Measurements of submicron aerosols in Houston, Texas during the 2009 SHARP field campaign

Transcription

1 JOURNAL OF GEOPHYSICAL RESEARCH: ATMOSPHERES, VOL. 118, 10,518 10,534, doi: /jgrd.50785, 2013 Measurements of submicron aerosols in Houston, Texas during the 2009 SHARP field campaign Misti E. Levy, 1 Renyi Zhang, 1 Alexei F. Khalizov, 1 Jun Zheng, 1,2 Don R. Collins, 1 Crystal R. Glen, 1,3 Yuan Wang, 1 Xiao-Ying Yu, 4 Winston Luke, 5 John T. Jayne, 6 and Eduardo Olaguer 7 Received 2 July 2013; revised 22 August 2013; accepted 24 August 2013; published 20 September [1] During the field campaign of the Study of Houston Atmospheric Radical Precursors/ Surface-Induced Oxidation of Organics in the Troposphere (SHARP/SOOT) in Houston, Texas, a suite of aerosol instruments was deployed to directly measure a comprehensive set of aerosol properties, including the particle size distribution, effective density, hygroscopicity, and light extinction and scattering coefficients. Those aerosol properties are employed to quantify the mixing state and composition of ambient particles and to gain a better understanding of the formation and transformation of fine particulate matter in this region. During the measurement period, aerosols are often internally mixed, with one peak in the effective density distribution at 1.55 ± 0.07 g cm 3, consistent with a population composed largely of sulfates and organics. Episodically, a second mode below 1.0 g cm 3 is identified in the effective density distributions, reflecting the presence of freshly emitted black carbon (BC) particles. The measured effective density demonstrates a clear diurnal cycle associated with primary emissions from transportation and photochemical aging, with a minimum during the morning rush hour, increasing from 1.4 to 1.5 g cm 3 on average over 5 h, and remaining nearly constant throughout the afternoon. The average BC concentration derived from light-absorption measurements is 0.31 ± 0.22 μgm 3, and the average measured particle single scattering albedo is 0.94 ± When elevated BC concentrations are observed, typically during the morning rush hours, single scattering albedo decreases, with a smallest measured value of about 0.7. Aerosol hygroscopicity measurements indicate that larger particles (e.g., 400 nm) are more hygroscopic than smaller particles (e.g., 100 nm). The measurements also reveal discernable meteorological impacts on the aerosol properties. After a frontal passage, the average particle effective density decreases, the average BC concentration increases, and the aerosol size distribution is dominated by new particle formation. Citation: Levy, M. E., et al. (2013), Measurements of submicron aerosols in Houston, Texas during the 2009 SHARP field campaign, J. Geophys. Res. Atmos., 118, 10,518 10,534, doi: /jgrd Introduction [2] Houston is the fourth largest city in the United States in one of the most rapidly expanding regions in the country [U.S. Census Bureau, Population Division, 2012], which 1 Department of Atmospheric Sciences, Texas A&M University, College Station, Texas, USA. 2 School of Environmental Science and Engineering, Nanjing University of Information Science & Technology, Nanjing, China. 3 Sandia National Laboratories, Albuquerque, New Mexico, USA. 4 Pacific Northwest National Laboratory, Richland, Washington, USA. 5 NOAA/Air Resources Laboratory, Silver Spring, Maryland, USA. 6 Center for Aerosol and Cloud Chemistry, Aerodyne Research, Inc., Billerica, Massachusetts, USA. 7 Houston Advanced Research Center, The Woodlands, Texas, USA. Corresponding author: R. Zhang, Department of Atmospheric Sciences, Texas A&M University, College Station, TX 77843, USA. (renyi-zhang@tamu.edu) American Geophysical Union. All Rights Reserved X/13/ /jgrd has a history of exceeding the Environmental Protection Agency s (EPA) National Ambient Air Quality Standards (NAAQS) [Lei et al., 2004; Zhang et al., 2004a]. Between 2000 and 2005, this region exceeded the 8 h O 3 standard of 75 parts per billion (ppb) on 219 occasions [Cowling et al., 2007]. Several atmospheric field campaigns, including the Texas Air Quality Study (TexAQS) I (2000) and II ( ), the Gulf of Mexico Atmospheric Composition and Climate Study in 2006, and the TexAQS- II Radical and Aerosol Measurement Project TexAQS-II Radical and Aerosol Measurement Project (TRAMP), have been conducted to assess emissions of volatile organic compounds (VOCs) and nitrogen oxides and ozone formation in southeast Texas [Daum et al., 2003; Olaguer et al., 2009; Parrish et al., 2009; Lefer et al., 2010]; photochemical oxidation of VOCs in the presence of NO x contributes to ozone formation [Nesbitt et al., 2000; Bond et al., 2002; Daum et al., 2003]. The overarching goals of the previous field campaigns have been focused to identify cost-effective solutions to the ozone problem for this region. For example, considerable 10,518

2 Figure 1. Map of the sampling site and back trajectories using the National Oceanic and Atmospheric Administration-Hybrid Single-Particle Lagrangian Integrated Trajectory (HYSPLIT) model to illustrate the synoptic flow patterns during the period of 30 April to 7 May The green trajectory with north winds corresponds to a frontal passage occurring on 3 May attention has been placed on the rapid formation and transport of ozone [Berkowitz et al., 2004; Jiang and Fast, 2004; Parrish et al., 2009], emissions of VOCs and NO x [Mellqvist et al., 2007], secondary organic aerosols (SOA), and PM 2.5 [Russell et al., 2004; Fan et al., 2005, 2006; Massoli et al., 2009]. In particular, rapid ozone formation (> 40 ppb h 1 ) associated with emissions of highly reactive olefins, such as ethene and propene from refineries and chemical plants, has been identified in the Houston region [Daum et al., 2003; Ryerson et al., 2003; Wert et al., 2003]. Furthermore, during high ozone days, the concentrations of organic aerosols have been frequently observed to exceed 12 μgm 3, which is the current NAAQS annual PM 2.5 standard [Dechapanya et al., 2003; Bahreini et al., 2009]. The oxidation products of VOCs emitted from biogenic and anthropogenic sources contribute importantly to secondary organic aerosol formation [Lei et al., 2000a, 2000b; McGivern et al., 2000; Hallquist et al., 2009]. In 2009, the