By continuing to browse this site you agree to us using cookies as described in About Cookies
Notice: Wiley Online Library will be unavailable on Saturday 7th Oct from 03.00 EDT / 08:00 BST / 12:30 IST / 15.00 SGT to 08.00 EDT / 13.00 BST / 17:30 IST / 20.00 SGT and Sunday 8th Oct from 03.00 EDT / 08:00 BST / 12:30 IST / 15.00 SGT to 06.00 EDT / 11.00 BST / 15:30 IST / 18.00 SGT for essential maintenance. Apologies for the inconvenience.
Radiocarbon (14C) analysis is a unique tool to distinguish fossil/nonfossil sources of carbonaceous aerosols. We present 14C measurements of organic carbon (OC) and total carbon (TC) on highly time resolved filters (3–4 h, typically 12 h or longer have been reported) from 7 days collected during California Research at the Nexus of Air Quality and Climate Change (CalNex) 2010 in Pasadena. Average nonfossil contributions of 58% ± 15% and 51% ± 15% were found for OC and TC, respectively. Results indicate that nonfossil carbon is a major constituent of the background aerosol, evidenced by its nearly constant concentration (2–3 μgC m−3). Cooking is estimated to contribute at least 25% to nonfossil OC, underlining the importance of urban nonfossil OC sources. In contrast, fossil OC concentrations have prominent and consistent diurnal profiles, with significant afternoon enhancements (~3 μgC m−3), following the arrival of the western Los Angeles (LA) basin plume with the sea breeze. A corresponding increase in semivolatile oxygenated OC and organic vehicular emission markers and their photochemical reaction products occurs. This suggests that the increasing OC is mostly from fresh anthropogenic secondary OC (SOC) from mainly fossil precursors formed in the western LA basin plume. We note that in several European cities where the diesel passenger car fraction is higher, SOC is 20% less fossil, despite 2–3 times higher elemental carbon concentrations, suggesting that SOC formation from gasoline emissions most likely dominates over diesel in the LA basin. This would have significant implications for our understanding of the on-road vehicle contribution to ambient aerosols and merits further study.
Particulate matter (PM) in ambient air has a negative impact on human health because it can cause respiratory and cardiopulmonary diseases that lead to increased mortality [Pope and Dockery, 2006; World Health Organization, 2006]. Aerosol particles also affect the climate by scattering and absorbing sunlight and modifying clouds [Intergovernmental Panel on Climate Change, 2007; Jacobson et al., 2000]. Carbonaceous particles (total carbon, TC) are a major fraction of the fine aerosol (PM1 or PM2.5, PM < 1 or 2.5 µm) and contribute from 10% up to 90% of the PM mass depending on the location and season [Gelencsér, 2004; Jimenez et al., 2009; Putaud et al., 2004]. TC is further classified into the subfractions elemental carbon (EC) and organic carbon (OC) [Jacobson et al., 2000]. EC is directly emitted from combustion processes of fossil and nonfossil fuels (e.g., coal, gasoline, diesel, oil, and biomass) [Pöschl, 2005]. OC, on the other hand, can be directly emitted as primary organic aerosol (POA) from biogenic sources (e.g., plant debris), biomass burning, cooking, and fossil fuel combustion or can be formed in the atmosphere as secondary organic aerosol (SOA) via gas-to-particle conversion of low-volatility reaction products of gas phase precursors [Hallquist et al., 2009; Jacobson et al., 2000]. As a result, organic aerosol (OA) is comprised of an immensely complex mixture of POA and SOA compounds from a myriad of sources [Jimenez et al., 2009]. The characterization of these sources and the quantification of their emission strengths is a key step in understanding the impact of OA on public health and climate and improving the implementation of effective mitigation strategies.
Several measurement techniques and methods exist for the characterization and quantification of POA and SOA. One among those is the Aerodyne aerosol mass spectrometer (AMS) which measures the nonrefractory organic and inorganic (sulfate, nitrate, chloride, and ammonium) composition of submicron particles [Jimenez et al., 2003; Drewnick et al., 2005; DeCarlo et al., 2006; Canagaratna et al., 2007]. Positive matrix factorization (PMF) of the AMS organic mass spectra [Paatero and Tapper, 1994; Lanz et al., 2007; Ulbrich et al., 2009] can then be used to retrieve different OA sources and formation processes. However, it is not possible to assign an anthropogenic versus biogenic or fossil versus nonfossil contribution for OA sources with PMF, especially for SOA.
Radiocarbon analysis (14C) is a quantitative tool for unambiguously distinguishing fossil and nonfossil sources of EC and OC. 14C is completely depleted in fossil fuel emissions (14C half-life = 5730 years), which can therefore be separated from nonfossil carbon sources (e.g., biomass burning, cooking, or biogenic emissions) which have a similar amount of 14C as atmospheric carbon dioxide (CO2) [Szidat, 2009; Currie, 2000]. The 14C measurement on the EC fraction is still challenging. Only a few groups worldwide perform such analyses, and there are still open questions concerning the best approach for the EC isolation for 14C measurements [Zhang et al., 2012; Bernardoni et al., 2013; Dusek et al., 2014]. However, a laboratory intercomparison shows that 14C results of the OC and TC fraction agree well and are not sensitive to different isolation methods [Szidat et al., 2013].
Since 14C measurements require a sufficient carbon loading, filters with sampling durations well over 12 h are typically analyzed. Szidat et al. [2004a, 2007] reported 14C results obtained from aerosol filters with a sampling time of 12 h and 8 h, respectively, but in order to collect enough carbon, each filter was used during the same period of day (morning or evening) on one to three consecutive days. Lewis and Stiles , Klinedinst and Currie , and Fushimi et al.  also used aerosol filters with relatively short sampling times (6–11.5 h) for 14C analysis. In general, carbonaceous aerosol emission sources, formation processes, and sinks are exceedingly variable and their characterization thus requires highly time resolved measurements, which are often impractical for filter sampling.
Combining 14C data with AMS-PMF results can help constrain the possible POA and SOA sources and precursors. Only a few studies [Lanz et al., 2008; Aiken et al., 2010; Minguillón et al., 2011] applied a combined 14C/AMS-PMF source apportionment where the 14C measurements were used to assign a fossil and nonfossil fraction to total SOA. However, only Hodzic et al.  and El Haddad et al.  presented such an apportionment of different SOA fractions.
In this study we present, to the best of our knowledge, the first highly time resolved (3–4 h) 14C measurements. Filter samples collected during the California Research at the Nexus of Air Quality and Climate Change (CalNex) 2010 field campaign are analyzed to obtain diurnal profiles of fossil and nonfossil carbonaceous aerosols. When combined with organic markers measured by gas chromatography interfaced to mass spectrometry (GC-MS) and PMF analysis of aerosol mass spectra, the 14C measurements provide detailed source apportionment of OA, especially focusing on SOA, in the Los Angeles basin. Some summarizing results were already shown in Bahreini et al.  (6 h average fossil and nonfossil TC concentrations for 2 days). Here the full highly time resolved data set as well as new results and analysis are presented.
