Corresponding author: Q. Zhang, Department of Environmental Toxicology, University of California, One Shields Ave., Davis, CA 95616, USA. (firstname.lastname@example.org)
 Organic aerosols (OA) were studied in Fresno, California, in winter 2010 with an Aerodyne High Resolution Time-of-Flight Aerosol Mass Spectrometer (HR-ToF-AMS). OA dominated the submicron aerosol mass (average = 67%) with an average concentration of 7.9μg m−3 and a nominal formula of C1H1.59N0.014O0.27S0.00008, which corresponds to an average organic mass-to-carbon ratio of 1.50. Three primary OA (POA) factors and one oxygenated OA factor (OOA) representative of secondary OA (SOA) were identified via Positive Matrix Factorization of the high-resolution mass spectra. The three POA factors, which include a traffic-related hydrocarbon-like OA (HOA), a cooking OA (COA), and a biomass burning OA (BBOA) released from residential heating, accounted for an average 57% of the OA mass and up to 80% between 6 – 9 P.M., during which enhanced emissions from evening rush hour traffic, dinner cooking, and residential wood burning were exacerbated by low mixed layer height. The mass-based size distributions of the OA factors were estimated based on multilinear analysis of the size-resolved mass spectra of organics. Both HOA and BBOA peaked at ∼140 nm in vacuum aerodynamic diameter (Dva) while OOA peaked at an accumulation mode of ∼460 nm. COA exhibited a unique size distribution with two size modes centering at ∼200 nm and 450 nm respectively. This study highlights the leading roles played by anthropogenic POA emissions, primarily from traffic, cooking and residential heating, in aerosol pollution in Fresno in wintertime.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
 Atmospheric particles have significant impacts on the earth's climate system [e.g., Pöschl, 2005], human health [e.g., Pope and Dockery, 2006] and ecosystems [e.g., Carslaw et al., 2010]. The organic portion of the atmospheric aerosol, a.k.a. organic aerosol (OA), usually constitutes a large fraction (∼20–90%) of the fine particulate mass worldwide [Kanakidou et al., 2005; Zhang et al., 2007]. However, atmospheric OA is poorly understood and more investigations are required to advance understandings of its characteristics, sources, evolutionary processes, and fates in the atmosphere. This information is also needed to elucidate the climate and health impacts of aerosols [Ghan and Schwartz, 2007].
 Although many studies were conducted on aerosols in Fresno, most of them used measurement techniques that acquire data with time resolution of at least a few hours. However, high time resolution measurements are necessary to better interpret aerosol sources and processes [Wexler and Johnston, 2008]. For example, Watson and Chow  observed a rapid increase of nitrate level during a wintertime PM2.5episode in Fresno that can only be captured by measurements at 5–10 min resolutions. So far, only a few aerosol measurements in Fresno used high time-resolution instruments, including sulfate and nitrate analyzers [Watson et al., 2006a], scanning mobility particle sizer (SMPS) [Watson et al., 2006a, 2006b; Hering et al., 2007], semi-continuous black carbon (BC) and elemental carbon (EC) measurements [Park et al., 2006], and single particle mass spectrometry (SPMS) [Qin and Prather, 2006; Bein et al., 2009].
 Here we report the deployment of an Aerodyne High-Resolution Time-of-Flight Aerosol Mass Spectrometer (HR-ToF-AMS) in Fresno in winter 2010. The AMS is a real-time instrument designed to measure the size-resolved chemical compositions of submicron aerosols (PM1) with time resolution of seconds to minutes [Jayne et al., 2000; Canagaratna et al., 2007]. The HR-ToF-AMS is further able to determine the high-resolution mass spectra (HRMS) and the average elemental ratios (e.g., O/C, H/C, and N/C) of organics [DeCarlo et al., 2006; Aiken et al., 2008]. In addition, factor analysis of the AMS mass spectral data can identify a small number of categories indicative of the origin and processing history of the OA [Zhang et al., 2011]. This study was part of the San Joaquin Valley Aerosol Inhalation Toxicology project led by the San Joaquin Valley Aerosol Health Effects Research Center (SAHERC: http://saherc.ucdavis.edu/). The goal of the project was to examine the sources of airborne PM and identify the link between source-specified PM with human health effects. In this paper, we focus on discussing the chemical composition, sources, atmospheric processes, dynamic variations, and size distributions of organic particulate matter based on analyzing the HR-ToF-AMS data.
2. Experimental Methods
2.1. Sampling Site and Instrumentation
 The sampling site was located at the University of California Center in Fresno (36°48′35.26″N, 119°46′42.00″W). The site is surrounded by residential area to the north and a commercial center to the south and is ∼600 m to the east of a highway (Yosemite FWY-41). The field campaign was operated from Jan 9 through Jan 23, 2010. The weather during this campaign was generally wet (average relative humidity = 85%) and cool (average temperature = 9.8°C). The time series of air temperature and relative humidity (RH) over the campaign are presented in Figure S1 in Text S1 in theauxiliary material. The first week of the campaign was characterized by three dense fog events, which all began at night and lasted till next morning. The second week was influenced by persistent rainfall. Details of the sampling site and other meteorological conditions are given by Ge et al. .
 An HR-ToF-AMS and an SMPS were deployed to characterize the chemical composition, number- and mass-based size distributions of submicron aerosols (PM1) [Ge et al., 2012]. Calibrations of the HR-ToF-AMS for particle sizing and ionization efficiency (IE) were performed before and during the study period following the standard protocols [Jayne et al., 2000; Jimenez et al., 2003; DeCarlo et al., 2006]. A diffusion dryer was used to remove moisture in aerosols before the SMPS but not the AMS. However, due to temperature differences between indoor and outside, the RH of aerosols sampled by the AMS was on average only 33% (20–60%), indicating that particles were overall dry. The residence time of particles in the sampling line was ∼7 s. The AMS was operated under V and W ion time-of-flight modes alternatively every 5 min to achieve both sensitive measurements of mass changes (V-mode) and better chemical characterization (W-mode, mass resolution of ∼5000 in this study) of the NR-PM1species. Under V-mode, the instrument also switched between mass spectrum (MS) and particle time-of-flight (PToF) modes, spending 6 and 9 s, respectively.
