Apportioned primary and secondary organic aerosol during pollution events of DISCOVER-AQ Houston

Abstract

Understanding the drivers for high ozone (O₃) and atmospheric particulate matter (PM) concentrations is a pressing issue in urban air quality, as this understanding informs decisions for control and mitigation of these key pollutants. The Houston, TX metropolitan area is an ideal location for studying the intersection between O₃ and atmospheric secondary organic carbon (SOC) production due to the diversity of source types (urban, industrial, and biogenic) and the on- and off-shore cycling of air masses over Galveston Bay, TX. Detailed characterization of filter-based samples collected during Deriving Information on Surface Conditions from Column and VERtically Resolved Observations Relevant to Air Quality (DISCOVER-AQ) Houston field experiment in September 2013 were used to investigate sources and composition of organic carbon (OC) and potential relationships between daily maximum 8 h average O₃ and PM. The current study employed a novel combination of chemical mass balance modeling defining primary (i.e. POC) versus secondary (i.e. SOC) organic carbon and radiocarbon (¹⁴C) for apportionment of contemporary and fossil carbon. The apportioned sources include contemporary POC (biomass burning [BB], vegetative detritus), fossil POC (motor vehicle exhaust), biogenic SOC and fossil SOC. The filter-based results were then compared with real-time measurements by aerosol mass spectrometry. With these methods, a consistent urban background of contemporary carbon and motor vehicle exhaust was observed in the Houston metropolitan area. Real-time and filter-based characterization both showed that carbonaceous aerosols in Houston was highly impacted by SOC or oxidized OC, with much higher contributions from biogenic than fossil sources. However, fossil SOC concentration and fractional contribution had a stronger correlation with daily maximum 8 h average O₃, peaking during high PM and O₃ events. The results indicate that point source emissions processed by on- and off-shore wind cycles likely contribute to peak events for both PM and O₃ in the greater Houston metropolitan area.

1. Introduction

The Houston metropolitan area has added nearly 1 million residents in the last decade (U.S. Census Bureau, 2019). Due to this rapid growth and expansion, traffic remains a major challenge for the city with an estimated 278 million km driven per day by the metropolitan area residents (Lubertino, 2019). Aside from urban emissions associated with residents (motor vehicles, cooking, etc.), the Houston metropolitan area also has strong influences from industrial and biogenic emissions (Bean et al., 2016; Dunker et al., 2019; Wallace et al., 2018). The Houston Ship Channel (HSC) is lined with dense zones of industrial facilities, including a petrochemical complex that is the largest in the U.S and second largest in the world (Port Houston, 2019). The Port of Houston, one of the busiest U.S. seaports (AAPA, 2019), contributes ship and heavy-duty diesel emissions to the city’s atmosphere (Schulze et al., 2018; Wallace et al., 2018). The Houston metropolitan area, much like other southeastern U.S. cities, is also highly vegetated with large forested regions north of the metropolitan area (Fig. 1); approximately 18.4% of land in the city of Houston is covered by tree canopy (Nowak et al., 2017).

Fig. 1.

Ground-based sampling sites: Moody Tower and Manvel Croix.

