Environmental Science and Pollution Research

, Volume 21, Issue 20, pp 11649–11660

On the origin of water-soluble organic tracer compounds in fine aerosols in two cities: the case of Los Angeles and Barcelona

Authors

  • M. Alier
    • Department of Environmental ChemistryInstitute for Environmental Assessment and Water (IDAEA-CSIC)
  • M. Dall Osto
    • Institut de Ciències del Mar (ICM-CSIC)
  • Y.-H. Lin
    • Department of Environmental Sciences and Engineering, Gillings School of Global Public HealthUniversity of North Carolina
  • J. D. Surratt
    • Department of Environmental Sciences and Engineering, Gillings School of Global Public HealthUniversity of North Carolina
  • R. Tauler
    • Department of Environmental ChemistryInstitute for Environmental Assessment and Water (IDAEA-CSIC)
  • J. O. Grimalt
    • Department of Environmental ChemistryInstitute for Environmental Assessment and Water (IDAEA-CSIC)
    • Department of Environmental ChemistryInstitute for Environmental Assessment and Water (IDAEA-CSIC)
14th EuCheMS International Conference on Chemistry and the Environment (ICCE 2013, Barcelona, June 25 - 28, 2013)

DOI: 10.1007/s11356-013-2460-9

Cite this article as:
Alier, M., Osto, M.D., Lin, Y. et al. Environ Sci Pollut Res (2014) 21: 11649. doi:10.1007/s11356-013-2460-9

Abstract

Water-soluble organic compounds (WSOCs), represented by anhydro-saccharides, dicarboxylic acids, and polyols, were analyzed by gas chromatography interfaced to mass spectrometry in extracts from 103 PM1 and 22 PM2.5 filter samples collected in an urban background and road site in Barcelona (Spain) and an urban background site in Los Angeles (USA), respectively, during 1-month intensive sampling campaigns in 2010. Both locations have similar Mediterranean climates, with relatively high solar radiation and frequent anti-cyclonic conditions, and are influenced by a complex mixture of emission sources. Multivariate curve resolution-alternating least squares analyses were applied on the database in order to resolve differences and similarities in WSOC compositions in the studied sites. Five consistent clusters for the analyzed compounds were obtained, representing primary regional biomass burning organic carbon, three secondary organic components (aged SOC, isoprene SOC, and α-pinene SOC), and a less clear component, called urban oxygenated organic carbon. This last component is probably influenced by in situ urban activities, such as food cooking and traffic emissions and oxidation processes.

Keywords

PMWSOCSecondary organic aerosolDicarboxylic acidsAnhydro-saccharides

Introduction

Organic carbon (OC) accounts for a major fraction of the atmospheric fine particulate matter (PM2.5) and water-soluble organic compounds (WSOC) are important contributors to OC (Hersey et al. 2011). Biomass burning is an important primary emission source for WSOC (Fine et al. 2004). Nevertheless, in the absence of biomass burning, most of the WSOC is thought to be derived from secondary organics of which many are oxygenated compounds, such as dicarboxylic acids (DCA) (Hallquist et al. 2009). Although DCA are emitted in small quantities from traffic and vegetation it is expected that the majority of these oxidized compounds are formed in the atmosphere after photochemical transformation of volatile and semi-volatile organic compounds from nonfossil (e.g., vegetation) as well as fossil origins (e.g., fossil fuel combustion) (Heald et al. 2010; Kleindienst et al. 2012; Paulot et al. 2011). The importance of oxidized organics is emphasized by their large contribution (40–90 %) to the total the organic fraction in PM2.5, especially in polluted areas (Jimenez et al. 2009).

The composition and evolution of secondary organic carbon (SOC) in the atmosphere is still not well known and prone to controversy (Hallquist et al. 2009). For example, several studies showed that diluted emissions from diesel emissions or biomass burning produce large quantities of SOC, whereas field measurements in the Los Angeles (LA) basin indicated an enhanced formation of SOC from gasoline emissions over diesel emissions (Bahreini et al. 2012). Although the net emissions of volatile organic compounds (VOCs) from vehicles have been reduced by a factor of two in LA over the past decades (despite that the number of vehicles has increased), it is still thought to be a dominant source (Warneke et al. 2012).

Nevertheless, the contributions of oxygenated OC from non-fossil sources, such as biogenic VOCs emitted from vegetation, are still uncertain. Recent results from radioactive carbon (14C) analysis of urban aerosols collected from both Barcelona (BCN) and LA showed that non-fossil sources contribute for about 50 % of the fine OC aerosol (Minguillón et al. 2011; Bahreini et al. 2012). Moreover, food-cooking activities seem to contribute substantially to OC in urban areas. In both LA and BCN, these contributions were estimated to be around 15 % of the OC based on online high-resolution aerosol mass spectrometry (HR-AMS) analysis (Hayes et al. 2013; Mohr et al. 2012).

