Ultrafine Particulate Matter Source Contributions across the Continental United States

Abstract. The regional concentration of airborne ultrafine particulate matter mass (Dp < 0.1 µm; PM0.1) was predicted with 4 km resolution in 39 cities across the United States during summer time air pollution episodes. Calculations were performed using a regional chemical transport model with 4 km spatial resolution operating on the National Emissions Inventory created by the US EPA. Measured source profiles for particle size and composition between 0.01–10 µm were used to translate PM total mass to PM0.1. PM0.1 concentrations exceeded 2 µg m-3 during summer pollution episodes in major urban regions across the US including Los Angeles, the San Francisco Bay Area, Houston, Miami, and New York. PM0.1 spatial gradients were sharper than PM2.5 spatial gradients due to the dominance of primary aerosol in PM0.1. Artificial source tags were used to track contributions to primary PM0.1 and PM2.5 from fifteen source categories. As expected, on-road gasoline and diesel vehicles made significant contributions to regional PM0.1 in all 39 cities even though peak contributions within 0.3 km of the roadway were not resolved by the 4 km grid cells. Food cooking also made significant contributions to PM0.1 in all cities but biomass combustion was only important in locations impacted by summer wildfires. Aviation was a significant source of PM0.1 in cities that had airports within their urban footprints. Industrial sources including cement manufacturing, process heating, steel foundries, and paper & pulp processing impacted their immediate vicinity but did not significantly contribute to PM0.1 concentrations in any of the target 39 cities. Natural gas combustion made significant contributions to PM0.1 concentrations due to the widespread use of this fuel for electricity generation, industrial applications, residential, and commercial use. The major sources of primary PM0.1 and PM2.5 were notably different in many cities. Future epidemiological studies may be able to differentiate PM0.1 and PM2.5 health effects by contrasting cities with different ratios of PM0.1 / PM2.5. In the current study, cities with higher PM0.1 / PM2.5 ratios include Houston TX, Los Angeles CA, Birmingham AL, Charlotte NC, and Bakersfield CA. Cities with lower PM0.1 to PM2.5 ratios include Lake Charles LA, Baton Rouge LA, St. Louis MO, Baltimore MD, and Washington DC.