2.1 Aerosol Sampling
Filter sampling was conducted during the CalNex 2010 field campaign from 15 May to 16 June 2010 in Pasadena, CA (34.1405°N, 118.1225°W). The measurement site is located on the campus of the California Institute of Technology (Caltech), about 17 km northeast of downtown Los Angeles (LA) and 7 km south of the San Gabriel Mountains. The site is located within the LA basin, which has a population of about 18 million inhabitants (United States Census Bureau: http://www.census.gov). The region is influenced by multiple emission sources from traffic, cooking, industries, agriculture, biomass burning, and shipping [Schauer et al., 1996; Williams et al., 2010; Wonaschutz et al., 2011; Bahreini et al., 2012; Hayes et al., 2013]. The general wind pattern in the LA basin during daytime is driven by a sea breeze blowing onshore nearly perpendicular to the coastline, developing around 11:00 and reaching the maximum in the afternoon [Lu and Turco, 1995]. This sea breeze advects emissions from the southwest toward the north and northeast of the LA basin. At night a land breeze develops above 200 m and lasts 5–6 h transporting air masses from the LA basin over the Santa Monica Bay [Wagner et al., 2012]. At the surface, the wind directions during the night were mostly southeast and the wind speeds were weaker than during the day [Thompson et al., 2012; Hayes et al., 2013]. The sea breeze/land breeze circulation prevailed during the sampling period as thoroughly described elsewhere [Duong et al., 2011; Washenfelder et al., 2011; Thompson et al., 2012; Wagner et al., 2012; Hayes et al., 2013]. Furthermore, the examination of the back trajectories showed that during the measurement period, the Pasadena measurement site was sometimes impacted by air masses from the California coastal mountain ranges to the north and northwest [Hayes et al., 2013].
PM1 particles were collected onto preheated (8 h in air at 800°C in a muffle furnace, model LE 14/11/B150, Nabatherm, Germany) quartz fiber filters (Pallflex 2500QAT-UP) using a high-volume sampler (Digitel DHA-80, Switzerland), at a flow rate of 30 m3h−1. The sampling time was varied between 3 h (15–26 May and 4–15 June) and 4 h (26 May to 4 June) resulting in 6 to 8 filters per day. Immediately after sampling, the filters were wrapped in aluminum foil, sealed in plastic bags, and stored at −20°C until analysis. Field blank filters (n = 4) were also collected, following 15 min exposures to ambient air and treated in the same way as the loaded samples. Filter sampling has well known but nonsystematic artifacts due to adsorption and volatilization of semivolatile compounds [Viana et al., 2006; Jacobson et al., 2000]. Since it was not possible to conduct a more complex sampling (e.g., using two sampling lines, one with and without a denuder system for volatile OC removal or using two filters in series), artifacts could not be quantified. It has to be noted that due to the low-filter loadings caused by the short sampling times, such artifacts could have a larger influence than for studies with longer filter collection times.
2.2 14C Measurement
2.2.1 Selection of Filters
Filter samples from 7 days from the second half of the CalNex campaign with high OA concentrations (afternoon peak 15–24 µgm−3) [see Hayes et al., 2013] were selected for the measurement of 14C in TC and the determination of the fraction of nonfossil carbon therein (fNF,TC). The second half of the campaign was more consistently impacted by urban pollution, with higher OA concentrations and consistent diurnal cycles, which was the target of our study, while during the first half of the CalNex campaign there was a period with higher dispersion and reduced impact of the urban emissions. Therefore, the results shown below are representative of days with high concentrations from the second half of the CalNex campaign and absolute concentrations represent an upper limit. 14C in OC was additionally analyzed for 4 out of the 7 days selected. Only days with high concentrations were chosen because the analysis methods used (14C analysis and EC/OC measurement) require a certain filter loading which was generally low due to the short sampling times. The reason for analyzing 14C in OC for only 4 days was that there is also the possibility to calculate the nonfossil fraction of OC with the 14C results of TC and the EC/OC concentrations (see section 2.2.5). This is because a comparison for measured and calculated values for nonfossil OC for the first four analyzed days showed really good agreement (see Figure 1). Therefore, we rather analyzed more days for 14C in TC and performed repeated analysis of some samples for quality assurance instead of having both results, 14C in OC and TC, for a smaller number of days. Table 1 indicates the chosen days, the sampling details, and the different types of analyses performed.
Table 1. Sampling Details and Information on the Chosen Days and Different Offline Analysis Methods
Day of the week
3 h and 4 h
14C in TC and EC/OC Concentrations
14C in OC
14C in OC
14C in OC
14C in OC
14C sample preparation technique
Sealed tube combustion
Sealed tube ombustion
2.2.2 Sample Preparation
The sample preparation for the 14C measurement was carried out using three different techniques, as summarized in Table 1.
First, OC was separated before 14C analysis using the THEODORE system [see Szidat et al., 2004b for more details]. In brief, OC is oxidized to CO2 from filter punches with a diameter of 11 to 16 mm at 340°C for 10 min in an oxygen stream. At this low temperature no significant amount of EC is expected to be released from the samples [Szidat et al., 2004b]. The evolving CO2 is separated from interfering reaction gases, cryotrapped, and sealed in glass ampoules for 14C measurements. TC samples are also prepared using the same system, where filter punches are combusted at 650°C for 12 min and the resulting CO2 is recovered as described above.
Second, OC and TC samples were also combusted with a thermo-optical OC/EC analyzer (model 4L, Sunset Laboratory Inc., USA), which is equipped with a nondispersive infrared detector. The instrument is modified such that it can be operated with pure oxygen as a carrier gas in addition to the conventionally used He and He/O2 mixture. The temperatures and combustion times were set to the same values as used in the THEODORE protocol. The Sunset analyzer is connected to the trapping part of the THEODORE system, and the resulting CO2 is recovered again as described above. A detailed description of this setup is described by Zhang et al. , where it was shown that the results obtained using THEODORE and Sunset are comparable.
Third, TC samples for 14C measurement were prepared following the approach described in Fahrni et al. , where filter punches sealed in quartz tubes together with ~0.25 g of copper oxide (small rods for elemental analysis, Fluka, Switzerland) are combusted for 4 h at 800°C in a muffle furnace (model LE 14/11/B150, Nabatherm, Germany). Good agreement (± 5% on average) for the 14C results was found between the samples prepared following this approach and those measured using THEODORE (see Figure S1 in the supporting information).