Air quality in Houston is a result of population growth, large-scale industrial areas, and vegetation. Historically, the Houston metropolitan area has been in nonattainment for ozone (O₃), based on the National Ambient Air Quality Standards from the U.S. Environmental Protection Agency (EPA, 2019). While particulate matter (PM) concentrations have not exceeded these national standards, a TCEQ monitoring site located between Houston’s urban core and the HSC has measured PM levels close to national standards: annual average fine PM (PM_2.5; PM with an aerodynamic diameter less than 2.5 μm) concentration ranging from 11 to 16 μg m⁻³ from 2005 to 2014 (H-GAC, 2015Allen and Fraser, 2006; Sullivan et al., 2013). To better understand the sources and atmospheric chemistry driving O₃ and PM in Houston, there have been several large-scale field campaigns starting with The Texas Air Quality Study (TexAQS) in 2000 which identified the importance of highly reactive volatile organic compounds (VOC) to O₃ production and PM (Allen et al., 2018; Kleinman et al., 2002; Ryerson et al., 2003). Following these studies, the State Implementation Plan (Allen et al., 2018) enacted strategies aimed at reducing nitrogen oxides (NO_x) and highly reactive VOCs, where then the follow-up campaign, TexAQS II, in 2005–2006, observed significant reduction in highly reactive VOCs and slower O₃ production rates (Parrish et al., 2009; Zhou et al., 2014). A companion study of TexAQS, the Gulf Coast Aerosol Research and Characterization was a longer-term (2000–2001) field experiment with the objective of identifying spatial and diurnal trends and key processes of PM_2.5 formation in southeastern Texas. The major findings from this campaign relevant to the current study include 1) the importance of secondary organic aerosol (SOA) to PM_2.5; 2) the need for more study of the impact of point sources on PM_2.5; and 3) an assessment that there was a lack of spatial variability in the PM concentration (Allen and Fraser, 2006). The objective of the Study of Houston Atmospheric Radical Precursors in 2009 was to better understand the influence of radical precursors, formaldehyde, and nitrous acid to O₃ production in Houston’s urban and industrial areas (Olaguer et al., 2014). These campaigns were vital steps in reducing emissions and improving air quality in the Houston metropolitan area. However, due to continuous population growth, improvements in control technologies, and the changing landscape of the metropolitan area, extrapolating from previous air quality field studies and/or from studies specific to only the most urban/industrial parts of the city is problematic when wanting to most effectively combat current air quality issues and reduce exposure to air pollution for all Houston metropolitan area residents.

The Deriving Information on Surface Conditions from Column and VERtically Resolved Observations Relevant to Air Quality (DISCOVER-AQ) field campaign is the most recent multi-site, high-intensity field experiment in Houston. This field campaign, conducted in 2013, was part of a broader NASA project with the goal to better interpret and improve measurement of ground-level pollutant concentrations via satellite data (NASA, 2019). This manuscript contributes to the spatial assessment of ground-level pollutants during DISCOVER-AQ, with a focus on organic aerosol (OA) sources and their relationship with daily O₃ concentrations at representative sites in the urban core and suburb of the metropolitan area.

Published results from the DISCOVER-AQ field experiment have demonstrated the increasing contribution of OA and SOA to PM concentrations at sites across Houston. During the DISCOVER-AQ, mobile and stationary measurements indicated significant contribution of oxygenated OA (OOA) to PM across the Houston metropolitan area during this period (Bean et al., 2016; Leong et al., 2017). However, Leong et al. (2017) observed more primary sources and sulfate emissions in urban/industrial regions (Leong et al., 2017). The current manuscript describes OA and OC measurements during a peak O₃ event at a downtown and suburb site, which allows for analysis of the evolving interaction between O₃ and PM in the metropolitan area during a peak event and additional characterization of the sources driving SOA production.

A peak O₃ event on Sept. 25–26, 2013 during DISCOVER-AQ provided an opportunity to study the conditions and emissions associated with this pollution regime in Houston (Baier et al., 2015). The production of O₃, especially in Houston, is known to be impacted by sea breeze re-circulation patterns (Tucker et al., 2010; Wang et al., 2016), which was present during a peak O₃ event during DISCOVER-AQ. For O₃ production, specific regimes were determined for Houston during DISCOVER-AQ. From a previous DISCOVER-AQ study, O₃ production in most parts of the metropolitan area was found to be VOC-dependent in the morning and NO_x-dependent in the afternoon (Mazzuca et al., 2016). However, in the urban/industrial zones, O₃ production was VOC-dependent throughout the day (Mazzuca et al., 2016). On Sept. 25, high O₃ production within the Houston urban core occurred, followed by advection to outlying areas due to transport and bay-breeze recirculation (Baier et al., 2015). Mazzuca et al. (2016) determined that the enhanced O₃ concentration during this peak O₃ event was likely due to higher concentrations of precursor compounds and/or meteorological conditions (e.g. lower boundary layer, stagnant conditions, bay breezes) rather than a faster production rate as the O₃ production efficiency, a measure of O₃ production to rate of NO_x oxidation, was similar to the rest of the week. These findings stress the importance of understanding emission sources and atmospheric processing across the metropolitan area. The current study is investigating how these sources and conditions impacted aerosol composition and concentration.