2.2. Data Analysis
 The HR-ToF-AMS data were processed using the standard AMS data analysis software SQUIRREL v1.51H and the PIKA module v1.10H (available for download athttp://cires.colorado.edu/jimenez-group/ToFAMSResources/ToFSoftware/index.html) written in Igor Pro 6.22A (Wavemetrics, OR, USA). The elemental ratios among oxygen (O), carbon (C), hydrogen (H), nitrogen (N), and sulfur (S) and the organic mass-to-carbon ratio (OM/OC) of OA were determined via analyzing the HRMS from W-mode measurements. A collection efficiency (CE) = 0.5 was assumed for determining aerosol mass concentrations (seeGe et al. [2012, Figure 2] for comparisons between the total NR-PM1mass from the HR-ToF-AMS versus the apparent particle volume from the SMPS). Details regarding modification of the fragmentation table, determination of relative ionization efficiency (RIE) and data inter-comparisons are described byGe et al. .
 The hourly RH data, CO, CO2 and NO2 concentrations were acquired from the California Air Resource Board (CARB) website (http://www.arb.ca.gov/html/ds.htm). The data reported in this paper are in the Pacific Standard Time (PST), which was the local time, and was 8 h earlier than the Coordinated Universal Time (UTC).
 PMF analysis was performed using the PMF2.exe algorithm (v4.2) in robust mode [Paatero and Tapper, 1994]. The input data were the ion-speciated HRMS matrix and the corresponding error matrix, and the analysis was conducted using the PMF Evaluation Toolkit (PET) v2.03 [Ulbrich et al., 2009] downloaded from: http://cires.colorado.edu/jimenez-group/wiki/index.php/PMF-AMS_Analysis_Guide. The OA mass spectral and error matrices were both refined prior to PMF analysis according to the protocol summarized by Ulbrich et al. . Only ions with m/zup to 120 were included in PMF because: 1) the mass resolution of the HR-ToF-AMS is not sufficient to accurately separate ions at largerm/z's, and 2) ions larger than 120 amu tend to have low signal-to-noise (S/N) ratios and account for only a minor fraction (<5%) of the total organic signal. Isotopic ions were removed since their signals are not measured directly but scaled to their parent ions. In addition, 16 runs (<1% of total number of runs) corresponding to huge spikes in mass loading were downweighted by increasing their errors by a factor of 3. These treatments largely improve the OA factorization, but have negligible impact on the mass concentrations. The error matrix, which was determined as the sum in quadrature of Poisson counting statistics and electronic noises for each ion [DeCarlo et al., 2010], is critical to proper PMF analysis. A minimum error threshold was set in order to take into account ions that have arrived yet were not counted by the mass spectrometer. Ions with S/N ratio <0.2 were removed from both the error and data matrices, while ions with S/N ratios between 0.2 and 2 were downweighed by increasing their errors by a factor of 2 [Ulbrich et al., 2009]. H2O+, HO+, CO+ and O+ ions were removed since their signals were only scaled to the signal of CO2+, which would overweight the contribution of CO2+, but they were re-included in the mass spectra of OA factors after the PMF analysis.
 The PMF algorithm is able to provide a number of mathematically meaningful solutions, but choosing one that gives physically meaningful factors requires careful evaluation. Following the recommendations outlined by Zhang et al. , we carefully examined the time series, diurnal patterns, spectra profiles for a number of alternative PMF solutions, such as solutions at different numbers of factors (p) from 1 to 10, and with different rotational parameters (fPeak)from −0.5 to 0.5 (step: 0.1). The 4-factor solution withfPeak = 0 (Q/Qexp = 1.92) was chosen. A summary of the key diagnostic plots is presented in Figure S2. The 4-factor solution is able to reconstruct the total OA mass and its temporal profile very well (Figure S2f). Variations among the results at different fPeaks are small (Figure S2b). The fitting residuals in terms of both time series and mass profiles show near normal distributions, indicating that the solution is robust and representative, not strongly influenced by a few outlier runs or m/z's. The mass spectra and temporal changes of the 3-factor and 5-factor solutions are illustrated inFigure S3. Obviously, the 3-factor solution does not separate correctly hydrocarbon-like OA (HOA) (Section 3.2.1) and cooking OA (COA) (Section 3.2.2). On the other hand, the 5-factor solution, which identifies two oxygenated factors (O/C = 0.35 and 0.46, respectively), shows indications of factor splitting [Ulbrich et al., 2009]. For example, the sum of the two oxygenated factors identified in the 5-factor solution correlates very well with the one OOA factor from the 4-factor solution (R2 = 0.96). The sum also shows much better correlations with secondary aerosol tracer species, e.g., nitrate, sulfate, and methanesulfonic acid (R2= 0.67 – 0.83), than individual oxygenated factors do. In addition, among all PMF solutions, the mass spectra of the 4-factor solution agree the best with the reference spectra from specific sources or other ambient AMS measurements (downloaded fromhttp://cires.colorado.edu/jimenez-group/AMSsd/). These results indicate that the 4-factor solution best represents organic aerosols of the present study.
2.2.2. Determination of the Size Distributions of the OA Factors
 The size distributions of individual OA factors were determined using a multivariate linear regression algorithm to decompose the mass spectra of OA corresponding to individual size-bins as the linear combination of the unit mass resolution (UMR) mass spectra of the four factors determined from PMF analysis of the HRMS, as shown in the following equation:
where mst,i denotes the measured UMR mass spectrum of OA corresponding to time period t and size bin i, msp denotes the UMR mass spectrum of factor p from PMF analysis of the HRMS, and cp,t,idenotes the corresponding fitting parameter. This method assumes that the spectral profile of each factor in different size bins is invariant, consistent with the two-dimensional PMF model [Ulbrich et al., 2009]. The mass spectra range from 10 to 120 amu. The size-resolved OA mass spectra in the size range of 40–1200 nm were averaged into 23 size bins to improve S/N ratio. In this study, the average OA mass spectrum over the entire campaign in each size bin (shown inFigure S4) was linearly decomposed according to the contributions of the four individual OA factors. The key diagnostic plots of the fitting results are shown in Figure S5. The reconstructed mass agrees very well with the measured OA mass in each size bin (R2 = 0.9992, Figure S5a), and overall, the linear superposition of the mass-weighted size distributions of the four OA factors reproduces that of total OA well (Figures S5b–S5d). The size distribution of each factor is normalized to its corresponding mass concentration.