2.2.3 MICADAS Measurements
Carbonaceous fractions separated as described above were analyzed for 14C content by the MIni radioCArbon DAting System (MICADAS) [Synal et al., 2007] at the Swiss Federal Institute of Technology (ETH) Zürich. The instrument is equipped with a gas ion source [Ruff et al., 2007; Wacker et al., 2013], which allows direct CO2 injection after dilution with He [Ruff et al., 2010]. The fraction of modern carbon (fM) is derived as the ratio of 14C/12C measured in the samples to the 14C/12C ratio in 1950, which is defined as 0.95 times the value of the standard SRM4990B [Levin et al., 2010];
Values of fM are corrected for δ13C fractionation [Wacker et al., 2010] and for the 14C decay between 1950 and the year of measurement.
2.2.4 Data Correction and Presentation
Before using the fM values obtained for source apportionment of fossil and nonfossil contributions, several corrections are carried out. First, a mass dependent blank correction is applied to the measured fM following an isotopic mass balance approach [Zapf et al., 2013]:
where fM,corr is the blank-corrected fM, and fM,sample and fM,blk are the fM measured for samples and blanks, respectively. The carbon mass in the samples and the blanks are denoted by mCsample and mCblk, respectively. The mean TC concentration in the blank samples (n = 4) is 0.51 ± 0.23 μgC cm−2 with an fM value for TC (fM,TC) of 0.72 ± 0.24 (in general, throughout the manuscript, we always report average values ± standard deviation of the mean, unless it is indicated otherwise). Since no EC was detected in blank samples, the mean blank fM,TC and TC concentration were also used for the blank correction of the fM values of OC (fM,OC). The mean blank concentration contributes on average 17% ± 5% and 15% ± 4% to the OC and TC filter loading of the samples, respectively. The mass-dependent blank correction (see equation (2)) changes the measured fM values on average by ±3% and increases the uncertainty by ~3%.
Second, to account for the thermonuclear weapon tests of the late 1950s and early 1960s which increased the radiocarbon content of the atmosphere [Levin et al., 2010], fM,corr values obtained from equation (2) are converted into nonfossil fractions (fNF), following equation (3) [Szidat et al., 2006; Zhang et al., 2012]
where fNF,ref is a reference value representing the modern 14C content during the sampling period compared to 1950 before the bomb testing. The calculation of this parameter takes into account the contribution of biogenic (bio) and biomass burning (bb) sources to modern carbon:
where fbio,ref, the fM reference value for biogenic aerosols, equals 1.036 ± 0.015 for 2010. It is taken from long-term 14CO2 measurements at the background station Schauinsland [Levin et al., 2010]. The fM reference value for biomass burning aerosols fbb,ref is estimated as 1.116 ± 0.050 for 2010 using a tree growth model as described in Mohn et al. . The fraction of biogenic sources to the total nonfossil sources pbio is assumed to be 0.9 ± 0.1 since no large biomass burning events were recorded during the measurement period. For fbio,ref, fbb,ref, and pbio the uncertainties are reported which will be described in detail below (see section 2.2.6). Hayes et al.  estimated the contribution of biomass burning OA to total OA to be at most 5% using AMS and Particle Analysis by Laser Mass Spectrometry (PALMS) as well as measurements of the biomass burning tracer acetonitrile. However, due to the uncertainties of these measurement techniques and the lack of tracers for identifying aged biomass burning particles, pbio was set to 0.9 instead of 0.95. Note that pbio has only a very little impact on fNF,ref compared to other measurement uncertainties (an increase of pbio from 0.8 to 1 would change fNF,ref for this study only by 1.5%). Consequently, with a pbio of 0.9 ± 0.1, an fNF,ref value of 1.044 with an uncertainty of ± 0.02 is calculated and ultimately used to determine the nonfossil fraction in OC and TC, fNF,OC and fNF,TC, respectively.
2.2.5 Calculation of fNF,OC
Radiocarbon in OC and thus fNF,OC was only measured for 4 days (see Table 1). However, for the days where these measurements were not available, fNF,OC was determined based on fNF,TC and EC/OC concentrations (see section 2.3), following equation (5) [Szidat et al., 2004b]
Assuming a contribution of nonfossil carbon to EC, fNF,EC, equation (5) can be expressed as
A sensitivity analysis for the assumed fNF,EC covering a wide range of results (0–0.25; see Table S1 in the supporting information), including summer results from urban European cities, was carried out. Finally, a fNF,EC value of 0.1 ± 0.05, which is the average of several urban European cities in summer [Szidat et al., 2006, 2009; Minguillón et al., 2011], was chosen for the calculation of fNF,OC. Good agreement was found between measured and calculated fNF,OC for the 4 days where both values are available (difference of ± 6.9%, slope ~1, and intercept ~0; see Figure 1). Therefore, the calculated fNF,OC will be presented in this study for the days where fNF,OC was not measured. Furthermore, the sensitivity analysis assuming different fNF,EC values shows that fNF,OC is only slightly affected by the assumed values and that the calculated fNF,OC values in this study are insensitive to a change in the biomass burning contributions to fNF,EC.
It should be noted that there are differences in the car fleets between Europe (higher diesel car fraction) and the U.S. However, on the one hand, biomass burning influence during the CalNex campaign was negligible (see section 2.2.4 above) and on the other hand, fNF,EC was varied within a broad range in the sensitivity analysis which should account for possible differences (e.g., biodiesel fraction) between Europe and the U.S.
We also note that uncertainties associated with the calculation of fNF,OC are larger than for the days where it was measured which is mostly due to the large uncertainties of the EC/OC measurement caused by the low filter loadings (see section 2.2.6 and 2.3 below). This is also the reason why the calculation of fNF,EC, for the days where fNF,TC and fNF,OC was measured, using equation (5), was not carried out since the uncertainties of the calculated fNF,EC would on average be 5 times higher than the value itself. The application of different thermal-optical protocols used for the EC/OC measurement on the calculation of fNF,OC is not expected to have a large impact due to the low EC/TC ratios during the campaign (on average ~12%). The assigned uncertainties used in the error propagation of equation (6) (see section 2.2.6 below) for the OC, EC, and TC concentrations are large and therefore should well represent possible differences due to different thermal-optical protocols (see section 2.3 below).
2.2.6 Uncertainty Assessment
The uncertainties (1 sigma) of fNF,OC and fNF,TC are the combined uncertainties from the 14C measurement, the 14C blank correction, and the 14C bomb peak correction (see section 2.2.4). The uncertainty of the calculated fNF,OC (on average 14%; see section 2.2.5) was obtained by an error propagation of equation (6) and includes the uncertainties of the OC,EC, and TC mass concentration measurement (19%, 31%, and 19%, respectively; see section 2.3) as well as the uncertainty of fNF,EC (0.05; see section 2.2.5) and fNF,TC and is on average 9% higher than the average uncertainty of the measured fNF,OC values (5%). The average uncertainties of fNF,OC (measured and calculated values) and fNF,TC were calculated to be 9% and 5%, respectively. In the following, all individual uncertainties contributing to fNF will be explained.