Finally, Dunker et al. (2019) investigated emission impacts on O₃ and OA during DISCOVER-AQ using the Comprehensive Air Quality Model with Extensions (CAMx). Dunker et al. (2019) provided more specific apportionment of OA and O₃ at several sites, identifying point sources to be a major contribution (29 and 21% fraction contribution, respectively) to both pollutants. The results from the current manuscript can also validate whether modeled changes in sources of OA match field measurements of OA components.

The objective of the current study was to quantify contributions of major sources and evaluate the trends and spatial distribution of daytime total particulate carbon (TC; organic plus elemental carbon; OC + EC) and O₃ in the Houston metropolitan area. During the DISCOVER-AQ campaign in September 2013, filter-based daytime samples were collected at two sites. Detailed chemical composition including molecular and isotopic analysis was performed on these filter samples to characterize carbonaceous aerosols at a downtown and suburb site. A molecular marker-based chemical mass balance (CMB) model was utilized in combination with ¹⁴C analysis to characterize TC and OC, respectively. Comparison with co-located high-resolution time-of-flight aerosol mass spectrometer (HR-ToF-AMS) at the suburb site further confirmed contribution of oxidized aerosol contribution. Particular focus was given to identifying major sources and drivers for PM formation during periods of peak PM and O₃. Day samples were chosen for detailed analysis, as daytime periods are most affected by O₃ and photochemical processes. In all, this analysis improves our understanding of sources and transport of PM across the metropolitan area and its connection to peak O₃ events.

2. Materials and methods

2.1. Sampling sites

PM_2.5 samples were taken in conjunction with NASA’s DISCOVER-AQ campaign at Moody Tower (MT; 29.7197, −95.3432) and Manvel Croix (MC; 29.5205, −95.3925) representing a downtown and suburb region, respectively, in the Houston metropolitan area (Fig. 1). MT is 4.6 km south of downtown Houston and located on top of a high-rise building at approximately 70 m above ground level (AGL). This site has been a focus for many air quality and meteorological studies for the Houston area (Czader et al., 2015; Lefer et al., 2010; Wong et al., 2012) as the site is representative of Houston’s urban atmosphere. MC is in a suburb 22.5 km south of MT located in a park near a residential area (Fig. 1).

2.2. Sampling

2.2.1. Filter-based sampling

PM_2.5 samples were collected on 90- and 102-mm-diameter quartz fiber filters (Pall Corporation, Port Washington, NY, USA) using medium-volume (90 L min⁻¹; URG Corporation, Chapel Hill, North Carolina, USA) and high-volume (Tisch Environmental, Cleves, OH, USA) samplers, respectively. For the high-volume samplers, the flow rate was 200 L min⁻¹ and 225 L min⁻¹ at MT and MC, respectively. All samplers were calibrated prior to field deployment; differences in flow rate were due to small differences in sampler design. The quartz fiber filters used for samples and field blanks were pre-cleaned by baking at 550 °C for 12 h in individual aluminum foil packets. Filters were stored in − 10 °C freezers pre- and post-sampling. A total of 14 field blanks were obtained from the three samplers: seven blanks from the high-volume at MT, three blanks from the medium-volume at MT, and four blanks from the high-volume at MC. Filter blanks were handled in the same manner as the ambient samples including filter preparation, travel (e.g. transport to Houston and to the sites), and storage (e.g. long-term laboratory and short-term onsite). For collection of the field blanks, the pre-cleaned filters were placed into sample holders and then into the samplers without turning them on.

Filter samples were collected from Sept. 4–28, 2013. A flexible sampling schedule was used for filter collection which included morning (06:30–10:00), afternoon (10:00–20:00), day (06:30–20:00), night (20:00 to 6:30), and 24-h (06:00–05:30) samples. This flexible schedule was implemented to allow intensive sampling periods within the month and to separate sources and chemistry that are time-of-day dependent. All samples were analyzed for bulk carbon (i.e. OC and EC) while select samples collected during Sept. 8–15 (week 2) and Sept. 21–28 (week 4) were further analyzed for chemical and carbon isotope measurements. These select samples included day samples from MT and MC, and a few 24-h MT samples (Sept. 8, 11, 12, and 14). MT 24-h teflon filters were also collected for gravimetric analysis of PM_2.5.