3. Results and Discussion
3.1. Bulk Characteristics and Temporal Variations of Particulate Organic Species
Figure 1shows an overview of the average compositions of NR-PM1 and OA during this campaign. The average concentration of organic species is 7.9 μg m−3and represents ∼67% of the NR-PM1 mass (Figure 1a). The organic mass is on average composed of 66% carbon, 24% oxygen, 9% hydrogen, 1% nitrogen, and a very small amount of sulfur (∼0.014%). The CxHy+ ion family accounts for over half of the OA mass spectral signal (∼56%), larger than the contributions from all the oxygenated ions combined (∼41% in total), which include CxHyO1+ (28.2%), CxHyO2+ (10.1%), HyO1+ (1.5%), CxHyNpOz+ (0.8%), and CxHyOzSn+ (0.03%) (Figures 1a and 1b). The rest ∼3% signal is contributed by the CxHyNp+ ions. The average OA spectrum shows a base peak at m/z 43 with a composition of 62% C2H3O+, 36% C3H7+, and 2% C2H5N+ (Figure 1b). m/z 44 is the second largest peak, accounting for ∼6% of the total OA signal with major contributions from oxygenated ions of CO2+ (∼88%) and C2H4O+ (∼9%). m/z 57, which has been used as an AMS spectral tracer for HOA at urban locations [Zhang et al., 2005b], is dominated by C4H9+ (66%) but also contains 29% C3H5O+ and 4% C3H7N+. m/z 55 (∼5.7% of the OA signal) is another major peak in the spectrum and it is mainly contributed by C4H7+ (∼63%) and C3H3O+ (∼33%).
 The nominal chemical formula of OA in Fresno during this campaign is determined at C1H1.59O0.27N0.014S0.00008, which yields an average OM/OC of 1.50. Compared to the elemental ratios observed at other urban locations, the average O/C ratio of 0.27 in winter Fresno is in the lower end (cf. 0.41 in Mexico city [Aiken et al., 2008], 0.27 in London [Allan et al., 2010] and 0.36 in New York City (NYC) [Sun et al., 2011]), while the H/C ratio of 1.59 corresponds to the higher end (cf. 1.62, 1.62 and 1.49 in Mexico City, London and NYC respectively). The relatively low O/C and high H/C ratios suggest a high concentration of chemically reduced organic molecules in Fresno aerosol. The N/C ratio of 0.014 translates to an average organic nitrogen (ON) mass concentration of ∼0.09 μgN m−3 in Fresno. This value is higher than the 0.055 μgN m−3 in NYC (calculated from Sun et al. ) but lower than the 0.24 μgN m−3 observed in Mexico City [Aiken et al., 2008, 2009]. Given the abundance of reduced nitrogen ions, such as CH4N+, C2H5N+, C3H8N+, and C3H9N+, in the OA spectrum, amino functional group is likely a major contributor to the total ON in Fresno aerosol. Previous studies observed relatively high concentrations of amino compounds in atmospheric particles and water droplets in central California and suggested that it was because of regional agricultural activities [Zhang et al., 2002; Zhang and Anastasio, 2003; Sorooshian et al., 2008]. Amines were also found to represent an important fraction of the ON in NYC due to ocean-related emissions [Sun et al., 2011] and in Mexico City due to local industrial emissions [Aiken et al., 2010]. The scarcity of sulfur-containing organic ions present in the Fresno OA spectra indicates a very low loading of sulfur-containing organic compounds in particles. Nevertheless, the presence of methanesulfonic acid (MSA) or mesylate (CH3SO3−; the deprotonated anion of MSA) is confirmed based on the unambiguous detection of MSA signature ions – CH2SO2+, CH3SO2+, and CH4SO3+and the signal-ratios of these ions. The average MSA/mesylate concentration is estimated at 0.018μg m−3 during this study [Ge et al., 2012].
 The diurnal changes of the O/C and N/C ratios in OA generally follow a similar trend showing relatively higher values at nighttime (Figure 1c). On the contrary, H/C is elevated during daytime and peaks during morning rush hour, lunchtime, and evening rush hour/dinnertime, reflecting intensified anthropogenic emissions from e.g., traffic and food cooking (Section 3.2). The relatively lower O/C and higher H/C ratios during daytime (Figure 1c) suggest that elevated POA emissions outweigh the relatively slow daytime SOA production in Fresno during winter. SOA formation via aqueous-phase reactions may also have been enhanced by the frequent occurrence of nighttime fog events [Ge et al., 2012]. In contrast, a gradual increase of O/C ratio from morning to late afternoon was often seen in urban locations during summer campaigns, e.g., in NYC [Sun et al., 2011] and Mexico City [Aiken et al., 2008], because of intense photochemical SOA formation during daytime.
3.2. Sources and Processes of Particulate Organics
 Atmospheric OA is a combination of species from multiple sources and with different levels of oxidation. Although the bulk characteristics of OA yield some insights into the sources of particulate organics, a more powerful investigation can be achieved via PMF analysis of the organic high resolution spectra to resolve distinct OA factors that may be associated with unique sources and process types [Zhang et al., 2011]. The PMF analysis in this study yields four OA factors. Three of them are found to represent primary emissions: a hydrocarbon-like OA (HOA) factor driven by vehicle emissions, a cooking-influenced OA (COA), and a biomass-burning OA (BBOA) related to the residential wood combustion for heating. In addition, one oxygenated OA (OOA) factor is identified, representing secondarily formed organic species. An overview of the time series, HR spectra, and diurnal patterns of the individual OA factors is shown inFigure 2, along with the time series of tracer species for different sources. In the following sections, we discuss in detail the characteristics, sources, and processes of each factor.
