under mixing conditions of the atmosphere (25±19 µgm-3 city site; 10±10 µgm-3 elevated site). PM10 had moderate correlations between sites during both periods of the day (r2=0.32–035, Figs. S4 and S5) and the overall concentration trends remained similar (18±7 µgm-3 city site; 16±7 µgm-3 elevated site), although the slightly higher warm-period PM10 at the elevated site may be another indication for variations in their organic composition.

The 5 clusters of 48 h-backward trajectories for both sampling periods resolved with the HYSPLIT clustering algorithm are showed in Fig. S3b. The first (warm) sampling period was dominated by southerly trajectories (red cluster) and trajectories coming from the east (dark blue cluster). Both clusters have their origins in the Mediterranean Sea indicting regional recirculation. The end of this period was dominated by northerly trajectories coming from the European continent (light blue cluster). This change in the air mass origin from regional to continental was reflected in the air quality (Fig. S4) involving a reduction of NO, NO2 and PM10 peaks when the air was from the north. In these conditions, the differences in O3 between both sampling sites were reduced. In the second (cold) sampling campaign, all trajectories reached Barcelona from the north, the first period coming more directly from the Atlantic Ocean (green cluster) while trajectories coming from the gulf of Lion were observed in the last sampling days (pink cluster). This change was again reflected in lower pollutant concentrations at the end of the campaign.

Table 1 shows detection frequencies (DF), MLOD, average, maximum and minimum concentrations at the city and elevated background sites and p-values for the Wilcoxon test for paired samples (WPS) between both sites, and average day and nighttime concentrations are given in Table S1 in the Supplement. The magnitude of these differences between sites are evaluated by calculating the Decreasing Index for each compound in each pair of samples as following: DIi=Celevated sitei-Curban sitei∑1nCj/n; where i is the specific pair of samples, and n the total number of samples. The average DIs for day and night samples are shown in Fig. 2.

The average concentrations of individual PAHs ranged from 8 to 230 pgm-3 at the city site and from 6 to 152 pgm-3 at the elevated site. These compounds are generated by incomplete combustion, mainly from biomass and fossil fuels. Their concentrations were higher in the city site compared to the elevated site (p<0.05). Similar trends were observed for the methyl-PAHs which are usually emitted alongside PAHs under relatively low temperature combustion conditions. The average levels of individual methyl-PAHs varied between 5 and 183 pgm-3 at the city site and between 2 and 157 pgm-3 at the elevated site. Retene (RET), which is an indicator of pinewood combustion in the absence of coal combustion (Bi et al., 2008; Ramdahl, 1983; Shen et al., 2012), was the more abundant methyl-PAH, with similar concentrations as predominant unsubstituted PAHs (pyrene (PYR), benzofluoranthenes (BBJKFL) and benzo[ghi]perylene (BGHIP)), which suggests contribution of pinewood combustion in the sites, since there is no coal combustion in the region (van Drooge and Grimalt, 2015).

As regards to oxygenated PAHs, the average concentrations of the individual compounds ranged from 19 to 314 and from 10 to 289 pgm-3 at the city and elevated sites, respectively. The most abundant compound was 9,10-anthraquinone (ANQ) followed by 2-methylanthraquinone (2mANQ) and fluoren-9-one (9FLO), which have been associated with several primary combustion emissions, but also to secondary transformation processes, such as the reaction of anthracene and fluorene with OH, NOx or O3 (Keyte et al., 2013; Kojima et al., 2010; Oda et al., 1998; Rogge et al., 1993). The fact that their concentrations are similar at both sites suggest that secondary formation processes contribute substantially to their presence. In contrast, the other oxyPAH are significantly higher in concentration at the city level, indicating stronger influence of local primary emissions.

The two nitro derivatives of PAHs reported in this study exhibited relatively low concentrations, but with higher 9-nitroanthracene concentration in the city site than the elevated site (11 and 5 pgm-3, respectively). These compounds can be primary emitted from combustion sources, but also formed from anthracene reactions in the presence of NOx (Bandowe et al., 2014; Saldarriaga et al., 2008), and their significant higher concentrations in the city site indicates a priory a dominant influence of primary emissions over formation processing.

In almost all cases, the decreasing indexes of PAHs and derivatives were negatives for both day and night periods agreeing with higher concentrations at the city level (Fig. 2). However, the DIs were more accentuated for nighttime samples, up to -1.3, indicating larger differences between sites at night when air masses are disconnected.

Hopanes are related to traffic emissions for their presence in lubricant oils, although they are also emitted through coal combustion (Bi et al., 2008; Schauer et al., 2007). In the present study, hopanes are traffic markers due to the absence of coal combustion in the studied region, and they showed an average concentration of 117 pgm-3 at the city site and 42 pgm-3 at the elevated site (Table 1), with negative DIs for both day and night samples (Fig. 2).