Introduction
Airborne particulate matter (PM) has been linked with premature mortality and numerous other health risks in cities across the world (see for example references (Laden, Neas et al. 2000, Pope, Burnett et al. 2002, Dominici, Peng et al. 2006, Ostro, Broadwin et al. 2006, Franklin, Zeka et al. 2007, Pope, Ezzati et al. 2009, Kheirbek, Wheeler et al. 2013, Aneja, Pillai et al. 45 2017)).Despite years of progress (EPA 2017), PM concentrations in many urban regions in the United States still exceed health-based standards resulting in an increase of non-accidental mortality (Franklin, Zeka et al. 2007, Baxter, Duvall et al. 2013).Toxicology testing suggests that ultrafine particles with diameter < 0.1 µm may be the most harmful size fraction within PM2.5 (Oberdorseter, Gelein et al. 1995, Pekkanen, Timonen et al. 1997, Oberdurster 2000, Li, 50 Siotas et al. 2003, Ostro, Hu et al. 2015).Initial attempts to analyze ultrafine particles in epidemiology studies have used particle number concentration as a surrogate for ultrafine particle exposure, but this approach has not found consistent relationships with health effects (Ostro, Hu et al. 2015).In contrast, a recent epidemiology study based on ultrafine particle mass (PM0.1)found significant associations with premature mortality (Ostro, Hu et al. 2015).Follow-55 up studies have also found significant associations between PM0.1 and reproductive outcomes including birth weight and preterm birth (Bergin, Russell et al. 1996, Laurent, Hu et al. 2016).
These findings have biological plausibility, since ultrafine particles may cross cell membranes and interfere with the internal cell function (Sioutas, Delfino et al. 2005).The toxic material found in ultrafine particles has greater surface area due to the small particle diameter making the 60 material more available for chemical reaction.Ultrafine particles can therefore have a larger impact when deposited deep into the lung cavity where they are not easily removed (Li, Siotas et al. 2003, Nel, Xia et al. 2006).
A national monitoring network for PM10 and PM2.5 has been operating throughout the continental US for almost 20 years.Multiple studies have performed source apportionment 65 calculations for coarse and fine PM using these measurements (Zheng, Cass et al. 2002, Reff, Bhave et al. 2009, Ham and Kleeman 2011, Zhang, Hu et al. 2014).In contrast, measurements of PM0.1 are limited to focused field campaigns lasting for short time periods with even fewer studies attempting source apportionment calculations (Kleeman, Riddle et al. 2009).Multiple barriers have prevented the widespread deployment of PM0.1 monitoring networks including (i) models have been used to track source contributions to primary organic matter, elemental carbon and in some cases particle number concentration (PNC) over areas in the Eastern U.S. and parts 80 of Europe and Asia (Gaydos, Stanier et al. 2005, Lane, Pinder et al. 2007, Wang, Hopke et al. 2011, Posner and Pandis 2015, Cattani, Gaeta et al. 2017, Wolf, Cyrys et al. 2017, Simon, Patton et al. 2018, Zhong, Nikolova et al. 2018).However, these methods are limited in one or more aspects of their ability to predict population exposure ultrafine particles over large analysis domains.Source resolved models, such as PMCAMx, have been demonstrated for PNC but not 85 for PM0.1 (Posner and Pandis 2015).CMB models need measurements of specific molecular markers at numerous sites to resolve the sharp spatial gradients of ultrafine particle source contributions.LUR models need comprehensive measurements that act as training data sets in order to extend throughout a modeling domain (Lane, Pinder et al. 2007).
Hu et al. (Hu, Zhang et al. 2014) calculated population exposure to PM0.1 in California 90 using a regional source-oriented chemical transport model supported by measured profiles for particle size and composition emitted by dominant sources.Predictions were compared to all available fine and ultrafine particle measurements over the period 2000-2010 with good agreement observed for the dominant chemical components of PM0.1 mass including organic aerosol, elemental carbon, and numerous trace metals (Hu, Zhang et al. 2014).The 4km spatial 95 resolution used in these calculations supported multiple epidemiological studies based on spatial gradients of exposure (Ostro, Hu et al. 2015, Laurent, Hu et al. 2016).These encouraging results motivate the expansion of the PM0.1 exposure technique to other locations.
Here we use the Eularian source-oriented UCD/CIT chemical transport model to predict the concentration of PM0.1 in thirty-nine urban regions throughout the US during summer 100 pollution events in 2010.The calculation tracks contributions from fifteen (15) primary particle sources through a simulation of all major atmospheric processes while retaining information about particle size, composition and source origin (Hu, Zhang et al. 2014).The results of this calculation reveal US national trends in PM0.1 concentrations for the first time and suggest locations where the differential health effects of PM0.1 and PM2.5 can best be studied.

Methods 2.1 Simulation Dates
Simulations within each target city were carried out during peak summer air pollution events in 2010.Peak air pollution events typically had measured 1-hr maximum ozone (O3) concentrations greater than 70 ppb.Regional pollution events caused by atmospheric stagnation were selected whenever possible as opposed to special events caused by unusual occurrences such as wildfires.Measured PM2.5 24-hr concentrations during peak summer pollution events ranged between 3.2-30 µg/m 3 depending on the location.The simulation dates in each city are listed in Table 1 and a map of the city locations is shown in the supplemental information Figure S1.The aggregation of these events across the US enables a comparison of typical air pollution episodes within different cities.

Model Description
The UCD/CIT model predicts the evolution of gas and particle phase pollutants in the atmosphere in the presence of emissions, transport, deposition, chemical reaction and phase change (Held, Ying et al. 2005)  where Ci is the concentration of gas or particle phase species i at a particular location as a function of time t, u is the wind vector, K is the turbulent eddy diffusivity, Ei is the emissions 130 rate, Si is the loss rate, Ri gas is the change in concentration due to gas-phase reactions, Ri part is the change in concentration due to particle-phase reactions and Ri phase is the change in concentration due to phase change (Held, Ying et al. 2005).Loss rates include both dry and wet deposition.
Phase change for inorganic species occurs using a kinetic treatment for gas-particle conversion (Hu, Zhang et al. 2008) driven towards the point of thermodynamic equilibrium (Nenes, Pilinis 135 et al. 1998).Phase change for organic species is also treated as a kinetic process with vapor pressures of semi-volatile organics calculated using the 2-product model (Carlton, Bhave et al. 2010).More sophisticated approaches for secondary organic aerosol (SOA) formation (Cappa, Jathar et al. 2016) were also tested in the current study but these required a larger number of assumptions and they did not produce higher SOA concentrations in the PM0.1 size fraction.