3.2.1. Hydrocarbon-Like OA (HOA)
 During this study, the average (±1σ) mass concentration of HOA is 1.7 (±2.6) μg m−3 and the average contribution of HOA to OA mass is ∼22% (Figure 2m). The mass spectrum of HOA (Figure 2a) is dominated by the CnH2n±1+ ions (82.3% of the total signal; Figure 3a), among which the top 4 ions are C4H9+ at m/z 57, C4H7+ at m/z 55, C3H7+ at m/z 43, and C3H5+ at m/z 41. These ions are the major fragments arising from the 70 eV electron impact (EI) ionization of normal and branched alkanes and cycloalkanes [McLafferty and Turecek, 1993] – the dominant organic components of fuel and lubricating oil [Tobias et al., 2001; Canagaratna et al., 2004]. The abundance of these ions and the picket fence fragmentation pattern are common features in the AMS spectra of POA associated with fossil fuel combustion. The correlation between the Fresno HOA spectrum (after summed to unit mass resolution, UMR) and the standard HOA spectrum derived from 15 ambient HOA profiles [Ng et al., 2011] is very high (R2 = 0.95, Figure S6a). The HOA mass spectrum is also highly similar to the mass spectra of organic particles emitted in diesel [Canagaratna et al., 2004] (R2 = 0.94, Figure S6b) and gasoline vehicle exhausts [Mohr et al., 2009] (R2 = 0.93, Figure S6c).
 The O/C and OM/OC ratios of HOA are low (0.09 and 1.29) while the H/C ratio is high (1.80), indicating that HOA is mainly composed of chemically reduced hydrocarbon species. In the HRMS, HOA accounts for ∼32% of the CxHy+ signal and less than 10% of the oxygenated CxHyO1+ and CxHyO2+ signals (Figure 3b). The link between HOA and local traffic activities is further supported by its temporal variations: 1) the time series of HOA (Figure 2e) correlates well with those of fuel combustion tracers – CO (R2 = 0.81, Figure S7a), CO2 (R2 = 0.68, Figure S7b) and NO2 (R2 = 0.41, Figure S7c), and 2) the diurnal cycle of HOA concentration (Figure 2i) displays a peak corresponding to the morning rush hour (8:00–10:00) and another peak to the evening rush hour (17:00–20:00). The highest mass fractions of HOA in ambient aerosol occur at 9:00 (∼38% of OA) and 18:00 (∼27% of OA) (Figure 2n).
 As shown in Figure S8, ions that display the tightest correlations with HOA are the CnH2n+1+ ions, particularly C4H9+, C5H11+, C6H13+, C7H15+, and C8H17+, and the CnH2n−1+ ions containing 5 or more carbon atoms. Figure S9a further shows that HOA contributes over 60% of the signal in each one of these ions, corroborating their utilities as the AMS spectral markers for HOA. However, in viewing of integer m/z's, HOA accounts for less than 50% of the signal across the board (Figure S9b), even for m/z 57, which has been identified as an important AMS tracer for HOA in urban environment [Zhang et al., 2005b]. A reason is that this fragment also receives significant contributions from COA and BBOA in this study. Together, HOA, COA, BBOA contribute a total 85% of the signal at m/z 57 (Figure 4e), indicating a strong association of m/z 57 with total POA. In fact, based on comparing the contributions by POA versus those by OOA (Figure S9b), m/z 57 appears to be the most representative POA tracer for the UMR data.
 Some previous studies reported the size distributions of HOA and OOA based on the measured size distributions of m/z 57 and 44 [Zhang et al., 2005a; Nemitz et al., 2008]. This method seems to work at locations where traffic emissions are predominant sources of POA. However, in the presence of more than one POA sources (such as the case of this study), the measured size distribution of m/z 57 does not represent accurately the size distribution of HOA. We therefore used the multilinear deconvolution method discussed in Section 2.2.2 to estimate the size distributions of individual OA factors (Figures 4 and 5). The mass-based size distribution of HOA shows the typical behavior of traffic-related POA, peaking at around 140 nm (Figures 4a and 5a). In comparison, the distribution of m/z 57 shifts to a larger median due to substantial contributions from BBOA, COA, and OOA (Figures 4a and 4e). Overall, HOA dominates the mass of ultrafine particles (<100 nm) and its fractional contribution gradually decreases with increase of particle size (Figure 5b). It on average occupies less than 10% of the mass in particles larger than 500 nm in Dva.
3.2.2. Cooking-Influenced OA (COA)
 A factor related to cooking activities, COA, was identified. m/z's 55, 41 and 43 are the major peaks in the spectrum (Figure 2b), similar to the COA factors identified in London [Allan et al., 2010], Manchester [Allan et al., 2010], Beijing [Huang et al., 2010], New York City (NYC) [Sun et al., 2011], and Barcelona [Mohr et al., 2012]. As a POA factor, COA has a major contribution from CxHy+ ions, but to a lesser extent (∼72%) compared to HOA (Figure 3a). However, COA contains significantly larger amounts of oxygen-containing ions than HOA does (e.g., CxHyO1+- 19.3%vs. 10.2% and CxHyO2+- 4.7%vs. 3.9%) (Figure 3a) and is thus overall more oxygenated with a higher O/C ratio of 0.11 and a lower H/C ratio of 1.72. The oxygenated ions in the COA spectrum are likely produced from the ionization and fragmentation of oxygenated species such as fatty acids which were detected at significant amount in cooking aerosols [e.g., Zheng et al., 1997; To et al., 2000; He et al., 2004; Mohr et al., 2009]. The O/C ratio of COA in this study is lower than that identified in NYC (0.18) [Sun et al., 2011] and Barcelona (0.21) [Mohr et al., 2012], but similar to that observed in Beijing (0.11) [Huang et al., 2010]. Nevertheless, the overall similarities in the mass spectra between the Fresno COA and the COA factors determined previously are high (R2 = 0.79, 0.90, 0.86, 0.79 and 0.96 for the correlations with the COA factors observed in NYC, London, Manchester, Barcelona and Beijing; Figures S6d–S6i). The Fresno COA spectrum (in UMR) also shows high similarity (R2 of ∼0.9) to the reference aerosol spectra for food cooking reported by Mohr et al.  (Figures S6j–S6m). However, the COA spectrum does not resemble the profile of charbroiling reported by Lanz et al.  (R2 = 0.06, Figure S6n), implying that the Fresno COA is more related to heating of cooking oils rather than burning of meat/food itself, similar to the observation reported by Allan et al. .