The anhydro-saccharides galactosan (GAL), mannosan (MANNO) and levoglucosan (LEV) are marker compounds of biomass burning (Simoneit, 2002). They showed predominance of LEV, with average concentrations at the city and elevated sites of 80 and 66 ngm-3, respectively. The concentrations in the cold period (up to 300 ngm-3) were ten times higher than the warm period, which is similar to results in previous studies in the urban area of Barcelona (Reche et al., 2012; van Drooge et al., 2018). Nighttime DIs were around -0.4, indicating an accumulation of these combustion products in the city while they were close to 0 or positive during the day, pointing to vertical mixing in the urban airshed (Fig. 2). The substantial concentrations of these biomass combustion markers at the elevated site at nighttime in the cold period suggest an influence of local emissions besides and influence of recirculation of biomass burning aerosols (BBOA) in the region.

The 12 dicarboxylic acids (DCA) that were analysed in this study show different trends than the combustion products mentioned before. The C2–C9 linear homologues, two hydroxylated compounds (malic acid (MA) and tartaric acid (TARA)) and phthalic acid (PHA) are mainly formed through photochemical reactions of VOCs from natural and anthropogenic primary emission sources such as vehicular exhaust, food cooking, or biogenic emissions (Cao et al., 2017; Kunwar and Kawamura, 2014; Lui et al., 2023; Rogge et al., 1991; Xu et al., 2020). Their average individual concentrations span from 0.3 to 17 ngm-3 at the city site and from 0.3 to 21 ngm-3 at the elevated site. Malonic acid and tartaric acid were higher at the elevated site, while pimelic, suberic, azelaic and phthalic acids were more abundant at the city site. In previous studies, these later longer chained C7–C9 DCAs had higher concentrations in an intensive traffic site in the populated city centre and were linked to the oxidation of unsaturated carboxylic acids, such as oleic acid (Alier et al., 2013; Kawamura and Gagosian, 1987). Other dicarboxylic acids, such as succinic acid, glutaric acid, and phthalic acid did not show statistically significant differences in concentration between the two sites.

Secondary organic aerosol markers from biogenic volatile compounds, i.e. isoprene (methyltetrols (MT) and 2-methyl glyceric acid (MGA)) and α-pinene (cis-pinonic acid (CPA), 3-methyl butane tricarboxylic acid (MBTCA) and 3-hydroxyglutaic acid (HGA)) oxidation products (Claeys et al., 2004, 2007; Szmigielski et al., 2007), ranged from 1 to 14 ngm-3 at the city site and from 1 to 27 ngm-3 at the elevated site. Their concentrations were around ten times higher in the warm period compared to the cold period, relating these compounds to the period with highest biogenic (vegetation) activity and photo-chemical reactions. Higher levels were observed at the elevated site compared to the city site, with the largest difference for CPA which was also the most abundant one. CPA is a fresh SOA from α-pinene oxidation product while 3HGA and MBTCA are further oxidation products (Claeys et al., 2007; Szmigielski et al., 2007). CPA decreasing index was 0.4 for daytime samples and 0.8 for nighttime samples, suggesting further oxidation of CPA in the city compared to the elevated site. This may be related to enhanced formation of aged pinene SOA in the presence of NOx from traffic emissions as has been observed in previous studies in the urban background in Barcelona (Alier et al., 2013; Minguillón et al., 2016; van Drooge et al., 2018). On the other hand, the presence of pine forests in the Collserola hills may also be a potential source for α-pinene emissions and CPA formation.

Lastly, other biogenic saccharides were also determined in this study which can be related to fungal sporulation (mannitol (MANNI) and sorbitol (SOR)), pollen grain and organic soil dust (glucoses (AGLU and BGLU) and meso-erythritol (MERY)) which are relevant in the PM coarse fraction (Bauer et al., 2008; Burshtein et al., 2011; Jia and Fraser, 2011; Medeiros and Simoneit, 2008; Simoneit et al., 2004). Their individual average levels ranged between 1 and 24 ngm-3 in the city site and between 1 and 42 ngm-3 at the elevated site. Generally, they showed higher concentrations at the elevated site agreeing with the higher density of vegetation in this site compared to the city site. However, this difference was only significant for MANNI, AGLU and BGLU which showed high positive DIs during the day (0.8–1; Fig. 2).

Table S2 in the Supplement compares the concentrations of the studied compounds at both sites and campaigns with similar studies in the region in urban, sub-urban, rural, industrial, and high-altitude areas. In general, the levels of the studied compounds were in the range of these former studies in terms of location and season. In particular, the concentration of combustion markers (biomass burning markers, PAHs and derivatives) in the cold period campaign were similar to urban and sub-urban sites in cold periods, but lower than those registered in rural areas with local biomass burning emission sources. The hopanes were lower than recorded in suburban, industrial and traffic sites under stagnant atmospheric conditions, while the biogenic SOA marker concentrations were lower than those observed in rural sites.