The average uncertainty of the measured fM values of the samples presented in this study is 1.8% and is mainly driven by the counting statistics but also includes the variability of the standards and the background plus additional instrumental uncertainties [Stuiver and Polach, 1977; Wacker et al., 2010]. The uncertainty of fM,corr (the blank-corrected fM) is on average 4.7%. It was derived from the error propagation of equation (2). This includes a conservative uncertainty of 30% for fM,blk to account for additional blank variability which was not covered by the blank filters [Ceburnis et al., 2011; Dusek et al., 2013]. The uncertainty of fNF,ref (0.02) is obtained by the error propagation of equation (4) and includes the uncertainties of fbio,ref, fbb,ref, and pbio. The former was assumed to be 1.5% and includes a conservative estimation of the influence of primary biogenic OC, which could be older than the year of sampling and the natural variability of 14CO2 [Szidat et al., 2006]. The uncertainty of fbb,ref was estimated by Mohn et al.  to be 5% using a Monte Carlo simulation. The uncertainty of pbio was set to 0.1 to account for uncertainties and single high events in the biomass burning contributions found by AMS and PALMS (see section 2.2.4).
2.3 EC/OC Measurements
All samples were measured for EC and OC content by a Sunset OC/EC analyzer, following the thermal-optical transmittance (TOT) protocol EUSAAR2 [Cavalli et al., 2010]. It has to be noted here that the OC/EC determination by TOT with thermal-optical instruments is not standardized yet and that different TOT protocols (e.g., NIOSH [National Institute for Occupational Safety and Health, 1999; Peterson and Richards, 2002], IMPROVE [Chow et al., 1993], and EUSAAR2 [Cavalli et al., 2010]) can lead to different results. TC measured with different protocols usually shows a good agreement (within 10%), whereas EC can differ significantly from method to method [Chow et al., 2001; Schmid et al., 2001; Piazzalunga et al., 2011]. However, since EC concentrations were really low (EC/TC ~12%), no large differences in OC due to different protocols are expected. The measurements were repeated 2 to 4 times depending on the variability caused by the relatively low filter loadings (maximum loading of 6.4 μgC cm−2 and 1.6 μgC cm−2 for OC and EC, respectively). In addition, a blank correction was carried out using the average of four blank filters (TC = 0.51 ± 0.23 μgC cm−2). Since no EC was detected in blank samples, the mean TC concentration was also used for the blank correction of OC. The mean measurement uncertainty for OC, EC, and TC was estimated as 19%, 31%, and 19%, respectively, using the variability of all samples (n = 18) that were measured 4 times and the mean uncertainty of the blanks. This overall uncertainty is large and therefore should well represent possible differences due to different thermal-optical protocols.
2.4 AMS-PMF Data
A colocated Aerodyne high-resolution time-of-flight aerosol mass spectrometer (AMS) provided time-resolved measurements of nonrefractory organic and inorganic (sulfate, nitrate, chloride, and ammonium) submicron particles [Jimenez et al., 2003; Drewnick et al., 2005; DeCarlo et al., 2006; Canagaratna et al., 2007]. Positive matrix factorization (PMF) of the AMS organic mass spectra was used to retrieve OA sources and formation processes [Paatero and Tapper, 1994; Lanz et al., 2007; Ulbrich et al., 2009]. Details regarding the AMS operating conditions and PMF analysis can be found in Hayes et al. . Briefly, five components were identified by PMF. These include hydrocarbon-like organic aerosol (HOA), i.e., a surrogate for POA from fossil fuel combustion [Lanz et al., 2007; Aiken et al., 2008; Jimenez et al., 2009; Ulbrich et al., 2009], cooking-influenced organic aerosol (CIOA) [Schauer et al., 2002a; Mohr et al., 2012; Crippa et al., 2013a], local organic aerosol (LOA) [Aiken et al., 2009; Docherty et al., 2011; Sun et al., 2011] and semivolatile and low-volatility oxygenated organic aerosol (SV-OOA and LV-OOA), i.e., as surrogates for fresh and aged SOA, respectively [Aiken et al., 2008; Jimenez et al., 2009; Ulbrich et al., 2009].
2.5 Combined 14C/AMS-PMF Source Apportionment
To gain more insight into the origin of SOA precursors, AMS-PMF results were combined with the 14C measurements [El Haddad et al., 2013; Minguillón et al., 2011; Aiken et al., 2010; Hodzic et al., 2010; Lanz et al., 2008]. The approach is explained in the following.
The PMF factors are first converted from organic mass (OM) to OC using the OM:OC ratios determined from the elemental analysis of high-resolution AMS (OM:OCHOA = 1.3, OM:OCCIOA = 1.4, OM:OCLOA = 1.6, OM:OCSV-OOA = 1.6, and OM:OCLV-OOA = 2.2) [Hayes et al., 2013]. For the OC mass concentration of the AMS factors, the following notations, hydrocarbon-like organic carbon (HOC), cooking-influenced organic carbon (CIOC), local organic carbon (LOC), semivolatile oxygenated organic carbon (SV-OOC), and low-volatility oxygenated organic carbon (LV-OOC), are adopted throughout the manuscript. OC measured offline by Sunset (OCSunset) and measured online by AMS (OCAMS) are overall in good agreement, as shown in Figure S2 in the supporting information (r = 0.86). However, the slope between the two measurements indicates that the OCAMS values exceed the ones of OCSunset. The average ratio OCAMS/OCSunset is ~1.3. While the reason for this discrepancy remains uncertain at this stage, possible explanations include uncertainties in AMS and Sunset measurements (± 30% for AMS [Hayes et al., 2013; Middlebrook et al., 2012] and ± 19% for Sunset), evaporation losses of semivolatile compounds from filter samples and uncertainties related to the determination of AMS OM/OC ratios.
Second, the nonfossil fraction of the sum of OCAMS and EC (TCNF,AMS) is calculated by multiplying it with fNF,TC from the 14C measurement.
Third, a fossil/nonfossil fraction was assumed a priori for the primary PMF factors HOC, CIOC, and LOC as well as for EC. The sources of EC, HOC, and CIOC are relatively well constrained, whereas the origin of LOC is more uncertain (see section 3.3 below). The average nonfossil fraction of the PMF factors SV-OOC (fNF,SV-OOC) and LV-OOC (fNF,LV-OOC) is then derived by solving the following mass balance equation using a multiple linear regression analysis:
In this method, all the nonfossil fractions of the PMF factors and EC are assumed to be constant, even though these, especially fNF,LV-OOC and fNF,SV-OOC, may vary over the course of the sampling period. However, in the absence of additional information, this approach provides the best estimate of contributions of fossil and nonfossil sources to the PMF factors averaged for the measurement period, without introducing additional degrees of freedom in the regression method that can result in additional uncertainties.