Study of Houston Atmospheric Radical Precursors (SHARP) campaign was conducted to further investigate significant undercounted primary and secondary sources of the radical precursors, including formaldehyde (CH 2 O) and nitrous acid (HONO), in both heavily industrialized and more typical urban areas of Houston (E. P. Olaguer et al., Overview of the SHARP campaign: Motivation, design, and major outcomes, submitted to Journal of Geophysical Research, 2013). [3] In addition to their impacts on air quality and visibility, aerosols exert profound impacts on atmospheric chemical processes [Zhang et al., 1996a; Kroll and Seinfeld, 2008; Hallquist et al., 2009], cloud formation, and regional and global climate [Haywood and Boucher, 2000; Ramanathan et al., 2001; Forster et al., 2007]. For example, aerosols influence the climate directly by reflecting solar radiation [Forster et al., 2007] and indirectly by modifying cloud formation [Fan et al., 2007, 2008], modulating photochemistry [Li et al., 2005; Flynn et al., 2010], and promoting multiphase chemistry [Zhang et al., 1995, 1996b; Zhao et al., 2005; Osthoff et al., 2008; Ziemba et al., 2010; Qiu et al., 2011, 2012]. The overall impact of aerosols is dependent on the size, concentration, and chemical composition of particles. Elevated concentrations of atmospheric aerosols also have an adverse effect on human health [Mokdad et al., 2004; Dominici et al., 2006]. Research on aerosols remains an active field since aerosols correspond to the greatest source of uncertainty in projecting climate changes [Forster et al., 2007; Zhang, 2010]. [4] Aerosols in the atmosphere are broadly categorized as primary and secondary particles according to the origins of their formation. One important primary aerosol type is black carbon (BC), produced from incomplete combustion of fossil fuel and biomass burning [Zhang and Zhang, 2005; Khalizov et al., 2012], while sulfate, ammonia, nitrate, and organics represent the key compounds responsible for formation and growth of secondary particles [Zhang et al., 2004b; Zhang et al., 2012]. Light-absorbing particles such as BC [Horvath, 1993; Bond and Bergstrom, 2006] and brown carbon [Bond, 2001; Andreae and Gelencsér, 2006; Moosmüller et al., 2009] have been identified as a major contributor to global warming [Jacobson, 2001; Bond et al., 2013]. One particular challenge in evaluating the aerosol direct radiative forcing occurs when nonabsorbing aerosols, such as sulfates and organics, and light-absorbing aerosols, such as BC, coexist in the same air mass [Ramanathan and Carmichael, 2008; Ramana et al., 2010]. Recent laboratory experiments [Khalizov et al., 2009a; Xue et al., 2009a, 2009b; Qiu et al., 2012; Khalizov et al., 2013], field observations [Knox et al., 2009; Moffet and Prather, 2009], and modeling calculations [Jacobson, 2000, 2001] show that, when mixed with other nonabsorbing aerosol constituents, such as sulfate and organics, BC is more absorptive and exerts a higher positive direct radiative forcing. Currently, there is a large uncertainty regarding the magnitude of absorption enhancement due to BC coating [Zhang et al., 2008; Subramanian et al., 2010; Cappa et al., 2012]. Also, information on the concentration and the mixing state of BC aerosols in different geographical regions is limited. [5] In this paper, we report field measurements of submicron ambient aerosols during late springtime in Houston, Texas. As a subproject under the 2009 SHARP field campaign, the Surface-Induced Oxidation of Organics in the Troposphere (SOOT) led by Texas A&M University was conducted to measure ambient gaseous nitrous acid and aerosol properties and to investigate HONO formation and BC aging using an environmental chamber, with the objective to assess the role of heterogeneous chemistry in the radical budget, VOC oxidation, and O 3 formation (Olaguer et al., submitted manuscript, 2013). We emphasize in this work the formation and transformation of submicron aerosols in this region on the basis of a comprehensive set of simultaneously measured ambient aerosol properties. 10,519

3 2. Methodology and Instrumentation 2.1. SHARP/SOOT Campaign [6] The SHARP Campaign extended from 15 April through 31 May 2009 (Olaguer et al., submitted manuscript, 2013). The SOOT project consisted of two components, i.e., (1) ambient measurements of gaseous HONO and aerosols and (2) environmental chamber studies of HONO formation and soot aging using experimentally generated BC particles. Our instruments were stationed on the University of Houston (UH) northern Moody Tower (70 m above the ground level), which is an 18-story building located 5 km south of downtown Houston, 5 km southwest of the Ship Channel, and within 8 km of three major highways. This location, shown in Figure 1, allows measurements of a representative sample of inner-city Houston air without the interference of extremely localized vehicular emissions. During the period from 30 April to 7 May, the SOOT project focused on ambient measurements. From 8 to 31 May, a series of environmental chamber studies were conducted. As a result, measurements of ambient particle sizes, effective density, and hygroscopicity were sporadic beyond the week period of 30 April and 7 May. [7] The data presented in this work span between 30 April and 7 May 2009, representing the normal springtime synoptic conditions in this region. Figure 1 illustrates the major wind circulation patterns during this period, i.e., the frequent southerly wind from the Gulf of Mexico and the northerly wind after the passage of a cold front Instruments [8] A suite of aerosol instruments inside an air-conditioned trailer was deployed to measure particle number size distributions, effective densities, hygroscopic growth factors, and light extinction and scattering. Atmospheric aerosols were sampled through a 6 m long thermally insulated 3/8 inch outer diameter copper tube at 10 liters per minute (lpm) flow rate. Multitube Nafion driers (PD T-24SS, Perma Pure, Inc.) were used to reduce the relative humidity of the aerosol flow below 10% and to convert particles to an anhydrous state. This procedure was necessary to retrieve the dry-particle diameter and effective density and to establish a closure with the optical and chemical composition