Regarding nitro-PAHs, few studies report their PM concentrations in southern Europe. In fact, the only measurements in ambient air in Barcelona were those by Bayona et al. (1994), conducted at heavily trafficked sites. They reported considerably higher concentrations of 9-nitroanthracene and 2-nitrofluorene than those observed in the present study, a pattern also reflected in the parent PAHs. In contrast, the oxy-PAH ANQ, which was also measured in that study, was found at lower concentrations. Studies conducted in European cities during the early 2010s similarly reported higher nitro-PAH concentrations than those observed here, particularly in winter (Alam et al., 2015; Alves et al., 2017; Tomaz et al., 2016), whereas more recent studies on the Iberian Peninsula reported comparable levels (Lara et al., 2022). Although further measurements are necessary, reduction of nitro-PAHs concentrations may be attributable to reductions in NO2 emissions over the past decade, resulting from the implementation of low-emission zones and cleaner vehicle engines, which have decreased atmospheric formation of nitro-PAHs while potentially enhancing oxy-PAH formation through reactions with O3.

The multi-variant analysis by MCR-ALS was applied on the joined dataset, and included the organic markers in PM10 samples, as well as the average concentrations of the air quality parameters NO, NO2, O3, and PM10 for the sample periods. This was done in order to study the similarities and differences among the quantified compounds and sampled sites and to identify the most significant sources and atmospheric processes influencing air quality in the sites without pretending to be a mass balance for PM, or the organic aerosol. The analysis resolved six components based on their chemical composition in relation to their environmental interpretability. Those components explained 95 % of the variance in the dataset. The components are represented in Fig. 3 showing the loadings of each compound in each component (bars) and the percentage of the compound present in the specific component as the proportion of the compound along all components (purple dots). Moreover, Fig. 4 shows the score of the resolved components in each sample. The six components are as follows:

As expected, NO and NO2 were highly represented in the traffic components, comprising 83 % and 68 % of the total loadings of these compounds, respectively. O3 was related with secondary components, 52 % and 29 % for the fresh and regional, respectively. PM10 was represented in multiple components, namely secondary (45 %) but also traffic (22 %) and Biomass Burning (17 %).

The highest PAH loading (46 %) was related to biomass burning component which contrasts with observations one decade ago that attributed these pollutants to traffic emissions (Alier et al., 2013; van Drooge et al., 2018). Although these previous studies also included urban traffic sites, a general decrease of benzo[a]pyrene has been observed in the Air Quality network with a 73 % reduction in urban traffic sites between period 2010–2015 and period 2020–2023, compared to a smaller decrease of 28 % in urban background sites. This change reflects improvement in engine efficiency of motorized vehicles and the traffic restrictions in the low emission zone, in combination with an increase of biomass burning emissions during the past decade in the background sites. In contrast, the sum of the two traffic components represents the highest loadings of the PAHs derivatives, they account for the 56 % of mPAH (excluding RET), the 46 % of oxyPAHs and the 62 % of nitro-PAHs. In fact, the individual oxidized compounds show some significant differences in their abundances among the components. The more volatile oxyPAHs, 9FLO and ANQ, are probably mainly secondary, formed with the Fresh secondary aerosol formation while 23dmANQ and BAF are associated with the diurnal traffic. For the nitro-PAHs, 2nFLU is probably secondary formed after diurnal traffic emissions (46 %) while 9nANT is predominantly related to nocturnal traffic (51 %).

The polar fraction, excluding saccharides, is related with the secondary and biogenic components. In particular, most of the DCA and oleic acid have a regional origin, 39 % and 43 %, respectively, while biogenic SOA compounds (isoprene and α-pinene SOA) are more uniformly distributed between the non-combustion sources. Finally, the glucoses and fungal markers have almost solely relevant loadings from the biogenic component with percentages from 42 % to 97 % in this component.

The relative contribution from each component to the sum of scores varied significantly by season, site and period of the day (Fig. S7 in the Supplement). During the warm campaign, traffic-related components dominated at the city site, contributing 40 % during the day and 49 % at night, though regional contributions were also substantial (29 % day; 36 % night). In contrast, at the elevated site, traffic contributions were lower during the day (17 %) and nearly absent at night (4 %), with the Regional Secondary component becoming dominant (35 % day; 67 % night). The biogenic POA component was also relevant, particularly at the elevated site (32 % day; 13 % night) compared to the city site (18 % day; 10 % night). On the other hand, in the cold period, biomass burning become the dominant component in all cases, especially at night. At the city background site, it accounted for 56 % of the scores during the day and 72 % at night, those percentages contributions to the OA are similar than those observed in OA source apportionment studies in Barcelona and Madrid (van Drooge et al., 2018). A similar dominance was observed at the elevated site, where it accounted for 57 % of the scores during the day and 62 % at night, highlighting the influence of biomass burning in urban background areas with an implication of local sources. The Fresh secondary component also had relevant contributions to the sum of scores during the cold campaign, particularly at the elevated site (30 % day; 22 % night) although it was also important in the city site during the day (19 %).