140
Nucleation was not included in the current study and so particle number concentrations will not be discussed.Likewise, model spatial resolution was 4km over the 4.2 million km 2 of simulated urban areas and so near-roadway concentrations of ultrafine particles on spatial scales of ~0.1 km will not be presented.
A total of 50 particle-phase chemical species are included in each of 15 discrete particle 145 size bins that range from 0.01-10 µm particle diameter (Held, Ying et al. 2005).Artificial source tags are used to quantify source contributions to the primary particle mass for a specific bin size, therefore allowing for the direct contribution of each source of PM2.5 and PM0.1 mass to be determined.Gas-phase concentrations of oxides of nitrogen (NOx), volatile organic compounds (VOCs), oxidants, ozone, and semi-volatile reaction products were predicted using the SAPRC-150 11 chemical mechanism (Carter and Heo 2013).

Model Inputs
Anthropogenic emissions were generated using the Sparse Matrix Operator Kernel Emissions (SMOKEv3.7)modeling system applied to the 2011 National Emissions Inventory. 155 Emissions from each of the four major source sectors (area, mobile, non-road and point were tagged to create fifteen (15) different emissions groups: on road diesel, on road gasoline, off road diesel, off road gasoline, biomass, food cooking, natural gas, process heaters, distillate oil, aviation, cement, coal, steel foundries, paper products and all other emissions.Size and composition-resolved source profiles were then assigned to the PM emissions within each of 160 these groups using the UCD/CIT emissions processor based on the most recent measurements available in the literature (Robert, VanBergen et al. 2007a, Robert, Kleeman et al. 2007b, Kleeman, Robert et al. 2008).Some of the fifteen (15) source categories were represented using weighted average source profiles from multiple sources as described in Table S1.
Daily values for 2010 wildfire emissions were generated using the Global Fire Emissions 165 Database (GFED) (Giglio, Randerson et al. 2013).Biogenic emission rates were generated using the Model of Emissions of Gases and Aerosols from Nature (MEGANv2.1).The gridded georeferenced emission factors and land cover variables required for MEGAN calculations were created using the MEGANv2.1 pre-processor tool and the ESRI_GRID leaf area index and plant functional type files available at the Community Data Portal (Guenther, Jiang et al. 2012). 170 Meteorology parameters used to drive the UCD/CIT CTM were generated using the Weather Research and Forecasting model (WRFv3.6)and WRF preprocessing system (WPSv3.6).Meteorological fields were created within 3 nested domains with horizontal resolutions of 36km, 12km, and 4km, respectively.Each domain had 31 telescoping vertical levels up to a top height of 12km.Four-dimensional data assimilation (FDDA) or "FDDA 175 nudging" was used to anchor meteorological predictions to measured values (Hu et al., 2010).
Meteorological data and gridded map projections needed for 2010 emissions modeling were taken from the corresponding WRF simulations using the meteorology-chemistry interface processor (MCIP). 180

Supporting Measurements
Ambient hourly ozone measurements and daily PM2.5 measurements were obtained from the EPA AQS API / Query AirData (EPA).Model predictions are compared to these measurements to build confidence in the accuracy of the overall modeling system since PM0.1 measurements are not available during any of the peak summer pollution events studied here. 185

Results
Predicted 1-hr ozone concentrations were compared to measurements averaged within each city to indirectly evaluate the accuracy of the emissions inventories and meteorology fields.
Many of the sources that emit ozone precursors also emit ultrafine particles.Likewise,  S2 in the supplemental information).

200
Predicted 24-hr PM2.5 concentrations were compared to measurements as a second check on the accuracy of model features needed to predict ultrafine particle concentrations.Many of the combustion sources that emit primary particles within the PM2.5 size fraction also emit PM0.1.
The Chemical Speciation Monitoring Network (CSN) operated by the U.S. Environmental Protection Agency (EPA) measures PM2.5 mass and chemical composition at more than 260 sites 205 throughout the U.S. including many of the 39 cities studied in the current analysis (Solomon, Crumpler et al. 2014).Elemental carbon (EC) and organic compounds (OC) are the chemical components most relevant for both the PM2.5 and the PM0.1 size fractions.Figure 2 illustrates predicted vs measured 24-hr PM2.5 EC and OC concentrations for all 39 cities while Figure S3 illustrates predicted vs. measured 24-hr PM2.5 total mass comparisons.In general, the model 210 slightly under predicts PM2.5 EC, OC, and mass with regression slopes ranging from 0.62 for EC to 0.97 for OC.The negative bias in model predictions may stem from the 4km spatial averaging inherent in the calculations vs. the influence of sources closer than 4 km to the measurement site in the urban environment such as highways, restaurants, etc.This trend is reflected in the performance of ozone predictions during the evening hours for Los Angeles and New York City