measurements. Also, removal of the aerosol water content prevented the variations in the size distributions and optical properties of hydrated particles due to changes in ambient relative humidity [Massoli et al., 2009; Taylor et al., 2011]. Two driers connected in series were used with a 6.5 lpm flow sampled by the aerosol optical instruments. Two additional driers were used with 1.0 and 2.5 lpm flows sampled by the size/density and hygroscopicity instruments, respectively. Evaporative losses of semivolatile species were minimized by placing the driers inside the air-conditioned trailer. Diffusion-controlled particle losses in the driers were taken into account by an inversion algorithm when calculating the particle size distributions. Losses of optically important particles were estimated to be less than 1%. [9] Particle size distributions and effective density distributions were measured by a system consisting of a differential mobility analyzer (DMA, model 3081, TSI, Inc.), an aerosol particle mass analyzer (APM, model 3600, Kanomax Inc., Japan), and a condensation particle counter (CPC, model 3760A, TSI, Inc.) [Khalizov et al., 2009b; Pagels et al., 2009; Qiu et al., 2012]. The DMA operated in a recirculating flow configuration, with a sheath flow of 6.5 lpm and a sample flow of 1 lpm, corresponding to a particle mobility diameter range of about 10 to 400 nm. A relatively low sheath-tosample flow ratio was used to achieve a better detection of small particle concentrations and a shorter residence time in the DMA-APM measurements. The DMA transfer function was broadened but remained symmetrical. The application of the DMA-APM technique in the mass-mobility measurements has been previously described [McMurry et al., 2002; Pagels et al., 2009]. The APM consisted of two cylindrical electrodes that rotated at the same angular speed [Tajima et al., 2011]. Charged particles were introduced axially into the annular space between the electrodes and rotated at the same speed as the electrodes, when particles traveled through the APM. The mass of a particle that passed through the APM was determined by the rotational speed of the cylindrical electrodes and the voltage applied to the inner electrode. The aerosol mass distribution was determined by stepping the voltage at a fixed rotation speed and measuring the concentration of particles passing through APM by CPC. There were 30 points observed between 0.10 and 2.60 g cm 3 for the five distinct particle mobility diameters (46, 81, 151, 240, and 350 nm) approximately every hour. Although the APM allows for direct determination of particle effective density (ρ eff ) based on the particle mass, substantially higher accuracy (better than 5%) can be achieved by measuring ρ eff relative to the polystyrene latex (PSL) particles according to ρ eff ¼ V APM ρ V APM;PSL PSL (1) where V APM and V APM, PSL are the peak APM voltages corresponding to the masses of ambient and PSL particles of identical initial mobility diameter, and ρ PSL = g cm 3 is the material density of polystyrene latex. [10] To determine the size-resolved hygroscopic growth factor (HGF), a hygroscopic tandem-dma (H-TDMA) was employed. Monodisperse aerosols produced by the first DMA were exposed to elevated humidity in a multitube Nafion drier/humidifier. The changes in the particle mobility diameters were measured by the second DMA [Gasparini et al., 2004; Khalizov et al., 2009b]. The growth factor corresponds to the ratio of the particle sizes measured by the two DMAs, HGF ¼ DP D o (2) where D p and D o are the processed and initially unprocessed particle diameters, respectively. [11] The optical properties of ambient aerosols were quantified from direct measurements of light extinction and scattering. Aerosol extinction coefficients at 532 nm, b ext, were measuredbyacavityring-downspectrometer(crds)[smith and Atkinson, 2001; Khalizov et al., 2009a; Radney et al., 2009]. The CRDS cavity consisted of two high-reflectivity dielectric mirrors ( % reflectivity, 6 m radius of curvature, Los Gatos Research, Inc., CA) and a stainless steel cell with an aerosol inlet in the center and two outlets near 10,520

4 Inlet Size Distribution Effective Density Hygroscopicity Optical Properties DMA CPC DMA APM CPC DMA Humidifier Cavity Ring- DMA CPC Cavity Down Ring Down 46, 81, 151, 240, 15, 20, 50, 100, Spectrometer. 350 nm 200, 400 nm Nephelometer Figure 2. Schematic of the instrumentation arrangements and measured aerosols properties. the ends. To prevent contamination, the mirror region was purged with a small flow of filtered dry air. A light pulse (λ = 532 nm) from a Q-switched laser (QG , CrystaLaser, CA) was injected into the cavity through the front mirror, the light leaking through the rear mirror was detected with a photomultiplier (H , Hamamatsu), and the signal was digitized by a 100 MHz, 12-bit resolution CompuScope card (GaGe Applied Technologies, LLC) operated by the LabVIEW software. During measurements, 3000 ring-down traces were averaged at a 50 Hz repetition rate, and the decay time was calculated by nonlinear fitting of the averaged decay data. The detection sensitivity of this instrument was better than 0.5 Mm 1. [12] The scattering coefficients at 532 nm, b sca, were derived from measurements by a commercial three-wavelength integrating nephelometer (3563, TSI, Inc.). In the nephelometer, the aerosol flow passed through the measurement volume, where particles were illuminated over an angle of 7 to 170 by a halogen bulb light source. The sample volume was viewed by three photomultiplier tubes (PMTs) through a series of apertures along the axis of the instrument. Light scattered by particles was split into three wavelengths (450, 550, and 700 nm) using color filters in front of the PMT detectors. Following the calibration/correction and measurement protocols suggested by Anderson and Ogren [1998], the detection limit of this instrument was better than 0.2 Mm 1. In order to obtain a closure with the measured extinction coefficient, the scattering coefficient at 532 nm was calculated from a power law fit to the scattering coefficients at the three nephelometer wavelengths [Khalizov et al., 2009a]. During field measurements, the CRDS and nephelometer were automatically zeroed every hour using filtered