2.2.2. Real-time measurements in mobile air quality laboratory (MAQL)

Mass spectral data of non-refractory submicron PM (PM₁) was collected using a HR-ToF-AMS (Aerodyne Research, Inc., Billerica, MA, USA). The instrument was housed in the University of Houston’s MAQL. Non-refractory PM₁ was sampled from a dried (<40% relative humidity) aerosol inlet, which was raised approximately 5.5 m AGL when the MAQL was stationary. Measurements made during stationary sampling at MC during week 4 was included in the current study. Further details of CO and PM₁ sampling in the MAQL, measurements methods, and detection limits and uncertainties have previously been published elsewhere (Leong et al., 2017). The PMF protocol applied to the HR-ToF-AMS data is outlined in the Supplementary Data (S1).

2.2.3. Additional measurements at TCEQ monitoring sites

The current study utilized measurements of daily maximum 8-h average O₃ and daily average PM_2.5 concentrations from Texas Commission on Environmental Quality (TCEQ) Continuous Air Monitoring Stations across the Houston-Galveston-Brazoria (HGB) regions including Park Place (C416), Galveston (C1034), Seabrook (C45), Deer Park (C35), Houston East (C1), Clinton (C403), UH Moody Tower (C695), Manvel Croix (C84) and Conroe (C78). Daily maximum 8-h average O₃ concentrations, which will be referred simply as “O₃ concentration”, (Sept. 1–29, 2013) were retrieved from the Texas Air Monitoring Information System web interface (TCEQ, 2018). Daily average PM_2.5 concentrations (Sept. 1–29, 2013) were measured using a tapered element oscillating microbalance and were received from TCEQ upon request (personal communication with Jim Price, TCEQ). Wind speed and wind direction data for MT and MC sites were also retrieved from the Texas Air Monitoring Information System web interface (TCEQ, 2018). Linear regression analysis of O₃ to OC, source apportioned TC or PM_2.5 concentration were made to better identify trends between O₃ and aerosol. Results and relevant parameters are included either in Fig. 9, S3 and S4.

Fig. 9.

Correlation between daily maximum 8 h O₃ average and a) apportioned contemporary and fossil TC from ¹⁴C analysis and b) biogenic and fossil SOC. The slope, intercept and p-value are listed, respectively: Contemporary TC vs. O₃ (0.036, 0.84, 0.004), Fossil TC vs. O₃ (0.033, − 0.01, 0.0004), Biogenic SOC vs. O₃ (0.035, 0.64, 0.001), Fossil SOC vs. O₃ (0.022, − 0.49, 0.0002).

2.3. Organic and elemental carbon

All collected filters were analyzed for OC and EC by a thermal optical transmission instrument (Lab OC-EC Analyzer; Sunset Laboratory Inc., Tigard, OR, USA) using the NIOSH 5040 method (Birch and Cary, 1996). Triplicate analysis was completed every tenth sample run with an average relative standard deviation of 3.03%. Samples were blank corrected using site- and sampler-specific filter blanks. Average blank subtraction was 14.4 ± 11.6% of OC concentration. There was no EC contribution in the blanks.

Day and night OC and EC measurements are reported here. When a daytime filter was not available, OC and EC masses from morning (06:30–10:00) and afternoon (10:00–20:00) samples were summed and volume normalized to represent daytime measurements. When a nighttime sample was not available, daytime masses (06:30–20:00) were subtracted from 24-h masses (06:00–05:30) and then volume normalized to represent nighttime measurements.

2.4. Detailed analyses

2.4.1. Radiocarbon analysis

The filter-based TC (i.e. OC + EC) was analyzed for ¹⁴C abundance as described previously (Barrett et al., 2015). Briefly, a fraction of each filter corresponding to ~100 μg of TC was acid fumigated in a desiccator over 1 N hydrochloric acid and then dried at 60 °C for 1 h. The samples were shipped on ice to the National Sciences Accelerator Mass Spectrometry facility in Woods Hole Oceanographic Institution (Woods Hole, MA, USA) for ¹⁴C analysis. Field and laboratory blanks were handled and analyzed in the same manner to allow for blank correction.