215
(Figure 1), where measured ozone concentrations fall to zero due to titration from nearby NOx emissions while predicted ozone concentrations remain greater than zero due to dilution of NOx emissions within 4 km grid cells.The MFB and MFE for PM2.5 predictions are summarized in the supplemental information Table S2.
As was the case for ozone predictions, PM2.5 model performance meets EPA criteria 220 (MFE<0.75) in 37 out of 39 cities, building confidence in the accuracy of the model results for predictions from these two independent techniques for PM0.1 concentrations associated with gasoline engines, diesel engines, food cooking, wood burning, and "other sources".Further details of this comparison are provided by (Yu, Venecek et al. 2018).This evaluation of the modeling procedures builds confidence in the PM0.1 source predictions across the US in the 235 current study.
Figure 3 illustrates a composite representation of PM2.5 and PM0.1 mass across the US during the summer pollution episodes listed in Table 1.The spatial plot in Figure 3 is constructed using the intermediate 12km simulation results from multiple simulations stitched together to cover a broader geographical area.Regional PM0.1 concentrations reach a maximum 240 value of 5 µg m -3 in a few isolated grid cells with wildfires but concentrations generally exceed 2 µg m -3 in major urban regions across the US including Los Angeles, the San Francisco Bay Area, Houston, Miami, and New York.The comparison between PM2.5 mass (Figure 3a) and PM0.1 mass (Figure 3b) shows that PM0.1 spatial gradients are sharper with less regional contributions between "hot spots".Locations in the Midwestern and Eastern US outside of cities with high simulation results during the pollution events listed in Table 1.Source contribution spatial plots for the entire US are shown in the supplemental information Figures S4-S7 and pie charts for PM2.5 source contributions in all 39 US cities are shown in the supplemental information Figure S8.As expected, on-road gasoline and diesel vehicles made significant contributions to regional PM0.1 in all 39 cities even though peak contributions within 0.3 km of the roadway were not 260 resolved by the 4 km grid cells.Food cooking also made significant contributions to PM0.1 in all cities but biomass combustion was only important in locations impacted by summer wildfires.
Aviation was a significant source of PM0.1 in cities that had airports within their urban footprints.
Industrial sources including cement manufacturing, process heating, steel foundries, and paper & pulp processing impacted their immediate vicinity but did not significantly contribute to PM0.1 265 concentrations in any of the target 39 cities. Natural gas combustion made significant contributions to PM0.1 concentrations due to the widespread use of this fuel for residential, commercial, and industrial applications.Natural gas contributions were especially significant in locations with high levels of industrial use such as chemical refineries or in locations with significant levels of natural gas fired power plants.

270
The major sources of primary PM0.1 and PM2.5 were notably different in many cities (compare Figure 3a and 3b).The sources that contribute most strongly to PM2.5 are on road diesel, gasoline, food cooking, coal and "other" which includes break and tire wear from mobile sources and dust.Natural gas combustion makes minor contributions to primary PM2.5 mass since particles from this source have a mass distribution peaking at ~0.05 µm particle diameter 275 (Chang, Chow et al. 2004) with all of the emitted mass in the PM0.1 size fraction.In contrast, other combustion sources using more complex fuels such as on-road vehicles have a mass distribution peaking at ~0.1 µm with at least half the emitted mass outside the PM0.1 size fraction (Robert, VanBergen et al. 2007a, Robert, Kleeman et al. 2007b).Likewise, food cooking contributes strongly to PM2.5 concentrations but the emitted particle mass distribution peaks at 280 0.2 µm with the majority of the mass outside the PM0.1 size fraction.39 study cities shown in Table 1.This calculation highlights the importance of natural gas combustion particles in the PM0.1 size fraction and the minor role that these natural gas combustion particles play in the PM2.5 size fraction.Natural gas typically consists of +93% methane with the balance of the fuel made up by higher molecular weight alkanes and trace 315 impurities.In addition to background sulfur compounds in the natural gas, sulfur-containing odorants such as mercaptans are commonly added to aid in leak detection.
Natural gas combustion does not emit high amounts of particulate matter per J of energy in the fuel, but the widespread use of natural gas suggests that it could still contribute significantly to ambient PM0.1 concentrations.Natural gas combustion accounted for 29% of 320 total US energy consumption in 2016 (Energy 2017).In contrast, gasoline combustion accounted for 17% of US energy consumption and diesel fuel combustion accounted for approximately 6% of US energy consumption in 2016.Gasoline and diesel fuel combustion in motor vehicles also emit most particles in the size fraction larger than PM0.1 (Robert, VanBergen et al. 2007a, Robert, Kleeman et al. 2007b) whereas natural gas combustion emits particles 325 entirely within the PM0.1 size fraction (Chang, Chow et al. 2004).Taken together, these facts support the potential importance of natural gas combustion for ambient PM0.1 concentrations.
The five (5) states with the highest consumption of natural gas in 2016 were Texas (14.7%),California (7.9%), Louisiana (5.7%),New York (5%), and Florida (4.8%).These consumption patterns are reflected in the natural gas distribution system (Figure 6a) and the 330 predicted PM0.1 concentration field associated with natural gas combustion (Figure 6b).Natural gas end-use included electric power generation (36%), industrial applications (34%), residential use (16%), commercial use (11%), and transportation (3%). of PNC.They found that PNC was made up of 36% on-road gasoline, 31% industrial, 18% nonroad diesel, 10% on-road diesel, 1% biomass burning and 4% long-range transport (Posner and Pandis 2015).The emissions particle number inventory was normalized based on PM10 mass 350 from each source and particle emissions from natural gas sources were assumed negligible, which effectively removed natural gas sources from the simulation.This has minor effects on PM2.5 and PM10 predictions, but the results of the current study suggest that natural gas combustion contributions significantly to ultrafine particle concentrations.