ambient air. [13] Single scattering albedo (SSA) of aerosols is represented as the ratio of scattering (b sca )toextinction(b ext )coefficients SSA ¼ bsca b ext (3) [14] During the observation period, less than 7.5% of the measured SSA values exceeded unity. The absorption coefficient, b abs, was calculated from the difference between the extinction and scattering coefficients. This procedure has been previously utilized to obtain aerosol absorption in laboratory experiments and field observations [e.g., Moosmüller et al., 2009, and references therein]. To minimize systematic errors due to temporal variations in the aerosol concentration, extinction and scattering measurements in the optical system were performed nearly simultaneously. The system was calibrated with nonabsorbing ammonium sulfate aerosols, for which the extinction coefficient equals the scattering coefficient. The detection limit for absorption using the difference method was within 0.6 Mm 1, on the basis of laboratory calibrations [Khalizov et al., 2009a; Xue et al., 2009b; Khalizov et al., 2013]. During the field measurements, less than 7% of the derived absorption coefficients were below zero. [15] A Sunset organic carbon and elemental carbon aerosol analyzer (Sunset Laboratories, Inc.) was used to determine the mass concentrations of organic carbon (OC) and elemental carbon (EC in the form of optical and thermal EC) according to the modified NIOSH (National Institute of Occupational Safety and Health) protocol [Bauer et al., 2009]. Negative optical EC values may be due to a bias from the loss of native EC at high temperatures, which is less than 0.01 μgm 3 and does not influence the BC concentration. To obtain a closure with the Sunset EC measurements, the mass concentration of BC was also calculated from light-absorption coefficients derived by the difference method, assuming uniform values of the mass absorption cross section (MAC) of 7.5 and 11 m 2 g 1 for fresh and aged BC, respectively [Bond and Bergstrom, 2006], BC ¼ babs MAC (4) [16] The absolute error in BC mass concentration due to uncertainties in MAC and b abs was estimated to be within 0.21 μgm 1. In addition, MAC was also calculated on the basis of the measured absorption coefficient and Sunset optical and thermal EC concentrations using equation (4). [17] Aerosol measurements by DMA-APM, H-TDMA, and the optical instruments were performed simultaneously. Each DMA-APM measurement cycle involved the acquisition of a single size distribution followed by effective density measurements for five different particle diameters; a full cycle took about 1 h to complete. Similarly, H-TDMA measurements involved acquiring a size distribution followed by six size-resolved hygroscopicity scans, each cycle occurring in about 20 min. Optical measurements were carried out continuously with a temporal resolution of 1 min. A schematic of the instrumentation arrangements is provided in Figure 2. [18] In addition to the size-average effective density determined by the DMA-APM, the bulk particle density was 10,521

5 Figure 3. Meteorological parameters: (a) wind speed (black) and direction (red) and (b) relative humidity (red) and temperature (black) during the period of 30 April to 7 May Northerlywindsoccurredbetween15hon3Mayand9hon 4 May at the UH site. calculated from the submicron aerosol composition data collected by an aerosol chemical speciation monitor (ACSM, Aerodyne Research, Inc.). The ACSM measured the aerosol total mass loading and chemical composition including ammonium, sulfate, organics, nitrate, and chloride in 30 min cycles. The ACSM overall density ρ was derived from the mass fractions f i and material densities ρ i of these components, assuming that ammonium sulfate, ammonium bisulfate, ammonium nitrate, and organics represent the only particle constituents, 1 ρ ¼ f i i (5) ρ i where the summation over f i equals 1, and the densities of the ammonium sulfate, ammonium bisulfate, and ammonium nitrate are 1.76, 1.78, and 1.73 g cm 3, respectively. The effective density of organic matter may be dependent on the origin of the air masses, since the UH site is impacted by variable aerosol emission sources. Sensitivity calculations over periods with well-defined air mass origins indicated that the best fit effective densities for marine and land originating organics were 1.34 and 1.45 g cm 3, respectively. These values are within the range of the organic aerosol densities reported previously [Dinar et al., 2006; Malloy et al., 2009]. For simplicity, an average value of 1.40 g cm 3 was adopted in our present analysis. [19] Trace gases were sampled through perfluoropolymer tubing. The VOC mixing ratios were obtained with a proton transfer reaction mass spectrometer (PTR-MS, Ionicon Analytik, Austria). SO 2, CO, NO x, and O 3 mixing ratios were measured with the modified Thermo 43, 48, 42, and 49, respectively. A detailed description of the measurements of trace gases has been provided in this special issue (Olaguer et al., submitted manuscript, 2013). 3. Results and Discussion 3.1. Meteorological Conditions [20] Air quality in the Houston region is also dependent on the local meteorological conditions [Lei et al., 2004]. The meteorological parameters from 30 April through 7 May 2009 are presented in Figure 3. This region is typically humid because of its proximity to the Gulf of Mexico and the frequent influx of marine air. There is a frontal passage at 17 h (inferred from the change in the wind direction) on 3 May 2009, which causes precipitation and brings in northerly winds and colder air. The wind is 88% southerly flow (ESE-SW, referred as to the southerly flow condition) and 12% northerly flow (WNW-ENE, referred as to the northerly flow condition) during 30 April and 7 May, with an average wind speed of 5 m s 1 (Figure 3a). The average temperature is 25 C (77 F), and the average relative humidity is about 75% during this period (Figure 3b). A northerly flow is characteristic of a relatively polluted continental air mass, while a southerly flow is associated with a relatively clean marine air mass (J. P. Pinto et al., Intercomparison of field measurements of nitrous acid (HONO) during the SHARP Campaign, submitted to Journal of Geophysical Research, 2013). There is little occurrence of wind direction reversal in the late afternoon and nighttime hours during this period (Figure 1), which otherwise circulates aged pollutant plumes characterized by elevated O 3 and NO x back to the Houston region (Pinto et al., NOx (ppb) SO 2 (ppb) VOCs (ppb) Ozone (ppb) CO (ppb) /30 05/01 05/02 05/03 05/04 05/05 05/06 05/ /30 05/01 05/02 05/03 05/04 05/05 05/06 05/07 2 Isoprene Toluene Benzene / /01 05/02 05/03 05/04 05/05 05/06 05/ /30 05/01 05/02 05/03 05/04 05/05 05/06 05/ A B C D E /30 05/01 05/02 05/03 05/04 05/05 05/06 05/07 Time Figure 4. Mixing ratios (in ppb) of ambient trace gases during the period of 30 April to 7 May The vertical lines correspond to the midnight of the local time (CST). 10,522