Significant correlations (p<0.01) between score values in the samples from the urban background site at city level and the elevated site were observed for Biomass burning, Regional secondary and Biogenic components (r2=0.86, 0.89 and 0.56, respectively; Fig. S8 in the Supplement) with similar score values at both sites which suggests vertical and horizontal mixing, or local primary emissions, such is the case of nighttime biomass burning emissions in the elevated site. The Diurnal traffic component was also significantly correlated between sites (r2=0.71), but shows 3 times higher scores at the city site compared to the elevated site (Fig. S8). The Fresh secondary also showed good correlation between sites (r2=0.42) but with higher scores at the elevated site, especially when the scores were low in the city site (Fig. S8). This behaviour indicates a common influence between the two sites for the distribution of these pollutants but pointing to specific sources and processes for their origin, e.g., Diurnal traffic in the city and Fresh secondary related to nucleation at the elevated site.

On the contrary, the Nocturnal traffic component does not show a significant correlation between sampling sites with substantial higher scores at the city site. This decoupled behaviour between altitudes suggests the accumulation of this traffic related component in the nocturnal urban atmosphere in absence of turbulent mixing in nighttime period. The score values of the biomass burning in the urban area at city level in the cold period were anti-correlated with the planetary boundary layer height (PBLH; r2=0.63; p=0.003; Fig. 5). This anti-correlation was not observed at the elevated site, and suggest local emissions in the elevated site, and input of biomass burning emissions and accumulation in the urban airshed for the city site.

The Fresh secondary component was significantly anti-correlated with relative humidity in all cases (Fig. 5). These correlations evidenced the importance of low RH in the formation through nucleation and conservation of fresh SOA (Brines et al., 2019; van Drooge et al., 2022). The chemical compounds involved in the formation of this fresh SOA likely result from reaction with O3, which is also dominantly present in this component. On the other side, the formation of the secondary compounds present in the diurnal and nocturnal traffic components could be more related with the oxidation with the NO2 as evidenced by the presence of nitro-PAHs in these components.

Contrarily to Fresh secondary, the Regional secondary component was significantly correlated with relative humidity in the city site in the warm period and elevated site in both periods (Fig. 5). The formation of the chemical compounds contributing to this SOA component involved aqueous-phase oxidation under high relative humidity. The formation of SOA compounds from biogenic VOCs is influenced by multiple factors, including atmospheric conditions, aerosol acidity, and the presence of oxidants such as NO2 (Surratt et al., 2010). Unlike Fresh and Regional secondary components, individual SOA compounds from isoprene oxidation do not exhibit significant correlations with RH (Fig. S9 in the Supplement) which is characteristic of non-acidic conditions (Nestorowicz et al., 2018) and aligning with the little influence of RH in isoprene SOA yields (Dommen et al., 2006). However, their relative composition (2MGA/∑2MT) shows an anticorrelation with ambient humidity with independence from site and sampling periods, suggesting that different isoprene SOA formation pathways dominate under varying RH conditions. Similarly, the correlation of isoprene SOA compounds with NO2 is significant only for samples from the city site during the spring campaign, where concentrations decrease with increasing NO2 (Fig. S10 in the Supplement). Notably, the 2MGA/∑2MT ratio exhibits opposite trends with NO2 in the warm and cold campaigns. In the warm campaign, higher 2MGA/∑2MT ratios are observed with increasing NO2, aligning with literature that describes the preferential formation of 2-methylglyceric acid under high NOx conditions (Surratt et al., 2010). In contrast, during the cold campaign, the ratio decreases as NO2 levels rise.

For α-pinene SOA, the observed trends resemble those of the Fresh and Regional components, with increased aged-to-fresh ratios ((MBTCA+3HGA)/CPA) under higher RH conditions (Fig. S9). In contrast, NO2 shows a negative correlation with this ratio only in the city site during the warm campaign (Fig. S10).

These opposite trends of individual SOA compounds with RH and NO2 may suggest a relatively minor contribution from local biogenic SOA formation and a greater influence from regional transport. In fact, notable differences in SOA concentrations and relative abundances are observed under different air mass regimes (Fig. S11 in the Supplement), which also reflect variations in RH. Isoprene SOA concentrations are higher in samples influenced by Mediterranean (Regional) and Continental air masses (dark blue and light blue clusters, Fig. S3b).