 The COA diurnal cycle (Figure 2j) shows increases during noon and early evening coincidental with the lunchtime and dinnertime, corroborating its link with cooking emissions. COA co-varies closely with a few oxygenated ions (such as C3H3O+, C5H8O+, C6H10O+ and C7H12O+; Figures 2f and S8) and it contributes significantly to the signals of those ions (Figure S9a). The correlations between COA and C5H8O+ (R2 = 0.88, Figure S7d) and C6H10O+ (R2 = 0.92, Figure S7e) are especially high, supporting the recommendation by Sun et al.  that these two ions can be used as the high resolution mass spectral markers for ambient COA.
 Similar to previous observations [e.g., Mohr et al., 2012], the main spectral differences between COA and HOA include: 1) the COA spectrum has significant contributions from C3H3O+ and C3H5O+ at m/z's 55 and 57 and 2) the signal ratio of m/z 55 to 57 is much higher in COA than in HOA (Figure 2b). To distinguish COA from the other OA factors, Mohr et al.  proposed to use the relationships between the organic mass fractions of m/z's 55 and 57 (i.e., f55 and f57) or between those of C3H3O+ and C3H5O+ (i.e., fC3H3O+ and fC3H5O+) after subtracting contributions from oxygenated OA factors. We made this plot for Fresno OA in Figure 6a, in which the V shape formed by the two broken lines defines the edges of the COA and the HOA factors from several urban AMS data sets [Mohr et al., 2012]. Indeed, COA locates close to the left arm of the V shape, while HOA lies on the right arm, and the slope of f55 versus f57 also appears to be steeper in the afternoon than in the morning (Figure 6a), consistent with the diurnal patterns of COA and HOA (Figures 2i and 2j). Note that unlike in Barcelona [Mohr et al., 2012], in the present study, the fC3H3O+ to fC3H5O+ ratio is almost the same between COA and HOA but the fC3H3O+ in COA is almost 6 times higher than that in HOA (Figure S10a). The fC3H3O+vs.fC3H5O+ space is thus not useful for distinguishing the two factors for the Fresno data set.
 The mass-based size distribution of COA is the broadest among the four OA factors and shows two modes centering at ∼200 nm and 450 nm inDva (Figure 5a). This broad size distribution of COA is consistent with previous studies, which observed that the sizes of cooking-related particles vary widely, depending on cooking types, operations, and distance from the cooking sources [e.g.,Li et al., 1993; Dennekamp et al., 2001]. For example, Tu and Knutson  and Tu et al.  found that particles emitted from frying food were generally larger than the ones from gas combustion probably due to scavenging coagulation. Buonanno et al.  reported that the size distributions of particles emitted from different cooking methods (frying and grilling) and foods (fatty and vegetable foods) show a general trend of shifting to a smaller diameter at higher temperatures. The bimodal size distributions of COA observed in this study is similar to the observations by Kamens et al. , in which indoor particles during cooking periods sometimes presented a small size mode in addition to a dominant large size mode. Tu and Knutson  also observed two size modes during cooking soup but only one large size mode when frying food. The size modes of COA of this study (∼200 nm and 450 nm in Dva) are also consistent with previous observations. For example, meat cooking aerosols were found to peak at around 200 nm in mobility diameter [Hildemann et al., 1991] and the volume (or mass) median diameters of submicron aerosols from cooking were found to centered around 300–350 nm [Li et al., 1993; Mohr et al., 2012]. Overall, COA contributes a relatively constant fraction of particle mass in all sizes and it represents a much bigger fraction of mass in accumulation mode particles compared to HOA and BBOA (Figure 5b). As shown in Figure 4b, the size distribution of m/z 55 is quite different from that of COA, because ∼65% of the m/z55 signal is contributed by aerosols of non-cooking sources (Figure S9b) and both C4H7+ and C3H3O+ contribute significantly at m/z 55 (Figure 4f).
 On average, COA (average ± 1σ = 1.5 ± 2.3 μg m−3) accounts for ∼19% of the total OA mass (Figure 2m) and 33% of the POA mass (POA = HOA + COA + BBOA). The fractional contribution of COA is comparable to that of HOA in winter Fresno, similar to HR-AMS observations in NYC, where COA and HOA were found to contribute 16% and 14% of OA mass, respectively [Sun et al., 2011], and in Barcelona, where the values were 17% and 16%, respectively [Mohr et al., 2012]. Significant amounts of COA, which are comparable to or higher than the HOA loadings, were also reported in Beijing, China [Huang et al., 2010], and in two large cities in UK – Manchester and London [Allan et al., 2010]. In fact, prior to the AMS reports, source apportionment studies based on the molecular tracer technique repeatedly reported the identification of cooking aerosols in urban areas. For example, cooking emissions were estimated to account for 5–12% of the OA mass in some southeastern U.S. cities [Zheng et al., 2002], 21% of the primary fine organic carbon particulate mass in the Los Angeles area [Rogge et al., 1991], and 5–19% of PM2.5 mass during wintertime in Fresno [Chow et al., 2007]. All these results emphasize the importance of cooking emissions in urban areas. These aerosols thus deserve more attention in future air pollution control strategies, especially in densely populated area since all COA factors identified so far are from aerosols sampled from metropolises.