6 Figure 5. Aerosol mobility size distributions: (a) absolute number concentration for nm diameter particles during the period of 30 April to 7 May The dates marked correspond to the midnight of the local time (CST). (b) The normalized number concentration. The normalized plot is derived by dividing each observation (consisting of approximately 150 points between 20 and 400 nm) by the maximum concentration in each observation. submitted manuscript, 2013). Figure 3 shows a period with a wind speed of less than 5 m s 1 after the frontal passage, indicating the stagnant conditions that are favorable for pollutant accumulation [Lei et al., 2004] Trace Gases [21] Ozone, VOCs (i.e., isoprene, toluene, and benzene), CO, NO x, and SO 2 measured between 30 April and 7 May are depicted in Figure 4. There are overall lower concentrations of gaseous pollutants during the southerly flow condition than those after the frontal passage. This is particularly noticeable in the O 3 concentration, which exceeds the EPA 8h O 3 standard of ppm on 4 May but is well below the standard on the southerly wind days with an average ozone concentration of ± 3.80 ppb. On 4 May, the concentrations of SO 2, toluene, and benzene rise noticeably after the frontal passage. The isoprene concentration is relatively independent of the synoptic pattern, but there exists a strong diurnal variation, indicating the biogenic production [Li et al., 2007]. The concentration of NO x is comparable to measurements by other groups during the same period [Wong et al., 2012] and during previous field campaigns in Houston [Stutz et al., 2010]. A peak NO x concentration of about 25 ppb occurs on 3 May during the transitory period of the two distinct wind patterns, likely attributable to local emissions. It is evident from Figure 4 that the increases in CO and NO x occur during the morning and evening rush hours throughout the measurement period, along with concurrent decreases in O 3 by NO titration, which are indicative of traffic emissions Aerosol Size Distributions [22] Figure 5 shows the aerosol size distributions observed between 30 April and 7 May. The normalized plot is obtained by dividing each observation (consisting of approximately 150 points between 20 and 400 nm) by the maximum concentration in each observation (Figure 5b). The normalized plot illustrates the most prevalent diameter, displaying the evolution of the peak particle mobility diameter during a day. For example, the morning rush hour emissions are distinguished by an increase in the concentrations of nm aerosols [Ban-Weiss et al., 2010]. The aerosol growth is also discernable, because the highest concentration of particles is located between 50 and 100 nm in the morning and then between 150 and 250 nm in the afternoon. During a typical day, high concentrations of small particles ( nm) are observed near sunrise and sunset, likely due to the morning and evening traffic rushes and consistent with the correlated increases in CO and NO x emitted by traffic and O 3 decreases by NO titration in the trace gas measurements (Figure 4). The greatest concentration of larger particles (> 150 nm) occurs in the afternoon, likely reflecting the growth of small particles emitted or freshly formed in the morning. [23] On several occasions, a high concentration of less than 30 nm particles is observed, which may be explained by new particle formation [Zhang et al., 2012]. Although our instrumentation does not observe particles below 10 nm, the observed changes in the aerosol properties and concentrations suggest the occurrence of new particle formation. New particle formation is particularly evident by the highest particle number concentrations and growth from 10 nm and above after the frontal passage (Figure 5): nm size particles account for 53% of the total aerosol concentration. New particle formation is likely explainable by elevated gaseous sulfur and organic compound levels (Figure 4), which enhance the nucleation and growth rates of nanoparticles [Zhang et al., 2009]. In particular, sulfur dioxide can be oxidized to sulfuric acid to enhance new particle formation [Wooldridge et al., 1995; McMurry et al., 2005; Zhao et al., 2009]. Also, a lower preexisting aerosol concentration may enhance the condensation of atmospheric gases onto nanoparticles (< 3 nm) and reduce scavenging losses of freshly nucleated particles [Wang et al., 2010a, 2010b]. On 4 May, the particle size distribution clearly exhibits the banana growth curve, characteristic of new particle formation [Yue et al., 2011]. New particle formation on this day occurs at 9 10 h, 10,523

7 40,000 30,000 Northerly Winds Southerly Winds Diameter (nm) a 20,000 10, Diameter (nm) Number Concentration (cm 3 ) b Time Figure 6. Averaged aerosol size distributions corresponding to southerly (black) and northerly winds (red), calculated by averaging the aerosol size distributions. Northerly winds occurred between 15 h on 3 May and 9 h on 4 May at the UH site. Figure 7. Diurnal trend of the (a) average aerosol diameter (nm) and (b) the average aerosol concentration (cm 3 ) observed between 30 April and 7 May when the concentration of nm particles reaches cm 3. The particle surface area is about 5 μm 2 cm 3, during this episode of new particle formation, indicating a lower condensation sink for condensable gases and freshly nucleated particles. The SO 2 concentration is about ppb between 9 and 10 h and increases to over 3 ppb in the late afternoon. Also, as shown in the normalized plot of Figure 5b, throughout the morning and afternoon of 4 May, newly