Once received by the facility, these filter samples were analyzed using accelerator mass spectrometry to determine the fraction of modern carbon (F_M). F_M is the ¹⁴C–¹²C ratio of the sample to a “Modern” reference sample (i.e. NBS Oxalic Acid I).

$$F_M=\frac{{\left({}^{14}C/{}^{12}C\right)}_{sample}}{{\left({}^{14}C/{}^{12}C\right)}_{A{D}_{1950}}}$$ (1)

The F_M was blank corrected using field blanks collected during the sampling campaign. The averaged blank concentration of 1.12 ± 0.50 μg C cm⁻² had a F_M value of 0.38 ± 0.01.

2.4.2. Organic tracer analysis

The MT samples from week 2 and MT and MC samples from week 4 were analyzed for organic tracers by solvent extraction followed by analysis on a gas chromatograph – mass spectrometer (GC-MS). These samples were analyzed for a suite of non-polar combustion (alkanes, PAHs, hopanes and steranes), polar combustion (levoglucosan), and secondary biogenic (2-methyltetrols, pinic and pinonic acids) organic tracer compounds. An adapted version of a previously published pressurized liquid extraction method was used for tracer analysis (Clark et al., 2015). For this method, a portion of each filter sample representing ~400 μg of OC was utilized. Filters were spiked with a known amount of isotopically-labelled internal standards (tetracosane-d⁵⁰, triacontane-d⁵⁸, dotriacontane-d⁶⁶, fluoranthene-d¹⁰, pyrene-d¹⁰, benz(a) anthracene-d¹², chrysene-d¹², benzo(b)fluoranthene-d¹², benzo(e)pyrene-d¹², benzo(k)fluoranthene-d¹², indeno (1,2,3-cd)pyrene-d¹⁴, dibenz (ah)anthracene-d¹⁴, coronene-d¹², cholestane-d⁴, levoglucosan-C¹³). Filters were extracted using an accelerated solvent extractor (Dionex 350, Sunnyvale, CA, USA) with a solvent mixture of 2:1 v/v dichloromethane and acetone. The extracts were concentrated to ~0.5 mL using a Caliper TurboVap II (Hopkinton, MA, USA). The extracts were then concentrated to a total volume of 125 μL using a gentle nitrogen stream and analyzed using a GC-MS with an electron ionization source (Agilent Technologies, Santa Clara, CA, USA). The organic tracers were quantified against prepared six-to seven-point external calibration solutions. A check standard, generally the mid-point of the dilution standard series, was run before and after each sample batch for quality control on the calibration curve.

To analyze levoglucosan, a polar compound, a 25 μL aliquot of each sample extract was blown down to dryness using a gentle nitrogen stream. The samples were then reconstituted using 25 μL of pyridine (Fisher Chemical, Waltham, MA, USA) with 50 μL of N,O-bis(trimethylsilyl)trifluoroacetamide trimethylchlorosilane (BSTFA-TMCS; Sigma Aldrich, St. Louis, MO, USA) (Simoneit et al., 1999). Within 24 h, the derivatized samples were run on the GC-MS with a six-point calibration curve and check standards as mentioned before.

2.5. Source apportionment tools for filter-based measurements

2.5.1. Radiocarbon apportionment

The Δ¹⁴C value is the relative difference between the ¹⁴C measurement of the sample and standard reference material, corrected to account for decay that took place between collection and time of measurement. Δ¹⁴C is calculated using the reported F_m, inverse of radiocarbon’s half-life (λ), and the year of the sample collected (Y_c):

$${\mathrm{\Delta }}^{14}C=\left[F_m\ast e^{\mathrm{\lambda }\left(1950-Y_c\right)}-1\right]\ast 1000$$ (2)