3.2.3. Biomass-Burning OA (BBOA)
 In this study, we identified that regional residential wood combustion for heating is another important POA source, in addition to vehicular and cooking emissions, in Fresno during wintertime. This factor, which is termed as biomass-burning OA (BBOA), includes considerable signals from oxygenated ions of CxHyO1+ (35.8%) and CxHyO2+ (14.7%) (Figure 3a); it thus consists of a higher concentration of oxidized species than HOA and COA do. The O/C of BBOA is the highest (0.33) and its H/C lowest (1.56) among three POA factors, similar to the findings by Aiken et al.  and Mohr et al. . The BBOA profile (Figure 2c) has strong signals at m/z 60 (100% of which is C2H4O2+) and m/z 73 (95% of which is C3H5O2+), which are the known fragment ions in the EI-MS of levoglucosan - a biomass-burning aerosol tracer [Alfarra et al., 2007; Aiken et al., 2008]. The BBOA spectrum of this study is also similar to the AMS BBOA factors observed at other locations and the correlation with the standard BBOA spectral profile derived from the average of two BBOA spectra obtained in Mexico city and Houston [Ng et al., 2011] is tight (R2 = 0.85, Figure S6o). Recently, Jimenez et al.  proposed a graphical method taking f60 as a primary marker for BBOA and f44 as a tracer for SOA or aged POA, to investigate the atmospheric evolution of BBOA. We demonstrated this graph in Figure 6b using fCO2+vs.fC2H4O2+ instead of f44 vs.f60. Note that for this data set, since ions other than CO2+ and C2H4O2+ contribute very little to signals at m/z 44 and 60 (Figures 4g and 4h), respectively, the f44 versus f60 plot (Figure S10b) is very similar to Figure 6b. The triangle shape is adopted from that reported by Jimenez et al. , which was determined from the measurement data influenced by fire plumes. The Fresno BBOA locates near to the right edge of the triangular region, separated well from the other OA factors which situate outside the triangle shape, owing to low fC2H4O2+ in them (Figures 2a–2d).
 The mass-based average size distribution of BBOA (Figure 5a) peaks around 160 nm, resembling more to the size distributions of HOA than to OOA, despite the larger difference in the oxidation degrees between BBOA and HOA (0.33 versus 0.09) than between BBOA and OOA (0.33 versus 0.42). Similar to HOA, BBOA contributes a large fraction of the small particle mass (Figure 5b). Since m/z 60 is only composed of C2H4O2+ and m/z 73 is dominated by C3H5O2+ (95%), their size distributions are thus highly similar to that of BBOA (Figures 4c and 4g).
 The average (±1σ) BBOA mass concentration is ∼1.3 (±2.2) μg m−3 (16% of OA mass), slightly lower than those of COA and HOA. BBOA presents a distinct diurnal pattern that shows substantially elevated concentrations in the evening from 19:00–24:00 (Figure 2k), indicating its direct link to the residential wood burning activities. The time series of BBOA correlates very well with that of C2H4O2+ (R2 = 0.95, Figure S7f) and C3H5O2+ (R2 = 0.91, Figure S7g), which are the AMS spectral markers for BBOA. It also has a moderate correlation with K+ (R2 = 0.38, Figure S7h), another important biomass-burning marker [Chow et al., 2007]. BBOA has the highest N/C ratio (0.021) among the four OA components (Figures 2a–2d), and it shows particularly good correlations with CN+ (R2 = 0.46, Figure S7i) and CHN+ (R2 = 0.58, Figure S7j), consistent with enhanced nitrile emissions during burning of biomasses [Simoneit et al., 2003].
 Additionally, BBOA should also contain various phenolic compounds which are important pyrolysis products of lignin in woods [e.g., Simoneit et al., 1993; Shakya et al., 2011]. Since lignin is a complex compound, unlike the low molecular weight methoxyphenols, polyphenols may also be emitted and present in aerosols from wood-burning [Gonçalves et al., 2011]. Specially, in this study, the time series of two fragment ions representative of polyphenols, C14H14O4+ (m/z 246) for guaiacol dimer and C16H18O6+ for syringol dimer (m/z 306) [Sun et al., 2010], correlate well with BBOA (R2 = 0.82 and 0.78 for m/z 246 and 306; Figure 7), indicating their utilities as tracers for biomass burning aerosols. However, these two ions were also observed in the AMS HRMS of the SOA products of phenolic compounds formed via photochemical aqueous-phase reactions [Sun et al., 2010], suggesting that the BBOA observed in this study may contain secondary organic species as well. In a single-particle mass spectrometry study conducted in Fresno during wintertime,Qin and Prather also detected biomass-burning particles of which the diurnal pattern was very similar to that of the BBOA observed in this study. They hypothesized that the diurnal profile was due to direct biomass emissions followed by gas-to-particle partitioning of semi-volatile species and aqueous-phase processing. Similar processes likely occurred during this campaign given the wet and cold atmospheric conditions throughout the campaign [Ge et al., 2012]. As a result, the BBOA factor of this study may contain secondary species, but likely only in a minor amount since no enhancement of BBOA concentration was observed during nighttime fog events, during which all secondary species including OOA, nitrate, and sulfate increased significantly [Ge et al., 2012]. In addition, the diurnal pattern of the BBOA factor also suggests that it is mainly governed by direct wood burning emissions, i.e., mainly of a primary origin. Overall, the average nighttime BBOA concentration (e.g., ∼2.9 μg m−3 between 19: 00 – 24:00) is more than 7 times higher than the average daytime BBOA concentration (e.g., ∼0.4 μg m−3 between 11:00 – 16:00).