A sensitivity analysis for the calculated fNF,LV-OOC and fNF,SV-OOC was performed using a pseudo Monte Carlo simulation by varying all the a priori-assumed factors (fNF,HOC, fNF,CIOC, fNF,LOC, and fNF,EC) and also including the uncertainties of the 14C and EC measurements. The detailed description about the defined range of the a priori-assumed factors and the uncertainties used in the pseudo Monte Carlo simulation as well as their influence on the result can be found in the supporting information.
2.6 Measurement of Organic Markers
The organic PM markers levoglucosan, polycyclic aromatic hydrocarbons (PAHs), oxygenated polycyclic aromatic hydrocarbons (o-PAHs), straight-chain alkanes (n-alkanes), and hopanes were measured for three complete days (see Table 1) using an in situ derivatization and thermal desorption method which is coupled to gas chromatography with mass selective detection (IDTD-GC-MS) (this method is described in detail in Orasche et al. ). Briefly, the filter punches were placed into glass liners suitable for an automated thermal desorption unit [Schnelle-Kreis et al., 2005]. Isotope-labeled standard compounds were spiked directly onto the filter surface to account for influences of the matrix for later quantification. Derivatization was performed on the filter by adding of liquid derivatization reagent N-methyl-N-trimethylsilyl-trifluoroacetamide (MSTFA, Macherey-Nagel, Germany). During 16 min of desorption time, in addition, an in situ derivatization with gaseous MSTFA was carried out to quantitatively silylate polar organic compounds and optimize the desorption process. The derivatization procedure has been automated as well. Derivatized and desorbed molecules were first trapped on a precolumn before separation by gas chromatography (BPX-5 capillary column, SGE, Australia). The detection and quantification of compounds was carried out on a Pegasus III time-of-flight mass spectrometer using the ChromaTOF software package (LECO, St. Joseph, MI).
3 Results and Discussion
3.1 Fossil and Nonfossil Fractions of TC and OC
Figure 2 presents the results for fNF,OC and fNF,TC for all days analyzed. The data for all the individual filter samples including the raw data from the 14C measurement are listed in Table S2 in the supporting information. The total average of fNF,TC (0.51 ± 0.15; individual values range from 0.29 to 1.00) was found to be slightly lower than the average of fNF,OC (0.58 ± 0.15, 0.32–1.03), due to the contribution of fossil fuel emissions to EC. The average of fNF,OC including only the days where it was measured (see Table 1) is 0.60 ± 0.17 and is statistically not different (t test, 95% confidence interval) from the average overall values (measured and calculated). Given the low contribution of biomass burning during the measurement period (see section 2.2.4 above), this result reveals the significant contribution of nonfossil sources such as biogenic SOA and cooking to atmospheric OA, despite extensive fossil emissions in the region (i.e., more than 50% of OC is nonfossil). Very similar 14C results (fNF,TC ~0.47) for the annual average of 2007 were reported for Wilmington, a district of LA adjacent to the ports of Los Angeles and Long Beach [Buchholz et al., 2012]. Previous 14C data reported from two stations for this region (in Long Beach and Azusa) in 1982 (four filters for each location with a sampling time of 3 months covering the whole year from January to December) suggest lower relative contributions from nonfossil sources (0.32–0.43 and 0.2–0.35) [Hildemann et al., 1994]. It is noted that while the absolute concentration of nonfossil TC (TCNF) remained relatively constant between 1982 and 2010 (~3 μgC m−3), a noticeable decrease is observed in fossil TC (TCF) during this time period (~7 μgC m−3 in 1982 compared to ~3 μgC m−3 in 2010). This decrease is expected considering the air quality mitigation strategies implemented since 1982.
The most prominent feature in Figure 2 is the fNF diurnal pattern, with increasing values during the night and early morning (on average 0.71 ± 0.12 for fNF,OC and 0.60 ± 0.15 for fNF,TC) and clear minima in the early afternoon (0.42 ± 0.08 and 0.38 ± 0.09 for fNF,OC and fNF,TC, respectively), coinciding with the transport of the western LA basin plume to Pasadena by the sea breeze. The diurnal patterns for all the selected days are similar, and the variability of the average diurnal cycle (standard deviation of the mean) is 5–14% for fNF,OC and 3–12% for fNF,TC (see Figure 2, left). Nevertheless, it should be noted that two samples from 5 June and all samples from 30 May show exceptionally higher nonfossil fractions (24 h average of fNF,OC is 19% and fNF,TC is 30% higher on 30 May; the average for both fNF,OC and fNF,TC for the two samples on 5 June from 09:00 to 12:00 is 25% higher than for the other days). The measurements of those filters were performed several times, excluding possible contamination during sample preparation. The elevated fNF appears consistent with an increase in the OC/EC and OM/HOA ratios (by a factor of ~2) during 30 May, suggesting an additional input from nonfossil sources to OC. Furthermore, levoglucosan concentrations on 30 May are a factor of ~2 higher than on 4 June and 14 June (see Figure S3). These findings suggest that enhanced charbroiling could be the reason for the higher fNF values, since in 2010, 30 May was the Sunday before the Memorial Day (a public holiday). No elevated fNF values were found on the other two Sundays, 6 June and 13 June. Buchholz et al.  identified anthropogenic inputs of TC in the LA basin close to the port of Long Beach with excessive 14C content (fNF,TC values of up to 3.4). However, sources of anthropogenic 14C are unlikely the reason for the elevated fNF values on 30 May and 5 June due to the following reasons:
No fNF values > 1 were measured in Pasadena in this study, whereas almost all samples in Buchholz et al. , which showed anthropogenic 14C inputs (8 out of 34), exhibit fNF values of >1.4.
A contamination during handling before and after the filter sampling before filters were frozen and shipped to Switzerland can be excluded since filters from no other day show elevated fNF values, and it is very unlikely that an artificial 14C source would contaminate only samples from 1 day during handling.
14C results from the measured blank filters (n = 4) showed no elevated fNF,TC values (results range from 0.62 to 0.88) including one blank from 30 May (fNF,TC = 0.75).
Anthropogenic sources of 14C (nuclear power plants, pharmaceutical industry, biochemical laboratories, …) are point sources and show a high temporal variability. Therefore, the high time resolution filter sampling conducted in this study together with the consistent daily change in wind direction (land/see breeze system) would be ideal to detect such inputs. However, fNF values of all analyzed days show very consistent diurnal cycles without large increases in the data.
The distance between Pasadena and Long Beach is large (~40 km), and Buchholz et al.  stated that sources of anthropogenic 14C with distances of >50 km to the sampling site are unlikely to have a large influence on 14C results of aerosol filter samples.