The Δ¹⁴C can then be used to apportion fractional contribution from fossil and contemporary sources in atmospheric PM (Hildemann et al., 1994). With known end-members for contemporary and fossil sources, the contribution of contemporary, f_cont, versus fossil sources, 1- f_cont, can be determined using a mass balance approach:

$${\Delta }^{14}{C}_{sample}=\left({\Delta }^{14}{C}_{cont}\right)\left(f_{cont}\right)+\left({\Delta }^{14}{C}_{fossil}\right)\left(1-f_{cont}\right)$$ (3)

End-member values of +67.5‰ and − 1000‰ were used to represent contemporary and fossil sources, respectively. The contemporary end-member is an averaged value of +107.5‰ to represent wood burning and +28‰ to represent emissions from annual growth (Zotter et al., 2014). A final combined uncertainty u_c, was calculated for each sample measurement (equation (4)) which included instrument standard error, σ_ins, relative difference of F_M blank correction d_{r blk}, and the standard deviation of using either of the two contemporary end-members (i.e. +28‰ and +107‰) for calculating the Δ¹⁴C_sample, σ_endmem. The σ_endmem accounted for the uncertainty when using an average (i.e. +67.5‰) of the two contemporary end-members.

$$u_c=\sqrt{{\sigma }_{ins}^2+d_{rblk}^2+{\sigma }_{endmem}^2}$$ (4)

2.5.2. Chemical mass balance

The CMB model is a receptor-based source apportionment model, which uses measured species concentration and known source profiles to calculate contributions of each source for each sample (Schauer et al., 1996). The model assumes that no change occurs in the emissions from source to receptor. However, this is an over-simplification, as oxidation and partitioning between gas- and particle-phases leads to changes in the plume during transport. For the purposes of the current study, the primary emissions apportioned by the CMB are assumed to be representative of the remaining primary plume which was collected on the filter during sampling. The current study used the EPA CMBv8.2 model (Coulter, 2004). The CMB source profiles, f_ij, included primary sources from vegetative detritus (Rogge et al., 1993), EPA Region 4 woodsmoke (Fine et al., 2002; Sheesley et al., 2007), diesel-powered motor vehicle exhaust, gasoline-powered motor vehicle exhaust, and lubricating oil-impacted motor vehicle exhaust (Lough et al., 2007). The motor vehicle sources are combined and presented as “motor vehicle exhaust” (MVE). The CMB model outputs were accepted if the χ-square < 4 and r² > 0.80. For all samples, the average χ-square and r² values are 1.97 ± 0.88 and 0.90 ± 0.04, respectively.

2.5.3. Combined ¹⁴C and CMB source apportionment of TC

The ¹⁴C source apportionment distinguished contribution of fossil and contemporary carbon of TC. The CMB source apportionment distinguished contribution from major primary sources from fossil (i.e. MVE) and contemporary (i.e. wood smoke and vegetative detritus) OC. The EC to OC ratios of each source profile was used to calculate the EC contribution for each primary source (Lough et al., 2007; Sheesley et al., 2007). The primary fossil and contemporary carbon concentration from the CMB analysis was subtracted from the respective ¹⁴C apportioned fossil and contemporary TC to calculate an estimate of secondary fossil and secondary contemporary carbon.

2.6. HYSPLIT back trajectory (BT) analysis

The BT calculations were performed for the sampling sites using the NOAA Hybrid Single Particle Lagrangian Integrated Trajectory (HYS-PLIT) model, v4, May 2012 release (Draxler and Rolph, 2010). The BTs were produced every 6-h starting at 00:00, 06:00, 12:00, and 18:00 each day from Sept. 4–28 during the sampling period. Individual BTs were run with an initial height of 0 m and 80 m AGL for MC and MT, respectively, using GDAS meteorological dataset. To understand the transport pathway and major sources of air mass impacting each site, the resulting BTs were clustered into groups with similar transport patterns using the clustering function. The predominant cluster for each day was assigned (Table 1). Final clustered BTs were plotted using ESRI’s ArcGIS 10.1 mapping software (Fig. 2). To model potential uncertainty in the HYSPLIT model BTs during DISCOVER-AQ, a sensitivity study was performed for MT at 10 km north and south of the site. MT was plotted at 0, 80, and 100 m AGL. The sensitivity study results show that the HYSPLIT model is robust and reproducible producing relatively similar BT pathways even when varying the locations and heights from the initial sampling site. The cluster analysis was run for the 0 m AGL and the 80 m AGL BTs separately.