3.2.4. Oxygenated OA (OOA)
 In addition to three POA factors, a secondary, oxygenated OA factor was identified and found to account for an average 43% of the OA mass (Figure 2m). OOA is a ubiquitous and often dominant aerosol component in the atmosphere [Zhang et al., 2007; Jimenez et al., 2009]. In many cases, OOA can be further separated into a low-volatility/more oxygenated OOA (LV-OOA/MO-OOA) and a semi-volatile/less oxygenated OOA (SV-OOA/LO-OOA), which represent different aging and oxidization degrees [Jimenez et al., 2009; Ng et al., 2010, and references therein]. However, in this study only one OOA factor was identified. The most abundant peaks in the OOA mass spectrum (Figure 2d) are CO2+ at m/z 44 and C2H3O+ at m/z43. It is characterized with an overall dominance of oxygen-containing ions (51.5% in total: 37.9% is CxHyO1+ and 13.6% is CxHyO2+) though CxHy+ signals still occupy the largest portion (42.1%) among the six ion categories (Figure 3a). The Fresno OOA spectrum (in UMR) shows overall good agreements with the standard spectra of OOA, LV-OOA and SV-OOA derived from 15 urban data sets [Ng et al., 2011] (R2 = 0.81 – 0.86; Figures S6p–S6r). Ng et al.  discovered that all OOA components fell into a triangular region defined by f44 versus f43 and that the positions within the triangle corresponds to their oxidation degrees. Since the HRMS allows us to exclude interferences from other ions (such as C3H7+ at m/z 43), we examined fCO2+versus fC2H3O+ instead (Figure 6c). OOA is close to the bottom of the triangular region, approximately at the same position as the SV-OOA identified from NYC does [Sun et al., 2011] and in the region within which the SV-OOA factors usually falls [Ng et al., 2010]. The f44 (∼9%) and the O/C ratio (0.42) of the Fresno OOA are in the range of the SV-OOA factors observed worldwide too (e.g., 7 ± 4% forf44 and 0.35 ± 0.14 for O/C ratio) [Jimenez et al., 2009]. These results imply that the OOA factor identified in winter Fresno is more similar to SV-OOA than to LV-OOA. On the other hand, HOA and COA are both outside of the triangular region, denoting their different chemical characteristics; BBOA lies the closest to OOA in the triangle, consistent with its higher content of oxygenated species. Compared to thef44 versus f43 space (Figure S10c), the fCO2+versus fC2H3O+ triangle apparently improves the separation of OOA from other primary OA factors.
 This semi-volatile characteristic of the Fresno OOA is further verified by its higher correlation with nitrate - a semi-volatile secondary species (R2 = 0.86, Figure S7k), than with sulfate - a low-volatility secondary species (R2 = 0.71, Figure S7l). Actually, the correlation coefficients between OOA and any secondary inorganic species (i.e., nitrate, sulfate, chloride, and ammonium) are higher than those between the OOA and the POA factors (i.e., HOA, COA, and BBOA; Table 1). OOA correlates especially well with total secondary inorganic anions (= sulfate + nitrate + chloride) (R2 = 0.87, Figure 8a) while the correlation between POA and secondary inorganic anions is poor (R2 = 0.13, Figure 8b). In addition, OOA does not correlate with combustion tracer gases such as CO (R2 = 0.10, Figure S7m), CO2 (R2 = 0.19, Figure S7n) and NO2 (R2 = 0.07, Figure S7o), nor with POA (R2 = 0.12, Figure 8c). These results are consistent with the association of OOA to secondary origins.
Table 1. Pearson's R Coefficients for Linear Regressions Between OA Factors and Gas Pollutants and Aerosol Species Measured at Fresno in Winter 2010a
Values that indicate a strong correlation (i.e., R > 0.5) are in bold.
POA = HOA + COA + BBOA.
Ammonium + sulfate + nitrate + chloride
Methanesulfonic acid (MSA) or mesylate
 The diurnal cycle of OOA (Figure 2k) is relatively flat without distinct peaks. The slightly higher concentrations of OOA during nighttime are likely due to: 1) enhanced gas-to-particle partitioning favored by lower temperature and higher RH at night, and 2) aqueous-phase production of secondary aerosol species in the presence of nighttime fogs [Ge et al., 2012]. Particularly, CH2SO2+ and CH3SO2+, which are the characteristic AMS spectral ions for methanesulfonic acid (MSA) [Ge et al., 2012] – a well-known secondary species, correlate tightly with OOA (R2 = 0.60 and 0.71, respectively, Figure S8). In addition, these two ions are almost contributed by OOA alone (Figure S9a. These results further support the hypothesis that the OOA factor represents SOA. The average mass-based size distribution of OOA is similar to those of secondary inorganic species [Ge et al., 2012], peaking at an accumulation mode size of ∼460 nm in Dva (Figures 4d and 5a). The mass fraction of OOA gradually increases with the increase of particle size and oxygenated species dominates the mass of particles larger than 300 nm in Dva (Figure 5b). The size distribution of m/z 44 is broader than that of OOA and extends deeper into the ultrafine size range (Figure 4d), because of significant contributions from POA (Figure 4h).
3.2.5. Relative Importance of Primary and Secondary OA in Winter Fresno
Figure 8c demonstrates that the high PM loading periods are generally associated with high POA concentrations and high mass fractions of organic aerosols, indicating the important roles that primary aerosol emissions play in aerosol pollutions in Fresno during wintertime. On average, POA dominates the OA mass (∼57%) with HOA, COA and BBOA being equally important (Figure 2m). The diurnal patterns of these three types of POA, however, are substantially different (Figures 2i–2k), indicating that the average POA composition varies considerably during the day. Primary emissions, in particular, play a leading role in aerosol pollution during nighttime, contributing an average 80% of the total OA at around 20:00 (Figure 2n), owing to elevated emission from evening traffic, dinner cooking and residential wood burning compounded with the collapsing of mixed layer height. Enhancements of POA are observed during morning rush hour and around noontime too, reflecting impacts from vehicle emissions and lunch cooking, respectively. A survey of some prior studies about PM (mostly PM2.5) sources in Fresno during wintertime is listed in Table 2. Generally, our findings about PM1 are consistent with those studies, highlighting the importance of wood burning, vehicle emission, and cooking as sources of aerosol pollution in Fresno.
Table 2. Prior Studies About the Sources of Fine PM in Fresno During Wintertime
CMB model using PM2.5 collected in Fresno during two severe wintertime (1995) air pollution episodes
Hardwood and softwood combustion (31 % of PM2.5), diesel (6 % of PM2.5) and gasoline-motor vehicles (8 % of PM2.5), and meat cooking (5 % of PM2.5) were found. Contributions from road dust and natural gas combustion were small but measurable.
CMB model using PM2.5 measurements in Fresno during 5 Dec. 2000–3 Feb. 2001
Residential wood combustion (31 % of PM2.5: 16–17 % of hardwood and 12–15 % of softwood); Motor vehicle (9–15 % of PM2.5: 3–10 % from gasoline and 5–6 % from diesel), and cooking (5–19 % of PM2.5) were found.