The comparison between LA and BCN is of interest as both urban areas have similar Mediterranean climates, which are characterized by high solar radiation and frequent anti-cyclonic atmospheric conditions, resulting in accumulation of primary and secondary aerosols. Moreover, both urban areas are under influence of intensive primary urban emissions from vehicles. The vehicle composition in the LA basin is dominated by gasoline-based engines (Bahreini et al. 2012; Hayes et al. 2013), whereas around 75 % of the vehicles in BCN have diesel engines and 30 % of the total vehicles are motorcycles (Reche et al. 2012). Previous results from the same sampling campaigns as the ones in the present study have shown that the urban primary emissions contributed for 12 % to the organic aerosol in LA (Hayes et al. 2013), which was similar to the 18 % estimated in the urban background site in BCN, while this contribution in the traffic intensive road site was 43 %.

As the number of chemical species is very large in fine aerosol, it is almost impossible to obtain a comprehensive chemical characterization of organic aerosol. As a result, the analysis of tracer organic compounds can provide insight into the origin and evolution of the organic aerosol if these compounds are related to source emissions and photooxidation processes. The tracer compounds measured in this study are:
  • Levoglucosan, galactosan, and mannosan are monosaccharide anhydrides that are thermal alteration products of cellulose and hemi-cellulose, respectively (Simoneit 2002). Levoglucosan is emitted in large quantities during biomass burning (Fine et al. 2004), therefore, easily identified in PM samples.

  • DCA, hydroxy-DCA and aromatic-DCA have received much attention because of their role in affecting the global climate and their value as organic tracers for secondary organic aerosol formation (Heald et al. 2010; Moise and Rudich 2002; Paulot et al. 2011; Yang et al. 2008).

  • Cis-pinonic acid, 3-hydroglutaric acid and 3-methyl-1,2,3-butanetricarboxylic acid (MBTCA) have been identified in fine aerosols and related to the photochemical oxidation of α-pinene (Claeys et al. 2007; Szmigielski et al. 2007), which is the most abundant monoterpene in the studied area derived from biogenic emissions. MBTCA is formed by OH-initiated oxidation of cis-pinonic acid (Szmigielski et al. 2007).

  • 2-methylglyceric acid and polyols, such as C5-alkene triols and 2-methyltetrols, have been related to isoprene oxidation (Claeys et al. 2004; Hallquist et al. 2009), a major VOC emitted from terrestrial vegetation.

The selection of these compounds allows analyzing the influence of the different potential sources (anthropogenic vs. biogenic; fossil vs. nonfossil) and oxidation processes on organic aerosol formation. Besides the chemical analysis of these WSOC constituents in the filter extracts from LA and BCN, the relationships among these chemical organic species was further investigated using multivariate curve resolution-alternating least squares (MCR-ALS) analyses on data matrices. MCR-ALS is based on a bilinear decomposition as it is described elsewhere in more detail (Tauler et al. 1995; Jaumot et al. 2005). One of the main advantages of MCR-ALS is that decomposes the data matrix by applying natural constraints, such as non-negativity, and thus, results can be interpreted more straightforwardly. MCR-ALS has been successfully used in the analysis of environmental data sets in air source apportionment studies (Tauler et al. 2009; Alier et al. 2013).

Methods and materials

Sampling sites in BCN

Two sites were sampled in BCN from 22 September 2010 to 18 October 2010 as part of the Solving Aerosol Problems by Using Synergistic Strategies project (SAPUSS). The urban background site (BCN-UB; 41.3899° N; 2.1161° E; 80 m above sea level (a.s.l.)) was located in a residential area, whereas the road site (BCN-RS; 41.3884° N; 2.1500° E; 40 m a.s.l.) was located in Urgell Street within the square-grid street network of BCN’s city center. More detailed information is described elsewhere (Dall’Osto et al. 2013).

Sampling site in LA

The sampling site in the LA Basin was located in Pasadena, CA (LA; 34.1408° N; 118.1223° W; 230 m a.s.l.), where continuous gas- and aerosol-phase sampling occurred from 15 May to 16 June 2010 as part of the CalNex (California Research at the Nexus of Air Quality and Climate Change) field study (Ryerson et al. 2013). This site is 18 km NE from downtown LA and 44 km from the Long Beach harbor area. Based on the characteristic of this site it is comparable to the background site in BCN.

Sampling details

In BCN, all filter samples were collected on pre-baked (450 °C overnight) quartz fiber filters (Tissuquartz™ Filters, 2500 QAT-UP, Pall Life Sciences), allowing for high-volume PM1 sampling (Digitel-DH80; 30 m3/h). Samples were collected in 12-h sample interval from 9:00 to 21:00 and 21:00 to 9:00 local time. In LA, filter samples were collected on quartz fiber filters (Tissuquartz™ Filters, 2500 QAT-UP, Pall Life Sciences), allowing for high-volume PM2.5 (TE-6001; 60 m3/h) of 23-h integrated filter samples from midnight to 11 pm local time.

Filter samples collected in both locations from each day were stored in a freezer at −18 °C until chemical analysis. Field blanks were collected every 7–10 days by placing a pre-fired quartz fiber filter into the sampler for 15 min before removing and storing in the same manner as the field samples. Analysis of both field and lab blanks showed no significant organic contaminants on pre-fired quartz filters, indicating that all of the organic compounds characterized in the present study are due to the aerosol collected from LA and BCN.