nucleated particles grow to 75 nm in diameter, and the growth rate of the fine particles is about 10 nm h 1. On 4 May, the temperature, relative humidity, and wind speed are noticeably low in the morning hours (Figure 3), in the range of C, 55 75%, and 4 5ms 1 during 8 12 h, respectively. Hence, during this northerly flow period, the environmental conditions are favorable for new particle formation, since the temperature, wind speed, and relative humidity are low, the existing aerosol surface area is low, but the anthropogenic sulfate and organic gases (i.e., SO 2 and toluene) are elevated. New particle formation also occurs during the frontal passage in the afternoon hours on 3 May. [24] Figure 6 shows two averaged aerosol size distributions, calculated by averaging all the 18 aerosol size distributions corresponding to southerly and northerly wind conditions. For the southerly flow condition, there exist two distinct peaks: the peak occurring between 100 and 200 nm has an average concentration of 10,000 particles cm 3, and the secondary maximum near 30 nm has an average concentration of 5,000 particles cm 3, likely representing primary aerosols emitted directly with subsequent growth and new particle formation, respectively. After the frontal passage, the total concentration approximately quadruples. While aerosols between 175 and 400 nm are comparable in concentration to that of the southerly flow, particles of less than 100 nm in diameter are significantly more prevalent after the frontal passage. There is one major peak in this distribution occurring near 40 nm with a concentration of nearly 40,000 particles cm 3, suggesting that new particle formation represents the dominant process. [25] Figure 7 depicts the averaged diurnal cycle of the aerosol diameter and number concentration between 30 April and 7 May. The averaged particle size increases during the morning hours, decreases during the afternoon hours, and remains nearly constant during the night. The highest particle concentration occurs prior to sunrise, decreases during the daytime, and increases after 16 h, likely to be jointly attributable to emissions and the variation of the planetary boundary layer height that regulates the vertical mixing Particle Effective Density [26] Atmospheric aerosols can be mixed externally or internally. Since individual particle constituents have distinct effective densities, the chemical makeup of particles can be assessed on the basis of the measured particle density distributions. Although the material density of EC, the major component of BC, is relatively high (1.7 to 1.9 g cm 3 )[Bond and Bergstrom, 2006], the effective density of freshly emitted BC particles can be as low as gcm 3 because of their open, fractal morphology [Zhang et al., 2008; Pagels et al., 2009]. The observations of lower density peaks with the values below unity provide an indicator of fresh or partially aged BC. Figure 8 demonstrates the effective density Figure 8. Effective density distributions of 151 nm particles. The red line displays a unimodal distribution observed at 9:30 h on 2 May, and the blue line demonstrates a bimodal distribution observed at 8:30 h on 4 May. 10,524

8 Figure 9. Normalized particle effective density distributions for (a) 46 nm, (b) 81 nm, (c) 151 nm, (d) 240 nm, and (e) 350 nm particles during the observational period. The white lines indicate the average effective densities. The normalized plot is derived by dividing each observation by the maximum concentration in each observation. The dates marked correspond to the midnight of the local time (CST). distributions of 151 nm particles measured on two different days (i.e., 2 and 4 May). On 2 May, the density distribution is dominated by a single peak, likely signifying an internal mixture. Several features are noticeable in the shown peak density distribution; the effective density distribution peak ranges from 1.3 to 2.1 g cm 3 with a maximum peak value of near 1.6 g cm 3, which may indicate the presence of sea salt (material density of 2.17 g cm 3 ), ammonium sulfate (material density of 1.76 g cm 3 ), ammonium sulfate internally mixed with organic materials (~1.65 g cm 3 ), or mostly organics (~1.5 g cm 3 )[Kostenidou et al., 2007]. On 4 May, the density distribution is bimodal, likely indicating an external mixture. There exists a prominent low-density peak at 0.55 g cm 3, which corresponds to the presence of freshly emitted or partially aged BC. The high-density peak at 1.4 g cm 3 reveals more organics but less ammonium sulfate. 10,525

9 Figure 10. Temporal variation of the peak aerosol effective densities for 46, 81, 151, 240, and 350 nm particles during the observational period. The plot is produced by determining the peak of each measured effective density distribution from Figure 8. The dates marked correspond to the midnight of the local time (CST). [27] Figure 9 presents the temporal trends of the effective density distributions for five particle sizes, i.e., 46, 81, 151, 240, and 350 nm. The temporal resolution of our measurements is sufficient to capture the diurnal density changes and is more detailed than previously published [McMurry et al., 2002; Geller et al., 2006; Malloy et al., 2009]. While the overall trends in the effective densities of the five particle sizes are similar, some variances between different particle sizes are apparent. For all particle sizes, the lowest effective densities occur in the early morning, and the largest effective densities are observed in the afternoon. The magnitude of the effective density changes varies between the particle sizes, showing a decreasing trend for larger particles. [28] Figure 10 shows the temporal trend of the measured effective density peak maximum of the five particle sizes. The effective density distributions exhibit a single peak between 1.40 and 1.70 g cm 3, consistent with particles dominated by ammonium sulfate (1.77 g cm 3 ) internally mixed with a variable fraction of organic materials ( g cm 3 ) [Dinar et al., 2006; Malloy et al., 2009]. In some of the observations, the effective densities show a second, lower density constituent that is attributable to the presence of fresh or partially aged BC. Particles of 151 nm diameter have the largest percentage of two effective density