 In addition, polycyclic aromatic hydrocarbons (PAHs), are known byproducts of incomplete combustion from sources like diesel and gasoline engines, biomass burning and wood smoke, etc., and many of them are potent mutagens and carcinogens [Marr et al., 2006]. The concentrations of particle-bound PAHs in this study (Figure 9a) were estimated based on the fragmentation table reported by Dzepina et al. . The average PAHs concentration is determined at 16 ng m−3. The time series of PAHs shows very good correlations with POA (R2 = 0.72, Figure 9b) (Table 1). In addition, PAHs shows a better correlation with BBOA (R2 = 0.70, Figure 9e) than with HOA (R2 = 0.44, Figure 9c) and COA (R2 = 0.41, Figure 9d) (Table 1), indicating that residential wood combustion is a significant source of PAHs in Fresno.
 OOA/SOA on average accounts for 43% of the OA mass during this study (Figure 2m). Similarly, OOA was found to account for less than half of the OA mass at three other urban locations (i.e., NYC, Tokyo, and Manchester) during winter months [Zhang et al., 2007]. In contrast, a dominant contribution from OOA was usually observed at urban locations during warm seasons [Zhang et al., 2011]. The slower secondary aerosol formation kinetics and the stronger accumulation of primary pollutants due to lower mixed layer depth during winter all contribute to a large mass fraction of POA in aerosol particles. During this study, OOA only outweighs POA at early morning (00:00–06:00) when all the anthropogenic primary emissions are largely reduced (Figure 2n).
4. Conclusions and Implications
 An HR-ToF-AMS was deployed in Fresno in winter 2010 to characterize submicron aerosols. The NR-PM1 mass on average is dominated by organics (∼67%) which has an average nominal chemical formula of C1H1.59N0.014O0.27S0.00008 and an average OM/OC ratio of 1.50. The average concentration of OA is 7.9 μg m−3, varied between 0.06 and 161 μg m−3. PMF analysis of the high resolution mass spectra revealed four sources of particulate organics in Fresno: 1) vehicle emissions that generate HOA; 2) food cooking that emits COA; 3) residential wood combustion that releases BBOA; and 4) chemical reactions that lead to the formation of secondary, oxygenated species (OOA). The diurnal cycles of the three POA components, i.e., HOA, COA and BBOA, show distinct patterns that are consistent with the temporal emission profiles of local sources. For example, HOA peaks during morning and evening rush hours; COA is enhanced during lunch- and dinner-time; and BBOA concentration is particularly high at night associated with enhanced residential wood burning for heating. OOA exhibits semi-volatile behavior, with relatively low O/C (0.42) and high H/C ratios (1.43) and good correlations with ammonium nitrate and ammonium chloride (R2 = 0.87 and 0.56, respectively). The correlation between OOA and ammonium sulfate is also quite good (R2= 0.71) during this study, likely arising from aqueous-phase reactions which strongly influence the formation of both species. The diurnal cycle of OOA concentration is relatively flat, with a slightly higher loading at night that may be attributed to enhanced gas-to-particle partitioning and aqueous-phase SOA production favored by the cool and moist meteorological conditions and the occurrence of nighttime fog events.
 Traffic, cooking, and residential wood combustion are important aerosol sources in Fresno in winter, on average contributing 14.5%, 12.8%, and 10.9%, respectively, of the PM1 mass. Primary aerosols from these sources are also found responsible for the frequent occurrences of high PM pollution episodes. Overall, POA is more important than SOA and constitutes an average 57% of the total OA mass observed during this study. The contributions of POA to PM pollution are especially high at night, with HOA, COA and BBOA together account for up to 80% of the OA mass between 18:00–22:00. The nighttime POA concentration was on average 5–6 times higher than its daytime concentration. The leading role of primary emissions, all of which are anthropogenic, in aerosol pollution observed in this study highlights the influences of human activities on regional air quality in central California. The severity of primary aerosol pollutions at urban locations during wintertime in this region also arises from factors including: 1) the winter meteorological conditions such as low mixed layer depth and low wind speeds which favor the accumulation of air pollutants, and 2) the low ambient temperature and weak solar radiation in wintertime that may further limit the volatilization and photochemical oxidation of POA.
 Furthermore, the mass-based size distributions of HOA and BBOA present the typical POA behavior, peaking at a small size mode (∼140 nm inDva), while that of COA is broad and contributes significantly to both the ultrafine and the accumulation modes. The mass-based size distribution of OOA peaks at an accumulation mode of ∼460 nm inDva, representing typical behavior of secondary species. Overall, POA contributes over 95% of mass in particles smaller than 100 nm (Dva) and likely represents a vast majority of ambient ultrafine particles in terms of number as well. Smaller particles can penetrate deeper into lung tissue and cause more damages [Nel, 2005]. In addition, since primary organic particles tend to have a high content of toxic species such as PAHs and are co-emitted with black carbon, they are probably the most damaging particulate component on human health. A tight correlation between the PAH and the POAs concentration was observed during this study, especially between BBOA and PAH. The fact that the PAH correlates significantly better with BBOA (R2 = 0.7) than with HOA (R2 = 0.44) and COA (R2 = 0.41) suggests that residential wood combustion may be an important health concern in the region. Our observations point out that in effort to mitigate the air pollution problem in the SJV of California in winter, more attentions and efforts should be laid upon controlling emissions of primary particles from local human activities.
 This research was supported by the Office of Science (BER), U. S. Department of Energy, grant DEFG02-08ER64627, DESC0002191, the California Agricultural Experiment Station (project CA-D-ETX-2102-H), and the UC Davis Atmospheric Aerosol Health (AAH) program. Support for the field experiment was provided by the San Joaquin Valley Aerosol Health Effects Research Center (SAHERC) funded by the U.S. EPA through grant RD-83241401-0 to UC Davis. However, the research described in the article has not been subjected to the EPA's required peer and policy review and therefore does not necessarily reflect the views of the Agency and no official endorsement should be inferred.