Organic tracer analysis

A detailed description of the analytical procedure can be found elsewhere (Alier et al. 2013). Briefly, a fraction (1/4 or 1/8 parts) of the blank and sample filter was ultrasonically extracted in a mixture of (2:1, v/v) dichloromethane and methanol (3 × 15 ml, Merck, Germany). After extraction, the extracts were filtered, concentrated by rotovap and evaporated under a gentle stream of N2 until dryness. WSOCs were derivatized to their methylsilylate esters by adding bis(trimethylsilyl)trifluoroacetamide + trimethylchlorosilane (99:1; Supelco, USA) and pyridine (Merck, Germany). A Thermo gas chromatography interfaced to mass spectrometry (GC/MS; Thermo Trace GC Ultra-DSQ II) equipped with a 60-m fused capillary column (HP-5MS 0.25-mm × 25-μm film thickness) was used for analysis operating in full scan (m/z 50–650) and electron ionization (70 eV) modes. Compounds were identified with their characteristic ions at the corresponding retention time in authentic standards or by comparison of mass spectrometric fragmentation patterns from literature and library data in the case of 3-hydroxyglutaric acid, 3-methyl-1,2,3-butanetricarboxylic acid (MBTCA), 2-methylglyceric acid, C5-alkene triols, 2-methylthreitol, and 2-methylerythritol (Claeys et al. 2007). Quantification was performed by calculating the concentrations using external standard calibration curves of authentic standard. Here, the chromatographic area of the ion of an analyte was divided by the area of internal standard, so a calibration curve can be formed with the know amounts of the standards. Then, the area of the analyte in the sample, divided by the area of the internal standard, was introduced to the regression line of the calibration curve in order to obtain the amount of the analyte in the extract (corrected by the known amount of internal standard). To correct for losses during the analytical procedure, the amounts of the analytes are corrected by the recoveries of the surrogate standards (succinic acid-d4 (Sigma Aldrich) and levoglucosan-d7 (Cambridge Isotopic Laboratories)) that were added to the filter samples before extraction.

Data arrangement

Experimental data obtained from GC/MS were arranged in a data matrix for every analyzed station and used for chemometric analysis in the MCR-ALS method. In the two BCN sites, the samples were stored in two separate data matrices in 12-h periods from 22 September to 18 October 2010. In LA site, samples were stored in a single data matrix in 24-h periods from 15 May to 12 June 2010. The samples of three locations were analyzed individually and simultaneously. Experimental data from the different samples (excluding field blanks) are set in the matrix rows and the measured compounds in the matrix columns. This gave a data matrix of 52 rows (samples) and 17 columns (measured compounds) for BCN-UB station, a data matrix of 51 × 17 for BCN-RS station and a data matrix of 22 × 17 for LA station. As the three data matrices (stations) had the same number of columns (analyzed compounds), the three individual data matrices were also arranged in a column-wise augmented data matrix with dimensions 125 × 17.

All data arrangements and pretreatments were performed into MATLAB 7.4 (The Mathworks, Natick, USA) for subsequent multivariate data analysis using MATLAB PLS 5.8 Toolbox (Eigenvector Research Inc, Masson WA, USA).

Results and discussion

WSOC analysis

Table 1 shows the mean, minimum and maximum concentrations of WSOC constituents (or tracers) in the two locations after analyses of the extracts by GC/MS.
Table 1

Mean (minimum–maximum) concentrations (nanograms per cubic meter) of analyzed WSOC in PM samples from LA and BCN (number of samples (N))

 

LA (N = 22)

BCN-UB (N = 52)

BCN-RS (N = 51)

Galactosan

1.4 (0.6–2.5)

1.3 (0.2–7.1)

0.7 (0.1–4.4)

Mannosan

1.5 (0.7–3.1)

1.2 (0.2–5.6)

0.6 (0.1–3.8)

Levoglucosan

14.1 (7.1–24.0)

9.4 (1.7–40.0)

5.3 (0.6–30.5)

Succinic acid

14.4 (4.2–38.4)

7.3 (3.1–24.7)

6.2 (2.5–14.8)

Glutaric acid

5.9 (1.8–13.9)

2.0 (0.6–7.5)

1.5 (0.6–3.2)

Pimelic acid

1.0 (0.5–1.8)

0.6 (0.2–2.1)

1.3 (0.6–3.7)

Suberic acid

0.9 (0.3–1.8)

0.9 (0.4–3.0)

1.7 (0.7–3.9)

Azelaic acid

5.1 (2.5–12.1)

2.9 (1.0–7.3)

6.1 (2.4–13.0)

Malic acid

19.4 (4.0–47.1)

14.2 (1.2–72.7)

9.9 (1.2–33.6)

Phthalic acid

5.2 (0.7–13.2)

3.9 (0.9–11.4)

3.6 (1.3–9.3)

Cis-pinonic acid

4.8 (2.2–11.3)

15.4 (3.6–49.7)

8.0 (3.1–18.4)

3-hydroxyglutaric acid

6.2 (1.1–14.1)

4.5 (0.5–20.2)

3.1 (0.8–8.6)

MBTCA

1.2 (0.3–2.7)

5.5 (1.0–23.5)

4.1 (0.7–10.3)

C5-alkene triols

5.7 (1.5–12.6)

1.0 (0.1–3.0)

0.7 (0.1–2.2)

2-methylglyceric acid

6.2 (1.1–14.1)

2.1 (0.4–6.4)

1.9 (0.6–4.5)