components (32%) followed by 240 nm particles (7%), indicating fresh BC mostly in the nm size range. This observation is in agreement with the previously reported ranges for the effective densities and size distributions of soot emitted from diesel engines [Park et al., 2003; Burtscher, 2005; Ban-Weiss et al., 2010]. For the other three sizes (i.e., 46, 81, and 350 nm), the frequency of observation of bimodal effective density distributions is less than 3%, indicating that smaller (i.e., 46 and 81 nm) and larger particles (350 nm) are less likely to contain freshly emitted BC. It should be noted that the absence of lower density particles does not necessarily imply the absence of BC but merely indicates that BC particles are heavily aged and internally mixed. Few effective density distributions exhibit the peaks between and gcm 3, which correspond to externally mixed sulfate and organics, respectively. Turpin and Lim [2001] suggested that SOA from the oxidation of aromatics and alkanes emitted in the Houston Ship Channel has a density between 1.20 and 1.40 g cm 3, whereas Kostenidou et al. [2007] found the SOA density to be in the range g cm 3. [29] Figure 11 depicts the diurnal cycle of the averaged effective density of the five particle sizes between 30 April and 7 May. For all particle sizes, the effective densities vary between 1.45 and 1.65 g cm 3. The effective densities increase from the morning to the afternoon (i.e., 7 to 17 h), likely due to the increase in the particle-phase sulfate and oxidized organic components. There are two minima in the averaged effective densities at 7 and h. Increased BC emissions from transportation are likely responsible for the two minima, since the times correspond closely to the morning and evening rush hours and are consistent with those trends in the trace gases (i.e., elevated NO x and CO and decreased O 3 shown in Figure 4). The effective densities remain nearly constant throughout the nighttime. Figure 11 also reveals that particles of 46 nm (350 nm) exhibit the largest (smallest) diurnal variability. [30] The size-averaged effective density from our APM measurements is compared with the particle density derived from the ACSM in Figure 12. Both the APM and ACSM density data show a similar temporal and diurnal trends. The average effective density is higher during the southerly flow (1.58 ± 0.04 g cm 3 ) than after the frontal passage (1.47 ± 0.09 g cm 3 ). The elevated BC level observed after the frontal passage (Figure 9) may be responsible for a reduced particle effective density on 4 May (Figure 12a). After the frontal passage, the effective density range is reduced and the diurnal cycle is less apparent. Figure 12b shows that between 30 April and 7 May the averaged effective density derived from APM and ACSM are 1.55 ± 0.07 and 1.56 ± 0.08 g cm 3, respectively. [31] When particle composition is dominated by organics and ammonium sulfate, the measured particle density can be used to infer the mass fractions of the two components. As shown in the ACSM measured chemical compositions (Figure 12c), the two-component assumption holds true for most times in the Houston area, since chloride (1.21% ± 1.08) and nitrate (4.25% ± 2.06) represents minor constituents in nm 81nm 151 nm 240 nm 350 nm Average Time Figure 11. Diurnal trend of the average effective density (g cm 3 )offive particle sizes and the weighted average effective density observed between 30 April and 7 May ,526

10 Effective Density (g cm 3 ) A ACSM APM /30 05/01 05/02 05/03 05/04 05/05 05/06 05/07 ACSM Mass Fraction C Organic Sulfate 04/30 05/01 05/02 05/03 05/04 05/05 05/06 05/07 Effective Density (g cm 3 ) B APM ACSM Time APM Mass Fraction D Sulfate Organic 04/30 05/01 05/02 05/03 05/04 05/05 05/06 05/07 Time Figure 12. Particle effective density between 30 April and 7 May 2009: (a) the size-averaged effective density (g cm 3 ) measured by the APM and the effective density calculated from the ACSM aerosol composition data. (b) A comparison of the effective density diurnal cycles observed by the APM (red) and ACSM (black). (c) Mass fractions of sulfate (black) and organics (red) from the ACSM aerosol composition measurements. (d) The mass fraction of sulfate (black) and organics (red) derived from the APM effective density calculations. The APM mass fractions are calculated by using the weighted effective density from each effective density distribution in equation (5) and assuming a two-component model, as explained in section 2. submicron aerosols [Russell et al., 2004; Bates et al., 2008]. Figures 12c and 12d compare the aerosol mass fractions derived from the APM effective density using equation (5) and from the ACSM measurements. During the southerly flow, the dominance in the aerosol mass fraction oscillates between sulfate and organics. The peaks of organics occur in the midafternoon, indicating a photochemically driven aerosol formation mechanism. Bates et al. [2008] suggested that sulfate constitutes 51% 61% of the particle mass with the land originating air masses and 20% when influenced by marine air in the Houston area. The mass fractions are 64 ± 16% for sulfate and 36 ± 16% for organics after the frontal passage and 46 ± 11 for sulfate and 54 ± 11 for organics under southerly winds. The high sulfate mass fraction during and after the frontal passage can be explained by the low wind speed, which is favorable for stagnation. Overall, there is a good agreement between the APM and ACSM time-averaged mass fractions for both organics (50 ± 13% and 54 ± 19%, respectively) and sulfate (49 ± 13% and 46 ± 19%, respectively). On 4 May, the derived mass fraction exceeds unity because the two-component assumption becomes invalid. This may be explainable by the presence of high concentrations of freshly emitted BC shown in Figure 9 (also to be discussed later). [32] From the measured size-resolved effective density and number size distributions, the mass concentration of particles between 10 and 400 nm (PM 0.4 ) is calculated PM 0:4 ¼ 400 nm 10 nm 1 6 dn π D3 o ρ eff ΔD o (6) [33] Figure 13 depicts the calculated PM 0.4 along with the BC mass concentration and fraction derived from the optical measurements (to be discussed below). The average PM 0.4 is 9.4 ± 5.1 μgm 3 ; the lowest daily average of 4.1 μgm 3 occurs on 30 April, and the highest daily average value of 13.3 μgm 3 is observed on 3 May. Note that although the particle number concentration maximizes after the frontal passage on 4 May Figure 13. Mass concentrations (μgm 3 )ofpm 0.4 (black), BC (blue), and the mass percent of BC particles (red) during the observational period. The dates marked correspond to the midnight of the local time (CST). 10,527