Nevertheless, since 30 May shows clearly different fNF values than the other days and as we cannot completely rule out a low contribution from sources emitting anthropogenic 14C during that day, this period was excluded from the analyses below. No anomalous OC/EC and OM/HOC ratios were found during 5 June compared to the other days to explain the unexpected nonfossil amplitude around noon, and therefore, the results from these two filter samples were included in the average diurnal cycle. As shown in Figure S4 (see supporting information), considering 30 May in the average has only a slight influence on the results (fNF,OC and fNF,TC are 5% and 6% more modern, and the variability increases from 8% to 12% and from 7% to 13% for fNF,OC and fNF,TC, respectively). We note that anthropogenic sources of modern carbon such as those identified by Buchholz et al.  could potentially increase the background 14C levels slightly. A 10% increase of the background 14C (i.e., fNF,ref in equation (3) increases from 1.04 to 1.14), which is unlikely due to the highly variable emissions of only a few anthropogenic 14C point sources, would decrease the fNF values on average by only ~5%. In view of this result, the absence of emission inventories for anthropogenic 14C sources enabling an accurate assessment of their influence, and as there is no evidence in our data that would point out to such contamination (see discussion above), we will not consider such correction in the following.
We also tested possible differences in the average diurnal cycle of fNF,OC using only the data from the measured days versus including all data (measured and calculated). As shown in Figure S5 (see supporting information) there are no significant differences between the two different average diurnal cycles. It should also be noted that only 7 days with high OA concentrations (on average 10.6 µgm−3) were selected. However, the relative fossil and nonfossil OC and TC contributions from the seven selected days which exhibit high OA concentrations may be considered representative of the second half of the CalNex campaign since the average relative contribution of the different sources to OA (HOA ~14%, LOA ~5%, CIOA ~12%, LV-OOA ~28%, and SV-OOA ~40%) and EC/OA (6%) for the seven selected days are very similar to the campaign average from 15 May to 16 June 2010 (HOA/OA ~12%, LOA/OA ~5%, CIOA/OA ~17%, LV-OOA/OA ~32%, SV-OOA/OA ~34%, and EC/OA ~7%) [see Hayes et al., 2013]. Furthermore, days from the second half of the CalNex campaign (also days with lower concentrations) show a very similar diurnal pattern (total OA and PMF factors), and also, the campaign average diurnal cycle of the PMF factors is very much consistent to the one obtained here [see Hayes et al., 2013] (section 3.3.2 below), signifying that OA dynamics may be driven by the same sources/processes.
3.2 Levels and Variability of Fossil and Nonfossil Carbon
Combining the 14C and OC/EC measurements gives access to the absolute concentrations of fossil and nonfossil OC and TC, whose time series and diurnal profiles are presented together with those of EC and PMF factors in Figure 3. OC and TC concentrations consistently rise during the day, with local maxima that extend from about 13:30 to 16:30 (6.5–8.1 μgC m−3 and 7.1–8.8 μgC m−3 for OC and TC, respectively), and coincide with the arrival of the western LA basin plume. The lowest OC and TC concentrations are observed during the night and early morning (2.4–3.6 μgC m−3 and 3.0–3.9 μgC m−3 for OC and TC, respectively). The EC concentration is on average 0.6 ± 0.4 μgC m−3 (0.2–2.1 μgC m−3) and accounts only for 12% ± 7% of TC. Examining the fossil and nonfossil origins of OC and TC clearly reveals that the nonfossil concentrations remain relatively constant throughout the day (see Figure 3). This indicates a nonfossil background of OC, ranging between 2 and 3 μgC m−3, with a minor afternoon increase of ~1 μgC m−3 (~30%–50% compared to the nonfossil background). In contrast, the observed afternoon increase in OC and TC concentrations is found to mainly stem from fossil emissions, with an average increase in OCF and TCF concentrations of ~3 μgC m−3 (~230% and ~160% for OCF and TCF, respectively) compared to the average fossil concentrations from 0 h to 9 h. We recognize that the concentrations shown here represent most probably an upper limit for the CalNex campaign, since only days with high concentrations from the second half were selected for the analysis, where the measurement site was more consistently impacted by urban pollution.
OCF shows moderate or low correlation with EC (r = 0.28) and HOC (r = 0.42), although the latter two are directly emitted species from fossil fuel combustion mainly present in diesel exhaust emissions (see Figure 4 and Table 2). Instead, OCF correlates more strongly with PAHs (r = 0.79) and hopanes (r = 0.76) which are other primary fossil fuel combustion markers [Robinson et al., 2006; Subramanian et al., 2006]. These results suggest that OCF may have a higher contribution from sources with elevated PAHs and hopanes, versus sources with significant concentrations of EC. This would be expected from sources such as gasoline vehicles [Schauer et al., 2002b]. Furthermore, these high correlations between the PAHs and hopanes with OCF demonstrate again the influence of the transport of the emissions of the western LA basin to Pasadena with the sea breeze. Additionally, OCF correlates also really well with SV-OOC (r = 0.87) and o-PAHs (r = 0.84), which both show increasing contributions in the early afternoon (see Figures 4 and S6). These significant correlations of OCF with secondary species together with their much higher concentrations compared to primary species (e.g., o-PAHs/(PAHs + hopanes) ~3) indicate that the increase of the OC concentration for the selected days in the early afternoon is mainly due to fresh anthropogenic SOC transported to Pasadena and formed within the western LA plume from fossil precursors. Combined with the low correlation between OCF and EC and the fact that we do not observe a weekend to weekday difference in the OCF levels and fractions, which would be expected due to the well-known large decrease of on-road diesel vehicle activity in the LA basin and thus significantly reduced specific VOCs emissions [Marr and Harley, 2002; Pollack et al., 2012], this observation supports the suggestions of Bahreini et al.  and Hayes et al.  that gasoline emissions may play a larger role than diesel emissions in the production of anthropogenic SOA in the LA basin. However, it should be noted that the correlation between OCF and the organic markers (PAHs, o-PAHs and hopanes; see Figure 4 (top)) includes less data than the correlation between OCF and the AMS/PMF data. In the next section, we assess the relative contribution of fossil and nonfossil precursors in the production of fresh and aged SOA.
Table 2. Pearson Correlation Coefficient for the Regression Analysis of the PMF Factors and EC as Well as the Total Mass Concentrations of PAHs, o-PAHs, and n-alkanes Versus OCNF and OCFa
aData from 30 May are excluded (see section 3.1). Correlation coefficients larger than 0.5 are marked in bold.
3.3 Combined 14C/AMS-PMF Source Apportionment
3.3.1 Range and Sensitivity of Results
Additional insights into the sources of OC can be gained by combining 14C measurements with AMS/PMF results. A major goal for the CalNex experiment is to improve the understanding of the formation and sources of SOA in California. The work presented in this section makes a contribution toward this goal by quantifying the fossil and nonfossil fractions of the secondary PMF factors, SV-OOC and LV-OOC. This apportionment was achieved using a multiple regression analysis (see equation (8)), as described in section 2.5. In the analysis, the fossil and nonfossil origins of the three primary PMF factors (CIOC, HOC, and LOC) and EC are assumed a priori, where fNF,CIOC, fNF,HOC, and fNF,LOC as well as fNF,EC denote the nonfossil fraction in CIOC, HOC, LOC and EC, respectively. The sensitivity of the analysis outputs to these assumptions is examined via a pseudo Monte Carlo simulation by varying fNF,CIOC, fNF,HOC, fNF,LOC, and fNF,EC in equation (8) within a predetermined range. This yields a range of nonfossil LV-OOC and SV-OOC fractions (fNF,LV-OOC and fNF,SV-OOC, respectively), which are plotted in Figure S7. A detailed description about the predetermined range of the a priori-assumed nonfossil fractions and uncertainties which were fed into the pseudo Monte Carlo simulation as well as their effects on the results can be found in the supporting information.