2-methylthreitol

3.7 (1.4–7.9)

2.8 (0.6–7.8)

1.5 (0.4–3.8)

2-methylerythritol

9.3 (3.1–22.4)

6.5 (1.7–20.7)

3.5 (0.6–9.5)

Anhydro-saccharides

Levoglucosan, mannosan and galactosan, as organic tracers for biomass burning, showed low concentrations, with mean levels for levoglucosan of 14, 9, and 5 ng/m3 in LA, BCN-UB, and BCN-RS, respectively (Table 1). The levoglucosan levels were comparable to those found in background sites where biomass burning has little influence on the air quality (Puxbaum et al. 2007). Ratios between the compounds have been used in the past to indicate the biomass source of the combustion (Fine et al. 2004, Schmidt et al. 2008). The ratios of levoglucosan to mannosan were very consistent during the samplings and similar in all sites (9.8 ± 1.7 in LA, 9.6 ± 3.4 in BCN-UB, and 9.2 ± 2.1 in BCN-RS). Based on combustion experiments using different wood types, where softwood and hardwood ratios are around 4 ± 1 and 18 ± 5, respectively (Fine et al. 2004; Schmidt et al. 2008), the observed ratios in the present study indicate a mixture of soft- and hardwood.

Both in LA and BCN, wood combustion is not a local energy source and the observed concentrations are representative for influence of regional biomass combustion. This is further indicated by the good correlation between the temporal concentration trends in the two sites in BCN (r2 = 0.8). Here, higher concentrations were observed in the last part of the sampling campaign (Fig. 1), which coincidence with the period when open fires in fields for biomass waste burning are legally allowed in the region. In LA, the concentrations showed fluctuations between 10 and 20 ng/m3, but there was a decreasing trend along the sampling period (Fig. 1; r2 = 0.3). The small influence of biomass burning in LA was also observed in the HR-AMS analysis during the CalNex campaign (Hayes et al. 2013). In the HR-AMS analyses, the biomass burning signal was below detection limit and probably mixed with secondary organic species, indicating the origin of the biomass burning aerosol was the result of long-range transport.
https://static-content.springer.com/image/art%3A10.1007%2Fs11356-013-2460-9/MediaObjects/11356_2013_2460_Fig1_HTML.gif
Fig. 1

Temporal trends of concentrations (nanograms per cubic meter) of the ∑anhydro-saccharides (green), ∑dicarboxylic acids (red), ∑SOC tracers for α-pinene (black), and ∑SOC tracers for isoprene (blue) in the three sampling sites

The 12-h resolution samples in BCN did not show any clear difference between day and nighttime samples for the concentrations of anhydro-saccharides, although during the end of the sampling period, slightly higher levels were observed during the nighttime, which probably indicates the influx of biomass burning influenced air masses from the inland by the land-breeze during this period.

Despite the detection of the anhydro-saccharides in the PM samples in both locations, as well as the indications that these compounds are sampled in aerosols after long-rang atmospheric transport, it is important to remember that these organic tracers for biomass burning may to be susceptible to oxidation (Hennigan et al. 2011; Hoffmann et al. 2010).

DCA, hydroxy-DCA, and aromatic-DCA

The mean concentration of these compounds ranged between 1 and 19 ng/m3 in LA, 1–14 ng/m3 in BCN-UB, and 1–10 ng/m3 in BCN-RS (Table 1), with the highest individual compound concentrations observed for malic acid (73 ng/m3 in BCN and 47 ng/m3 in LA) and succinic acid (24 ng/m3 in BCN and 38 ng/m3 in LA). The higher levels of malic acid (hydroxy-succinic acid) than its parent compound, succinic acid, was also observed in other sites that were influenced by anthropogenic emission sources although a non-fossil origin of these compounds cannot be excluded (Yang et al. 2008, and references therein). Malic acid concentrations were similar to those measured in winter in the urban area of BCN (van Drooge et al. 2012) and comparable to other European sites, such as K-Puszta, Hungary (38 ng/m3) (Ion et al. 2005) and Jülich, Germany (39 ng/m3) (Kourtchev et al. 2008).

Phthalic acid, a possible oxidation product of gas-phase polycyclic aromatic hydrocarbons from fossil fuel combustion, had similar concentrations in the three sites (∼4 ng/m3; Table 1). Phthalic acid concentrations in LA were consistent with those obtained from GC-ITMS analyses by Kleindienst et al. (2012) in collocated filter samples collected from the CalNex study. Moreover, they identified 2-methylphthalic acid in their sample extracts at concentrations around 3 times lower than phthalic acid, but with a similar temporal trend. Posterior analysis of the GC/MS chromatograms in our study confirms the presence of 2-methylphthalic acid in similar concentrations (LA = 1.5 ng/m3; BCN-UB and BCN-RS = 0.9 ng/m3) as Kleindienst et al. (2012) and showing good correlations with phthalic acid (r2 = 0.8 in LA; 0.7 in BCN-UB and 0.4 in BCN-RS). In the 12 h sampling resolution in BCN, the daytime samples also showed higher concentrations than the nighttime samples, indicating the formation of these products in the city. Kleindienst et al. (2012) discussed the application of phthalic acid (and isomers) for the source apportionment of SOC contribution from PAH emissions (i.e., naphthalene oxidation). The temporal multiday trend in LA as well as the similar concentrations in two sites in BCN (r2 = 0.7; slope = 1.1) agrees with their statement that primary emissions of phthalic acid are probably very small in comparison to secondary formation. Nevertheless, the weaker regression coefficients in BCN compared with the one in LA as well as the smaller slopes in BCN between 2-methylphthalic acid vs. phthalic acid (0.29 in LA and 0.18 in BCN), may indicate that the apportion of primary phthalic acid is higher in BCN than in LA. Conversely, in the two locations phthalic acid did not show any difference between weekend and weekdays and highest multiday concentrations were observed during regional circulation of air masses (e.g., between 2 and 6 June in LA and between 15 and 17 October in BCN (Fig. 1) (Dall’Osto et al. 2013; Hayes et al. 2013; Thompson et al. 2012)).