11 LEVY ET AL.: AEROSOL PROPERTIES IN HOUSTON Figure 14. Optical properties of aerosols observed between 30 April and 7 May 2009: (a) measured extinction (black), scattering (red), and absorption (green) coefficients (Mm 1). The Mie scattering coefficient (orange) is shown for comparison, which is calculated using the number size distribution. (b) The angstrom exponents (red) and the single scattering albedo (the right axis, blue). (Figure 5), the total PM0.4 is relatively low, since the nucleation mode particles contribute negligibly to the total particle mass. Furthermore, a lower concentration of preexisting particles on 4 May favors new particle formation [Zhang et al., 2012]. Our calculation is in agreement with previous submicron mass concentrations measurements of 6.5 to 20.8 μg m 3 in this region [Bates et al., 2008] Aerosol Optical Properties [34] The temporal trends of the aerosol extinction, scattering, absorption, and single scattering albedo are illustrated in Figure 14. The average extinction and scattering coefficients are 49.7 ± 23.9 and 46.9 ± 22.7 Mm 1, indicating that scattering dominates the attenuation process. The scattering coefficient is also calculated using the Mie theory on the basis of the measured particle size distributions and components, and the predicted values are in agreement with the measurements (Figure 14a). Since SSA is closely correlated with the BC concentration, SSA exhibits a lesser day-to-day variation than the scattering coefficient, because of little variation in daily vehicle emissions. After the frontal passage on 4 May, the SSA variation is significantly enhanced, ranging from 0.70 to 1.00, and the average SSA decreases to 0.89 ± The lowest SSA is observed to be around 0.7 on 4 May. The temporal trend of the Ångstrom exponent derived from wavelength dependence of the scattering coefficient and the aerosol surface area is also illustrated in Figure 14b. The high values of the Ångstrom exponent indicate that there is little presence of dust during the measurements. [35] The diurnal averaged optical coefficients are presented in Figure 15a. The scattering, absorption, and extinction coefficients all increase during the daytime, attributable to the greater concentration of larger particles during the daytime (Figure 5). The average SSA is 0.94 ± 0.04 and exhibits a clear diurnal cycle (Figure 15b). SSA is the lowest in the morning hours (between 6 and 7 h), typically with a value between 0.8 and 0.9 and remains above 0.9 during the rest of the day. Occasionally, it is evident that there exists a secondary decreased SSA near 21 h, likely due to an increase of BC emissions from the evening rush hours. Figure 15. Diurnal cycles of (a) extinction, scattering, and absorption coefficients, and the (b) single scattering albedo measured between 30 April and 7 May ,528

12 Figure 16. Aerosol hygroscopicity measurements from 30 April through 7 May (a) The weighted average hygroscopic growth factors for three individual particle sizes (i.e., 20, 100, and 400 nm). The dates marked correspond to the midnight of the local time (CST). (b) The hygroscopic growth factor distributions for 20, 100, and 400 nm particles. The size bins are in intervals of Hygroscopicity [36] The measured hygroscopic growth factors range from 0.9 to 2.0 for six different particle sizes (i.e., 15, 20, 50, 100, 200, and 400 nm). Similar to the effective density measurements shown in Figure 9, the weighted average for each hygroscopic distribution is calculated. Three representative particle sizes (20, 100, and 400 nm) of the six observed particle sizes are shown in Figure 16a, along with the average hygroscopicity of all six particle sizes. Figure 16b displays the frequency of the hygroscopic growth factor by 0.10 intervals for the three selected particle sizes. Figure 17 shows the hygroscopic growth factor distributions averaged over the northerly and southerly winds. Since the H-TDMA system operates in multiple configurations during the field campaign, the ambient hygroscopicity measurements are sporadic to resolve an appropriate temporal trend. [37] A hygroscopic growth factor of unity indicates no change in the particle size after exposure to 90% relative humidity (RH), a value less than 1 indicates a decrease in mobility particle size, and a value greater than 1 indicates an increase in size. Pure ammonium sulfate has a hygroscopic growth factor of 1.70 at 90% RH [Wise et al., 2003]. Organics exhibit lower growth factors, ranging from 1.08 to 1.17 at 90% RH, although there is still a degree of uncertainty [Meier et al., 2009]. While freshly emitted BC particles are often initially hydrophobic, partially aged BC particles may absorb water and restructure upon humidification, resulting in HGF below unity [Khalizov et al., 2009b]. Thus, the hygroscopicity measurements provide additional information on the chemical composition and mixing state of aerosol particles. [38] Figures 16 and 17a show that 20 nm particles are less hygroscopic than larger particles. The primary peak between 1.00 and 1.20 accounts for 75% of all 20 nm particles, and the secondary maximum near 1.30 accounts for the remaining 25%. The two distinct peaks of low and high HGF most likely represent the organics ( ) and an organic-sulfate mixture ( ). The average HGF values of 20 nm particles are not appreciably distinct between northerly Hygroscopicity Distributions A 20nm B 100nm C 400nm Northerly Southerly Hygroscopicity Growth Factor (HGF) Figure 17. Averaged aerosol hygroscopic growth factor distributions during the two meteorological periods for the three particle sizes (a) 20 nm particles, (b) 100 nm particles, and (c) 400 nm particles. The two distributions representing hygroscopicity measured during southerly (black) and northerly winds (maroon). Northerly winds occurred between 15 h on 3 May and 9 h on 4 May. 10,529