In brief, it appears that fNF,SV-OOC is insensitive while fNF,LV-OOC is quite sensitive to the a priori-assumed nonfossil fractions (see Figure S7). An increase/decrease of 0.1 of fNF,CIOC, fNF,HOC, fNF,LOC, and fNF,EC changes fNF,LV-OOC by ± 6%, 3%, 7%, and 4%, respectively, whereas fNF,SV-OOC is modified by less than 1%. It should be noted though that a lot of combinations yield “unrealistic” values for fNF,LV-OOC >1, but for realistic combinations, the analysis shows that fNF,LV-OOC never decreases below 0.69, even for very high contributions of nonfossil carbon assumed for the primary fractions (e.g., fNF,CIOC = 1 fNF,HOC = 0.1, fNF,LOC = 1, and fNF,EC = 0.2). This illustrates that LV-OOC, a proxy for aged SOC, is dominated by nonfossil precursors. By contrast, the oxidation of fossil emissions seems to produce substantial amounts of fresh fossil SOC, significantly contributing to SV-OOC, on average by 71%. An advantage of the pseudo Monte Carlo simulation used here is that it clearly shows that this difference in the origin of SV-OOC and LV-OOC is observed independently of the assumptions and the uncertainties related to the calculation.
The uncertainties of fNF,SV-OOC and fNF,LV-OOC (fNF,OOC) are defined as the maximum change in the corresponding fNF,OOC due to changes in the a priori assumptions and/or the consideration of the uncertainties of fNF,TC and/or fNF,EC as well as EC and are estimated by the simulation on average to be 18% for fNF,LV-OOC and 3% for fNF,SV-OOC (see Figure S7 in the supporting information).
3.3.2 Best Estimate
Figure 5 presents our best estimate of the fossil and nonfossil contributions to the PMF factors for the seven selected days, illustrated as an average diurnal cycle. In this solution fNF,HOC and fNF,EC were set to 0.05 and 0.1, respectively, to account for the utilization of biofuels (see supporting information). The value 0.1 for the latter also represents the average of 14C measurements of EC from several urban European areas in summer. Fraction fNF,LOC was fixed at 0.5 since the origin of LOC is not known (see supporting information), and considering its small contribution (~5% of OCAMS mass), different fNF,LOC values do not have a large effect on the calculation of fNF,LV-OOC and fNF,SV-OOC (a change of fNF,LOC by ± 0.3 would lead to an increase/decrease of fNF,LV-OOC and fNF,LV-OOC of 7.6% and <2%, respectively; see Figure S7). Considering these values, the only possible solution with fNF,LV-OOC < 1 is obtained for fNF,CIOC equal 0.9 or 1. We consequently set fNF,CIOC to 1 which is expected if CIOC is completely due to nonfossil carbon in cooking emissions. Values of fNF,LV-OOC and fNF,SV-OOC were found to be 0.89 and 0.29, respectively. Note that this is the most likely solution, and that while fNF,SV-OOC is insensitive to assumption, fNF,LV-OOC can vary with the a priori assumptions, albeit with values strictly above 0.69.
From Figure 5, it appears that nonfossil OC is mainly composed of LV-OOCNF, SV-OOCNF, and CIOCNF. CIOCNF contributes on average 13.9% ± 8.2% to total OC (0.77 ± 0.37 μgC m−3) and increases in the evening due to enhanced cooking emissions into a shallower boundary layer. The model attributes the slight but regular increase (~1 μgC m−3) observed in the nonfossil carbon (Figure 3) in the early afternoons to SV-OOCNF. Such an increase could be due to SOC formed from local biogenic emissions, as has been previously reported at several locations and attributed to the daytime increase of biogenic emissions (mainly terpenes) and photochemical activities [Slowik et al., 2010; El Haddad et al., 2013] and/or due to SOC formation from cooking emissions, which can be substantial. This regular increase in nonfossil SV-OOA in the early afternoon may partially be induced by the anthropogenic emissions of NOx, OH, and additional particle mass, which enhance SOA formation [Donahue et al., 2006; Hoyle et al., 2007, 2009]. The SV-OOCNF levels range from 0.22 μgC m−3 during the night to 1.89 μgC m−3 in the early afternoon and account on average for 10.8% ± 5.3% of OCAMS (0.77 ± 0.63 μgC m−3).
LV-OOCNF concentrations show no or only a weak diurnal variation, indicative of an aged background aerosol, as opposed to a fresh SOA locally formed during midday peaks of photochemistry (see Figure 5). This is consistent with the findings by Hayes et al. , showing the omnipresence of LV-OOC even in nonprocessed air masses and the association of OOA with aged OA in previous studies [Jimenez et al., 2009]. Marine emissions might be suspected as a potential source of the background nonfossil OC, considering the proximity of our site to the Pacific and the observation of contributions from these emissions to biogenic OC in urban and marine environments [Crippa et al., 2013b; Ceburnis et al., 2011]. However, no important contribution of marine sources to PM1 OC was found during the CalNex 2010 campaign [Hayes et al., 2013]. Meanwhile, the examination of the back trajectories brings evidence that during the measurement period, the sampling site was sometimes impacted by air masses from the California coastal mountain ranges to the north and northwest where biogenic emissions are expected to be strong [Hayes et al., 2013]. This suggests that most of the LV-OOC may be dominated by continental biogenic SOC which accounts to a large extent for the nonfossil background in the LA basin on the seven selected days presented here. LV-OOCNF contributes on average 1.28 ± 0.13 μgC m−3 (22.1% ± 5.2%) constituting the largest fraction (~30%) of OCAMS during nighttime and in the early morning.
Fossil OC is mainly composed of SV-OOCF and HOCF. While HOCF (0.99–1.58 μgC m−3) levels show a relatively small increase around midday (following the EC concentrations as they are both advected from the morning emissions in the downtown LA area), the large increase in fossil carbon contribution in the early afternoon is almost entirely attributed to SV-OOCF (0.51 μgC m−3 in the night and 4.42 μgC m−3 in the afternoon), which accounts for ~40% of the total OC at that time. This corroborates the above hypothesis that the increase in the PM concentrations after the arrival of the western LA basin plume to Pasadena is mainly caused by fresh anthropogenic SOC formed from fossil precursor gases.