These results indicate that most of the detected phthalic acid (and other DCAs, such as succinic acid, glutaric acid and malic acid) were SOC, which were relatively rapidly formed in the atmosphere after oxidation of precursors. Under stagnant atmospheric condition these compound accumulate in the urban atmosphere. This could lead to relatively higher day than nighttime levels, but not necessarily to higher weekday concentrations.

Longer-chained DCAs (C7–C9), i.e., pimelic acid, suberic acid, and azealic acid, seemed to behave differently from the shorter DCAs and their isomers (including aromatic DCAs). For example, the C7–C9 concentrations were two times higher in the urban center of BCN-RS compared with the background site BCN-UB. In all sites, azelaic acid concentrations were always higher than pimelic acid and suberic acid (Table 1), with good correlations between the temporal trends (r2∼0.7). In BCN-RS the concentrations of these compounds showed a clear daytime maximum and nighttime minimum, while this was less clear in BCN-UB. The presence of azelaic acid in ambient PM2.5 has been related to fast oxidation of unsaturated fatty acids, i.e., oleic acid, by ozone (Moise and Rudich 2002). Oleic acid itself may have many sources in the urban atmosphere, such as food cooking, traffic and non-fossil sources, including the marine environment (Schauer et al. 2002). In the present study, this possible precursor for azelaic acid was detected only in low concentrations (<2 ng/m3; including in field blanks). It is very well possible that the C7–C9 DCAs are related to rapid oxidation processes within the urban atmosphere that are not yet well characterized. An indication for this hypothesis is the high concentrations in the city center (BCN-RS, 6.1 ng/m3) in comparison to the background site in BCN (BCN-UB, 2.9 ng/m3). Nevertheless, azelaic acid itself is also susceptible to relatively fast oxidation in the presence of OH radicals to form shorter chained DCAs. As discussed in detail by Alier et al. (2013), in BCN there was a linkage of these compounds and primary organics typically emitted by gasoline engine vehicles as well as air mass dependency, with higher concentrations under low-nitrate conditions.

Biogenic SOC tracers for α-pinene

The biogenic SOC tracers for α-pinene oxidation, which includes cis-pinonic acid, 3-hydroxyglutaric acid (3-HGA) and MBTCA, had average concentrations of 5, 6, and 1 ng/m3 in LA, 15, 5, and 6 ng/m3 in BCN-UB, and 8, 3, and 4 ng/m3 in BCN-RS, respectively (Table 1). The LA levels were in the lower range as those observed in July–August in Riverside, situated east within the LA Basin (∑α-pinene SOA∼38–60 ng/m3) (Stone et al. 2009), which may indicate seasonal dependence of the concentrations of these compounds, since the present study was conducted in May–June. On the other hand the present concentrations were very similar to the ones measured during June–July in urban sites in Marseille, along the Mediterranean coast in France (15, 4, and 5 ng/m3, for cis-pinonic acid, 3-HGA, and MBTCA, respectively; El Haddad et al. 2011). The higher concentrations in BCN-UB could be caused by the proximity of this site the coniferous forest, whereas the lower cis-pinonic concentrations in BCN-RS may reflect further oxidation of this compound in the urban atmosphere. Studies performed in forested areas showed slightly higher levels of MBTCA, such as K-Puszta, Hungary (12 ng/m3) (Kourtchev et al. 2009), and Jülich, Germany (7 ng/m3) (Kourtchev et al. 2008). The α-pinene SOC tracers showed different diurnal and multiday trends. While 3-HGA and MBTCA were highly correlated in the two locations (r2 = 0.8 in LA and 0.7 in BCN), cis-pinonic acid did not show any correlation with 3-HGA and MBTCA. Cis-pinonic acid is a first-generation product of α-pinene oxidation, whereas MBTCA is formed by OH-initiated oxidation of cis-pinonic acid (Szmigielski et al. 2007) and 3-HGA is also thought to be a further-generation oxidation product (Claeys et al. 2007). Therefore, this possibly explains why there was no correlation observed between cis-pinonic acid and the other two oxidation products. Despite the substantial correlations between 3-HGA and MBTCA in both sites, the ratio between the compounds in LA was 5.1 ± 1.7 and 0.8 ± 0.4 in BCN. It is unclear what mechanisms are behind these differences, but it is possible the time of sampling (May–June for LA; September–October for BCN) influences these ratios.