Conclusion
The UCD/CIT regional chemical transport model was used to predict source contributions to PM0.1 across the continental United States during peak photochemical smog spatial gradients were sharper than PM2.5 spatial gradients due to the dominance of primary aerosol in PM0.1.This finding suggests that PM0.1 measurement networks needed to support 420 epidemiology must be denser than comparable PM2.5 measurement networks.Non-residential natural gas combustion was identified as a major source of PM0.1 across all major cities in the United States.On-road gasoline and diesel vehicles made significant contributions to regional PM0.1 in all 39 cities even though peak contributions within 0.3 km of the roadway were not resolved by the 4 km grid cells.This is consistent with other studies that have found an 425 exponential decrease in ultrafine particle concentrations outside of major roadways (Wang, Hopke et al. 2011).Food cooking also made significant contributions to PM0.1 in all cities but biomass combustion was only important in locations impacted by summer wildfires.Aviation was a significant source of PM0.1 in cities that had airports within their urban footprints.The major sources of primary PM0.1 and PM2.5 were notably different in many cities.Future 430 epidemiological studies may be able to differentiate PM0.1 and PM2.5 health effects by contrasting cities with different ratios of PM0.1 / PM2.5 sources.

70
the low concentration of PM0.1 mass, which challenges the detection limits of analytical methods, Atmos.Chem.Phys.Discuss., https://doi.org/10.5194/acp-2018-833Manuscript under review for journal Atmos.Chem.Phys.Discussion started: 3 September 2018 c Author(s) 2018.CC BY 4.0 License.(ii) the artifacts associated with collecting PM0.1 samples, (iii) the additional workload involved in operating the collection devices, and (iv) the sharp spatial gradients of PM0.1 fields.Expensive investments in PM0.1 monitoring are unlikely to occur without compelling evidence linking PM0.1 to public health.Early epidemiological studies for PM0.1 must therefore use some 75 other technique besides direct measurements to calculate population exposure.Various methods such as the source-resolved PMCAMx chemical transport model, the chemical mass balance (CMB) model, photochemical box models and land use regression (LUR) Atmos.Chem.Phys.Discuss., https://doi.org/10.5194/acp-2018-833Manuscript under review for journal Atmos.Chem.Phys.Discussion started: 3 September 2018 c Author(s) 2018.CC BY 4.0 License.