While it is generally thought that much LV-OOC arises from additional aging of SV-OOC species [Jimenez et al., 2009], the above discussion clearly shows that in the LA basin, for the seven selected days presented here, the two oxygenated factors derive from mostly distinct precursors. LV-OOC is overwhelmingly composed of nonfossil biogenic carbon (>69%), whereas a large fraction of SV-OOC is comprised of anthropogenic fossil secondary OC (SOCF) (~71%), produced from fossil precursors within the western LA basin plume during advection to Pasadena. At least a fraction of the SV-OOCNF is likely due to SOA formation from cooking emissions and thus is also anthropogenic. It is likely that the further oxidation of SV-OOCF, which would produce LV-OOCF can only be observed downwind of Pasadena. Overall, we conclude that in our conditions, 2.0 ± 1.5 μgC m−3 and 2.1 ± 0.7 μgC m−3 of SOC originate from fossil and nonfossil precursors, respectively. This implies that on a 24 h average basis, 56% ± 10% of SOC (33% ± 2% of total OCAMS) are related to local and regional biogenic and other nonfossil (e.g., cooking) SOA, despite extensive fossil fuel usage (traffic, industry, refineries, …) in the area.
3.3.3 Comparison With Results From Other Studies
The fossil contributions of SOC found in the LA basin on the seven selected days contrast significantly to the moderate levels reported for urban background stations in several European cities (SOCF = 0.5 ± 0.4 µg C m−3 and SOCNF = 1.2 ± 0.2 μgC m−3, Barcelona [Minguillón et al., 2011], Paris [Freutel et al., 2013; Crippa et al., 2013b], Marseille [El Haddad et al., 2013], and Zürich [Lanz et al., 2007, 2008; Sandradewi et al., 2008]). At these European cities, the fraction of SOCF to total SOC (25% ± 13% in Europe) is on average ~20% lower than in Pasadena, despite the higher EC concentrations (ECEurope = 1.9 ± 0.8 μgC m−3 versus ECLA = 0.6 ± 0.4 μgC m−3). This, together with the higher fraction of gasoline cars in the U.S. compared to Europe, support the hypothesis proposed by Bahreini et al.  and Hayes et al.  and the findings in section 3.2 that gasoline emissions are likely to dominate over diesel in the formation of fossil SOA in the LA basin during the 7 days selected for this study. It has to be noted that possible different photochemical ages at the different locations were not taken into account for this comparison. However, Platt et al.  found no significant changes in the SOA yields for gasoline cars after photochemical aging in chamber experiments after 4 h. Therefore, the different photochemical ages at the different stations may have a limited effect on the comparison of the fossil SOA levels.
In this study we provide time-resolved measurements of radiocarbon for OC and TC in PM1 for seven selected days from the second half of the CalNex 2010 study at the Pasadena ground site located in the Los Angeles (LA) basin. To the best of our knowledge, this is for the first time that 14C measurements were carried out on such highly time resolved ambient aerosol filter samples. As a consequence, a diurnal cycle for the fossil and nonfossil fractions of OC and TC could be demonstrated.
The results show a distinct and consistent diurnal pattern for all selected days. The nonfossil fractions are highest during nighttime and early mornings (fNF,OC = 0.71 ± 0.12 and fNF,TC = 0.60 ± 0.15 for TC from 24:00 to 06:00). In contrast, the carbonaceous aerosol is clearly influenced by fossil sources in the early afternoon (0.42 ± 0.08 and 0.38 ± 0.09 for fNF,OC and fNF,TC, respectively), coinciding with the transport of the western LA basin plume to Pasadena by the sea breeze. The campaign average of 0.51 ± 0.15 for fNF,TC was found to be slightly lower than for fNF,OC (0.58 ± 0.15), due to the contribution of fossil fuel emissions to EC. OC and TC concentrations also show a clear diurnal cycle with a maximum after the arrival of the western LA plume. The nonfossil OC and TC concentrations stay nearly constant throughout the day (2–3 μgC m−3), indicating a nonfossil background in the LA basin on the seven selected days. Furthermore, EC concentrations are low (0.6 ± 0.4 μgC m−3 on average) and show only a weak diurnal cycle. The afternoon increase of the concentrations primarily stems from fossil fuel combustion emissions since fossil OC and TC exhibit a strong increase (~3 μgC m−3).
We combined the 14C results with measurements of organic markers by gas chromatography mass spectrometry (GC-MS) and the source apportionment of OA by positive matrix factorization (PMF) analysis of aerosol mass spectrometry (AMS) data. The high correlation of fossil OC with semivolatile oxygenated organic carbon (SV-OOC) (r = 0.84), a surrogate for fresh secondary organic carbon (SOC), and the organic marker oxygenated polycyclic aromatic hydrocarbons (o-PAHs) (r = 0.87) indicate that the increase of the concentrations when the plume of the western LA basin arrives to Pasadena is mainly caused by fresh fossil SOC. Furthermore, we were able to assign a nonfossil fraction to the secondary PMF factors semivolatile oxygenated organic carbon SV-OOC and low-volatility oxygenated organic carbon LV-OOC. The latter primarily (by >69%) stems from nonfossil sources, whereas SV-OOC is mainly (68%–74%) formed from fossil precursors. Cooking is estimated to contribute at least 25% of OCNF (not accounting for possible SOA formation from this source), underlining the importance of urban nonfossil OC sources that are often ignored in such analyses.
Finally, the correlation between OCF and the primary species EC (r = 0.28) and HOC (r = 0.42) is much weaker than for the secondary species. Furthermore, no weekend to weekday difference in the OCF levels and fractions, which would be expected due to the well-known large decrease of on-road diesel vehicle activity on weekends in the LA basin and thus significantly reduced specific VOCs emissions, was observed. In addition, SOCF levels and SOCF to total SOC ratios for several European cities (Barcelona, Paris, Marseille, and Zürich), where the fraction of diesel passenger cars is higher than in the U.S., are on average lower by a factor of ~4 and 20%, respectively, despite the higher EC levels (2–3 times higher than in Pasadena). These findings corroborate the results of Bahreini et al.  and Hayes et al.  that gasoline emissions are likely to dominate over diesel in the formation of fossil SOA in the LA basin. This would have significant implications for our understanding of the contribution of on-road vehicles to ambient aerosols and merits further study.
Organization of the Pasadena ground site was led by the Jochen Stutz (University of California, Los Angeles) and John Seinfeld (California Institute of Technology, Caltech) as well as Joost de Gouw (National Oceanic and Atmospheric Administration, NOAA). Furthermore, we would like to thank CARB and NOAA for support of the site setup and the U.S. National Science Foundation (NSF) for the funding of the travel and shipping costs. José L. Jimenez and Patrick L. Hayes acknowledge support from CARB 08–319 and CARB 11–305 as well as DOE (BER, ASR Program) DE-SC0006035, DE-SC0006711, and DE-FG02-11ER65293. Patrick L. Hayes acknowledges a fellowship from the CIRES Visiting Fellows Program.