SOC tracers for isoprene

The SOC tracers for isoprene, which includes C5-alkene triols, 2-methylglyceric acid (2-MGA), 2-methylthreitol and 2-methylerythritol (2-MT), had average concentrations of 6, 6, 4, and 9 ng/m3 in LA, 1, 2, 3, and 7 ng/m3 in BCN-UB and 1, 2, 1, and 4 ng/m3 in BCN-RS, respectively (Table 1). Similar concentrations were found for Lewandowski et al. (2013) in a study done also at the CalNex campaign for 2-methylglyceric acid (3 ng/m3), 2-methylthreitol (1 ng/m3), and 2-methylerythritol (3 ng/m3). C5-alkene triols are known isoprene SOC tracers and have been reported at levels around 3 ng/m3 in other European sites (Kourtchev et al. 2008) and at very high levels around 500 ng/m3 in sites located in the southeastern USA (Lin et al. 2013a). Similarly high concentrations of 2-MT (570 ng/m3) were observed in southeastern U.S. during late summer (Lin et al. 2013a). 2-MT and C5-alkene triols are formed by the photooxidation (i.e., OH-initiated oxidation) of isoprene under NO-limited conditions. 2-MGA is formed by the further oxidation of volatile methacryloylperoxynitrate (MPAN), which is a major second-generation product of isoprene oxidation under initially high-NO conditions (Surratt et al. 2010). Recent work by Lin et al. (2013b) found that 2-MGA is directly formed from methacrylic acid epoxide, which is a gas-phase oxidation product of MPAN. The 2-MGA concentrations were in the same range as those observed in Marseille, France (El Haddad et al. 2011), while the levels of 2-MT were higher in the present study. In the present study, both locations had 2-methylerythritol concentrations that were ∼2.5 times higher than 2-methylthreitol concentrations, but the two compounds had the same temporal trend (r2 = 1). This ratio between these isomers of the 2-methyltetrols has also been observed in other studies (El Haddad et al. 2011; Ion et al. 2005).

Source apportionment of WSOC

MCR-ALS was applied to the individual and column-wise augmented data matrices (see “Data arrangement”) in order to investigate the similarities and differences of the analyzed WSOCs in both LA and BCN. In all cases, non-negativity and loadings normalization constraints were applied. Loading normalization allow the comparison of the compounds inside one component and also between different components. The total explained variances for the individual analyses were 98.7, 97.7, and 97.8 % for LA, BCN-UB, and BCN-RS sites, respectively, and 95.9 % for the augmented data matrix, when the three sites were analyzed simultaneously. This latter analysis emphasizes the composition of the analyzed compounds that the three stations had in common. The explained variance of the obtained components was normalized to 100 % in order to compare the contributions (score values) of the components to the WSOC in the three sites.

Five consistent components (Fig. 2) were identified in the three sites for the individual and augmented database analysis. The five resolved components could be described as potential WSOC sources: one from regional biomass burning and the other four from oxidation processes (SOC). A primary urban organic carbon (POC) component, related to traffic, was not obtained in this study. However, previous studies from the same sampling campaigns showed that 12 % of the organic matter in LA was attributed to primary emissions that were related to traffic, whereas this was 18 % in the urban background site (BCN-UB) and 43 % in the traffic intensive road site (BCN-RS) (Alier et al. 2013; Hayes et al. 2013). These percentages were applied here in order to estimate the contribution of the obtained WSOC components to the mean OC concentration. The mean OC concentration in LA was 3.3 μg/m3, while this was 2.1 μg/m3 in BCN-UB and 3.5 μg/m3 in BCN-RS. The OC concentrations versus the MCR-ALS sum of scores were correlated in both locations (r2 = 0.82 in LA and 0.24 in BCN (p < 0.01)). The lower correlation in BCN was due to the relatively high POC to OC contributions. Hence, the correlation coefficient improved when the POC was included (r2 = 0.75). The correlation coefficient in LA was high and remained high if the POC contributions were added. These findings suggest that the obtained WSOC components in this study together with the contributions of POC from previous studies covered most of the OC in the two locations. In this way it was estimated that about 88 % of the OC was attributed to WSOC (2.8 μg OC/m3), while this was 82 % in BCN-UB (1.7 μg OC/m3) and 57 % in BCN-RS (2.0 μg OC/m3). In the following paragraphs, the five resolved WSOC components are presented and their contributions to OC are discussed.

Regional biomass burning

The first component (Fig. 2a) was related to biomass burning and represented by levoglucosan, mannosan, and galactosan in all sites. The influence of regional biomass burning on the WSOC was 17 % in LA, with a decreasing trend towards the end of the sampling period (Fig. 3). In BCN (9–16 % of WSOC), the score values increase only in the last part of the sampling period, related to the period when open fires in fields for biomass waste burning are legally allowed in the region (Fig. 3). The loadings of DCA in this component were similarly low in all sites (Fig. 2), indicating a clear separation of biomass burning contributions and those from other sources and/or processes.
https://static-content.springer.com/image/art%3A10.1007%2Fs11356-013-2460-9/MediaObjects/11356_2013_2460_Fig2_HTML.gif
Fig. 2