Table 1.

September 2013 Daily Back Trajectory Cluster. The daily cluster was assigned based on the predominant 6-h BT cluster for each day.

	Moody Tower	Manvel Croix
Day	Daily Cluster	Daily Cluster
4-Sept.	ESE	S
5-Sept.	ESE	ESE
6-Sept.	ESE	ESE
7-Sept.	ESE	ESE
8-Sept.	ESE	ESE
9-Sept.	ESE	ESE
10-Sept.	ESE - SE	ESE
11-Sept.	ESE - SE	ESE
12-Sept.	ESE	ESE
13-Sept.	ALL - 3	ESE
14-Sept.	ESE	ESE
15-Sept.	ESE	ESE
16-Sept.	SE	ESE
17-Sept.	ESE - SE	ESE
18-Sept.	ESE	ESE
19-Sept.	SE	ESE
20-Sept.	SE	ESE
21-Sept.	NNE	All - 3
22-Sept.	NNE	NNE
23-Sept.	ESE	NNE
24-Sept.	ESE	NNE
25-Sept.	NNE	NNE - S
26-Sept.	ESE	S
27-Sept.	ESE	ESE
28-Sept.	SE	ESE

^*North-northeast (NNE), East-southeast (ESE), Southeast (SE), South (S).

Fig. 2.

HYSPLIT 12 h clustered back trajectories were performed to display percentages of air masses that traveled to the sampling sites during DISCOVER-AQ.

3. Results and discussion

Bulk carbon measurement was used as the initial characterization tool to identify trends of Houston’s carbonaceous aerosols during DISCOVER-AQ (Fig. 3). These trends were also observed in the PM_2.5 and O₃ measurement made at the TCEQ monitoring sites across the HGB area (Fig. 4). Based on these plots, two periods of interest were determined: week 2 (Sept. 8–15) with peaks in OC and PM and week 4 (Sept. 21–28) with peaks in OC, PM and O₃ (Figs. 3 and 4). The two periods were separated by a week of intermittent precipitation (Figs. 3 and 4).

Fig. 3.

Day (14 h) and night (10 h) PM_2.5 OC (dark gray) and EC (black) measurements at a) Moody Tower and b) Manvel Croix during full sampling campaign. Daily maximum 8 h average ozone concentration on right y-axis. Week 2 and Week 4 are highlighted in light gray.

Fig. 4.

TCEQ Monitoring Sites O₃ and PM_2.5 concentrations across the southeastern regions of Texas. The PM_2.5 measurements were 24 h (06:00–5:30) averages and O₃ data are daily maximum concentrations provided from Houston-network of Environmental Towers (http://www.hnet.uh.edu).

The two weeks of interest were chosen for comparison with co-located measurements and for more detailed chemical analysis, including organic tracer and ¹⁴C analysis. To better understand relationships between O₃ and PM in Houston, this detailed analysis focused on daytime samples at MT, the downtown Houston site, during both weeks 2 and 4. The detailed analysis of MC samples during week 4 were also daytime samples. Discussion of the results of the detailed analysis will follow the overview of the bulk carbon measurements.

3.1. Trends in carbonaceous aerosol and O₃ during DISCOVER-AQ

Average ambient concentration of daytime TC during the full sampling period at MT and MC was 3.74 ± 1.65 and 3.36 ± 1.29, μg C m⁻³, respectively. Of these samples, fractional contribution of OC at MT and MC was 0.87 ± 0.06 and 0.89 ± 0.04, respectively, while EC was 0.13 ± 0.06 and 0.11 ± 0.04, respectively. The OC and EC concentrations were highest at MT, the most urban site (Fig. 3a and b). Ambient OC concentrations varied during the sampling periods (Fig. 3a), where OC was significantly enhanced during the high pollution weeks (weeks 2 and 4) compared to the rest of the sampling days (Mann-Whitney Rank Sum test; p < 0.001). Ambient EC concentrations observed similar trends as the OC (e.g. peak EC concentration on Sept. 24 night) (Fig. 3a), however, the EC was not significantly enhanced during the high pollution weeks (Mann-Whitney Rank Sum test; p < 0.05). Although both weeks had building concentrations of OC, the peak concentration was higher for week 4 at both MT and MC (Fig. 3). For example, the OC concentration at MT dramatically peaked on Sept. 24 night (9.5 ± 1.1 μg C m) followed by Sept. 25 day (6.9 ± 0.7 μg C m⁻³) when a peak O₃ event had occurred (Fig. 3a). The EC concentration was comparable during both weeks 2 and 4 at both MT (avg: 0.50 ± 0.16 and 0.34 ± 0.19 μg C m⁻³, respectively) and MC (average: 0.35 ± 0.09 and 0.29 ± 0.15 μg C m⁻³, respectively).