190
photochemical smog episodes.Figure1illustrates the time series of predicted vs measured ozone concentration for four (4) representative cities spanning the South, East, Midwest and Atmos.Chem.Phys.Discuss., https://doi.org/10.5194/acp-2018-833Manuscript under review for journal Atmos.Chem.Phys.Discussion started: 3 September 2018 c Author(s) 2018.CC BY 4.0 License.PM concentrations.MFB values lower than 0.15 and MFE values lower than 0.35 are considered the goal or "excellent" in model performance.In the current study, the average MFB and MFE across all 39 cities was 0.126 and 0.379 for O3, and -0.27 and 0.38 for PM2.5 respectively.PM0.1 measurements are not available for model evaluation in the 39 cities across the US 225 in 2010 at the core of the current study, but measurements are available in California in the years 2015 and 2016 that can serve to evaluate similar modeling procedures.Yu et al. (Yu, Venecek et al. 2018) compared PM0.1 concentrations in Los Angeles, Fresno, East Oakland, and San Pablo, California predicted using the UCD/CIT air quality model to receptor-based source apportionment calculations based on measured concentrations of molecular markers in the 230 ultrafine particle size fraction(Xue, Xue et al. 2018).Good agreement was found between due to secondary formation (sulfate and secondary organic aerosol) did not have corresponding high concentrations of PM0.1.Most major urban centers had noticeable peaks of both PM2.5 and PM0.1.This pattern presents a challenge for epidemiological studies seeking to differentiate the effects of PM2.5 and PM0.1 because the locations with differential exposure (high PM2.5 but low PM0.1) have low population density which will reduce the power of 250 the analysis.Atmos.Chem.Phys.Discuss., https://doi.org/10.5194/acp-2018-833Manuscript under review for journal Atmos.Chem.Phys.Discussion started: 3 September 2018 c Author(s) 2018.CC BY 4.0 License.The UCD/CIT model explicitly tracks source contributions to particle mass in each size bin using artificial source tags.Pie charts of PM2.5 and PM0.1 source contributions are illustrated in Figure 3 for selected major cities.Pie charts for PM0.1 source contributions in all 39 US cities are shown in Figure 4.The detailed source profiles within each city are based on the nested 4km 255 Atmos.Chem.Phys.Discuss., https://doi.org/10.5194/acp-2018-833Manuscript under review for journal Atmos.Chem.Phys.Discussion started: 3 September 2018 c Author(s) 2018.CC BY 4.0 License.

Figure 1 .Figure 2 .
Figure 1.Time series of 1-hr measured vs predicted ozone concentration (ppm) for 4 selected city scenarios representative of the major geographical regions across the Continental United States

Figure 4 .Figure 5 .Figure 5
Figure 4. PM0.1 source contribution for 39 cities across the continental US Lane et al. (2007) used a source-resolved version of PMCAMx and individual emission inventories to determine source contributions of primary organic material (POM2.5)(Lane,Pinder   335   et al. 2007).Lane et al. note that POM2.5 associated with natural gas sources ranged from 0.1 to 0.8 µg/m 3 .Chang et al in 2004 measured emitted particle size distributions for gas-fired stationary combustion that fell between 10-100nm(Chang, Chow et al. 2004).The combination of these two results indicates that the natural gas mass component of POM2.5 predicted by Lane et al. is consistent with the magnitude of the PM0.1 mass associated with natural gas combustion 340 Atmos.Chem.Phys.Discuss., https://doi.org/10.5194/acp-2018-833Manuscript under review for journal Atmos.Chem.Phys.Discussion started: 3 September 2018 c Author(s) 2018.CC BY 4.0 License.found in the current study.Lane et al. were not studying PM0.1 and so the major role of natural gas in this size fraction was not identified.Posner and Pandis (2015) utilized PMCAMx with the LADCO 2001 BaseE sourceresolved mass emissions inventory for a July 2001 prediction of PNC over the Eastern United States with 36 km resolution (Posner and Pandis 2015).Posner and Pandis used a "zero-out" 345 method in combination with source-specific size distribution to study the percent contribution of six major sources (on road gasoline, industrial, non-road diesel, on road diesel, biomass and dust)

410
periods during the year 2010.Model performance for PM2.5 and ozone predictions met the recommendations for regulatory applications building confidence in the emissions inputs and meteorological fields used to drive the calculations.Similar model exercises carried out for episodes in California in 2015 and 2016 find good agreement between predicted PM0.1 source Atmos.Chem.Phys.Discuss., https://doi.org/10.5194/acp-2018-833Manuscript under review for journal Atmos.Chem.Phys.Discussion started: 3 September 2018 c Author(s) 2018.CC BY 4.0 License.contributions and receptor-based PM0.1 source contributions calculated using measured 415 concentrations of molecular markers (Yu, Venecek et al. 2018).Regional PM0.1 concentrations exceeded 2 µg m -3 during summer pollution episodes in major urban regions across the US including Los Angeles, the San Francisco Bay Area, Houston, Miami, and New York.PM0.1

Table 1 .
City, Simulation Date, 2010 Population and Geographical