MCR-ALS loadings of components for individual analysis in BCN-RS, BCN-UB, LA, and simultaneous analysis (RS + UB + LA) with the relative contribution to the overall variation in the legends. a Regional biomass burning, b aged secondary organic carbon, c oxygenated organic carbon urban, d isoprene secondary organic carbon, and e biogenic α-pinene secondary organic carbon

https://static-content.springer.com/image/art%3A10.1007%2Fs11356-013-2460-9/MediaObjects/11356_2013_2460_Fig3_HTML.gif
Fig. 3

Simultaneous MCR-ALS score resolved profiles for the LA (left) and BCN sites (right) using non-negativity constraints

These estimates showed that in LA about 0.5 μg/m3 of OC corresponded to “biomass burning organic carbon” (BBOC; 15 % of OC), while this was 0.3 μg/m3 in BCN-UB (13 % of OC) and 0.2 μg/m3 in BCN-RS (5 % of OC) (Fig. 4). The BBOC contribution in LA is higher than the <5 % estimated in aerosol mass spectrometry (AMS) analysis by Hayes et al. (2013). The low contributions in these analysis was possibly caused by a generally low biomass burning signal (i.e. low levoglucosan concentrations) in combination with the high abundance of secondary organics in the LA atmosphere.

Aged secondary organic carbon

The second component (Fig. 2b), which accounted for 28 % of the WSOC in LA and about 15 % in BCN, was composed mainly of oxygenated organic aerosol tracer compounds, such as short-chained diacids (e.g., succinic acid and glutaric acid), phthalic acid, and more oxygenated compounds from biogenic origin, such as malic acid, 3-HGA, and MTBCA.

Phthalic acid, possibly formed in several steps by oxidation of napthalene (Kautzman et al. 2010), can be related to processed aerosols. Interestingly, in both locations, phthalic acid showed good correlations with MBTCA and 3-HGA (r2 > 0.5); two products of further α-pinene oxidation. In all sites, there was a 1:1 relationship between 3-HGA and phthalic acid. Conversely, MBTCA showed five times lower concentrations compared with phthalic acid in LA, while in BCN the concentrations of MBTCA were slightly higher than phthalic acid concentrations. Succinic, glutaric, and malic acid also showed similar relationships between 3-HGA and MBTCA in the two locations. As mentioned before, it could be possible that the sampling performance in different seasons (late spring in LA and early fall in BCN) may have an influence on the formation pathways of α-pinene SOC formation.

Generally, the compounds represented in this component were products of further oxidation processes and can therefore be linked to processed air from urban and biogenic emissions in relation with circulation of air masses within the studied urban areas. In LA and BCN, the temporal trend of the Aged SOC component (Fig. 3) was very similar to the trend of the low volatile oxygenated organic aerosols (LV-OOA) measured by AMS (Hayes et al. 2013), indicating the high oxidation state of these aerosols. Moreover, at both sites, the highest abundance of this component (Fig. 3) coincides with generally high aerosol loadings during the sampling campaigns, between 3–5 Oct and 14–16 Oct in BCN, and 4–7 June in LA (Dall’Osto et al. 2013; Hayes et al. 2013; Thompson et al. 2012). This component contributed to about 30 % of the OC (0.8 μg OC/m3), whereas this was about 8–13 % (0.3 μg OC/m3) in BCN (RS and UB, respectively) (Fig. 4).

Oxidized organic carbon of mainly urban origin

The third component (Fig. 2c), which accounted for 16 % of the WSOC in LA, 19 % in BCN-UB and 50 % in BCN-RS, was mainly composed of longer-chained DCA (C7–C9). There was a clear separation between these longer-chained DCAs and shorter ones (succinic acid and glutaric acid) in the augmented data base, while this separation was less clear for the phthalic acid, malic acid, 3-HGA, and MBTCA in the case of LA when the database was analyzed separately (Fig. 2c). The presence of these oxidized compounds in this component in LA may be related to a more equilibrated influence of potential sources/processes to this component in this site compared with the sites in BCN. In the BCN-RS there was a strong day–night fluctuation of the component (Fig. 3c) and the contribution to OC (1.0 μg OC/m3) was significantly higher than the one in BCN-UB (0.3 μg OC/m3) (Fig. 4), indicating that this component could be related to specific in-site urban oxidation processes, which may include cooking activities (see “DCA, hydroxy-DCA, and aromatic-DCA”). In the urban background site of BCN-UB this day–night fluctuation is less pronounced (Fig. 3c), whereas in LA the major compounds (C7–C9 DCAs) mix up with the “Aged SOC” compounds (Fig. 2b, 3b). Therefore, it is seems that the “oxidized organic carbon urban” component was formed after rapid oxidation in the urban atmosphere and then accumulates in the background areas. However, the exact precursor and formation pathway remains unclear. In a previous study in BCN (Alier et al. 2013), the abundance of this component was related to primary urban emissions (i.e. higher molecular weight PAH), but there was also an air mass dependence.

In LA, the temporal trend of this component showed similarities with the cooking influenced organic aerosol, although it also correlated with the LV-OOA component (Hayes et al. 2013). Therefore, further study is necessary to elucidate the origin of this component.

Isoprene SOC

The fourth component (Fig. 2d) accounted for about 36 % of the WSOC in LA and 6–11 % in BCN (RS and UB, respectively), and was composed of the products of isoprene oxidation; specifically, the C5-alkene triols, 2-methyltetrols, and 2-methylglyceric acid.

Overall, there were substantial correlations between C5-alkene triols and the 2-MT (r2 = 0.8 in BCN and 0.7 in LA). A similar good correlation was observed for 2-MGA in BCN (r2 = 0.6 and 0.7, respectively), however weaker in LA (r2 = 0.3 and 0.5, respectively). Although the results in this study suggest that the isoprene oxidation products have similar origins, the weaker correlations with 2-MGA points to different transformation pathway, which were discussed in detail in “SOC tracers for isoprene”: 2-MT is formed under NO-limited conditions. Despite the slightly lower NO concentration in LA (4 μg/m3) compared with the ones in BCN (5 and 8 μg/m3 in UB and RS, respectively), there was no correlation between 2-MT and NO (Dall’Osto et al. 2013; Pollack et al. 2012). Overall, the temporal trends of this component (Fig. 2d) does not coincident with the highest bulk load for PM and organic aerosols (Dall’Osto et al. 2013; Thompson et al. 2012) nor with any of the AMS components obtained by Hayes et al. 2013. This may indicate that these secondary organics from isoprene oxidation probably do not contribute substantially to the urban OC here. However, the substantial contribution of 31 % to OC of this component in LA (1.0 μg OC/m3; Fig.4) suggest the contrary, that at least in this site the influence of isoprene oxidation is important.

Biogenic α-pinene SOC

The fifth component (Fig. 2e) was composed of cis-pinonic acid, and contributed to 4 % of the WSOC in LA, whereas this was 21–37 % in BCN (RS and UB, respectively). The separation of this compound from all the other analyzed compounds indicates that a unique source or process and specific conditions must be involved.
https://static-content.springer.com/image/art%3A10.1007%2Fs11356-013-2460-9/MediaObjects/11356_2013_2460_Fig4_HTML.gif
Fig. 4

Mean contributions of WSOC components (micrograms per cubic meter) to OC for the studied periods in LA, BCN-UB, and BCN-RS. The red line above the columns indicates the mean OC concentration (micrograms per cubic meter)

The locations investigated in the present study did not show correlations between the concentrations of cis-pinonic acid and 3-HGA or MBTCA (r2 < 0.2). The absence of these correlations can be explained by the different formation modes of these compounds as mentioned in “Biogenic SOC tracers for α-pinene”: cis-pinonic acid is a first-generation product of pinene oxidation, while MBTCA and 3-HGA are probably formed from further reaction of cis-pinonic acid in the presence of NOx. However, no correlations were observed between these parameters in any of the sites. Nevertheless, the ratio between cis-pinonic acid and MBTCA showed peak ratios (>10) in BCN on 25 September and 5 October, when urban new particle formation events were detected (Dall’Osto et al. 2013). Cis-pinonic acid has been linked to nucleation processes in forested areas as a first step in the formation of aerosols from organic vapors (O’Dowd et al. 2002; Laaksonen et al. 2008). In LA, the ratio between cis-pinonic acid and MBTCA were highest (>10) between 24 and 29 May, coinciding with a period of generally low aerosol loadings (Hayes et al. 2013).

Implications and conclusions

GC/MS with prior derivatization allows for the identification of several known organic tracer compounds for WSOC in PM filter samples. Moreover, the MCR-ALS was successfully applied on the WSOC dataset from LA and BCN, obtaining five consistent components in these urban atmospheres. The estimated contributions of WSOC to OC were 88 % in LA, 82 % in BCN-UB and 57 % in BCN-RS. In both locations, the biomass burning contributions to WSOC were low but could be separated from components that were related to oxidation processes, from biogenic and anthropogenic origin. These components made up about two thirds of the OC is related to SOC in LA and the urban background site in BCN, while this was half of the OC in BCN road site. Fresh biogenic SOA species, such as cis-pinonic from α-pinene oxidation and 2-methyltetrols from the oxidation of isoprene, showed different temporal concentration trends in comparison to further-generation oxidation products (aged SOC), including those from biogenic emissions as well as vehicle emissions. Isoprene oxidation was more important in LA than in BCN. Although this could be related to the different periods of sampling, the influence of oxidants, such as NOx, can also play a role and should be studied in more detail. The aged products from biogenic and anthropogenic origin clustered in another component that was strongly related to the organic aerosol loadings in the studied areas. Interestingly, the component represented by longer-chained C7–C9 DCA that are probably related to urban activities and in situ oxidation processes, may also involve specific atmospheric conditions. This later component was dominant in the road site in BCN, and further study should elucidate the role of primary sources (cooking and traffic), secondary aerosol processing, and atmospheric conditions on this component.

Acknowledgments

The authors thank Xavier Querol, Manuel Dall’Osto, Joost de Gouw, Jochem Stutz, Jason Surratt, John Seinfeld, and Jose Luis Jimenez, the organizers of the SAPUSS and CalNex-LA campaigns, for their service to the community. Technical assistance from R. Chaler and D. Fanjul is acknowledged. Financial support for this study was provided by projects: CTQ2009-11572, SAPUSS (FP7-PEOPLE-2009-IEF, Project number 254773) and AEROTRANS (CTQ2009-377 14777-C02-01).

Copyright information

© Springer-Verlag Berlin Heidelberg 2014