Carbonaceous aerosols are a significant fraction of PM_2.5 based on previous Houston studies (Allen and Fraser, 2006; Fraser et al., 2002). As observed by the TCEQ monitoring sites, the PM_2.5 concentrations (Fig. 4) and measured bulk carbon concentrations (Fig. 3) had very similar trends. TC was an important component of the increasing PM_2.5 concentrations during both weeks (excluding Sept. 15 when PM_2.5 peaks while TC concentration drops). Over the full sampling campaign, TC concentrations at MT and MC ranged between 0.62 and 10.44 μg C m⁻³ and 1.67–6.34 μg C m⁻³, respectively. The campaign mean for the contribution of TC to PM_2.5 was 43%. There is a relatively consistent intercept of 3.3–3.6 μg m⁻³ (Figure S4) for the PM_2.5 vs OC relationship at MT. This indicates that even in the absence of OC, there are significant inorganic contributions to PM_2.5 (e.g. sulfate, ammonium, nitrate).

The trends and concentration of O₃ were relatively consistent throughout the Houston metropolitan area based on the O₃ daily maximum concentrations measured at the TCEQ monitoring sites (Fig. 4). The average O₃ concentration during the full campaign at MT and MC were 41 ± 15 and ±14 ppbv, respectively. The highest O₃ concentration was on Sept. 25 during week 4 with concentrations of 83 and 82 ppbv at MT and MC, respectively (Fig. 3). The lowest O₃ concentration was on Sept. 20 during the intermittent precipitation period between weeks 2 and 4 with concentrations of 16 and 20 ppbv at MT and MC, respectively. The O₃ concentrations were well correlated with OC at MT and MC (r²: 0.59 and 0.44, respectively).

3.2. Sources of air masses based on BTs, wind direction and speed

The clustered BTs and wind rose plots were used to evaluate air mass transport to MT and MC during the sampling period. The BTs were clustered into five groups: north-northeast, east-southeast, southeast, northeast, and south (Table 1 and Fig. 2). The BT clusters at MT and MC were mainly from onshore east-southeast (57–68%) and offshore north-northeast (15–21%) winds (Table 1). Earlier in the campaign, winds were generally from the east-southeast direction at MT and MC (Table 1, Figure S1). A shift in wind direction occurred during week 4 where a combination of northern continental (e.g. north-north east) and onshore marine (e.g. east-southeast) air masses were observed (Table 1, Figure S1). During week 4, lowest wind speeds were observed on Sept. 25 at both MT and MC with average speeds of 3.8 ± 2.9 and 3.2 ± 1.6 miles/h, respectively, relative to average speeds of 8.2 ± 2.1 and 7.0 ± 1.4 miles/h, respectively, for the full campaign.

In or near coastal regions, shifting wind directions and lower wind speeds during the daytime is indicative of converging onshore and offshore winds due to the sea/bay breeze phenomenon (Banta et al., 2005, 2011; Caicedo et al., 2019; Stauffer and Thompson, 2015). During these periods, stagnant atmospheric conditions are produced allowing buildup of O₃ and pollutant concentrations (Banta et al., 2005; Loughner et al., 2011; Stauffer and Thompson, 2015); this was the case for the Houston metropolitan area on Sept. 25, 2013 (Baier et al., 2015; Caicedo et al., 2019). The lower wind speeds on Sept. 25 were accompanied by shifting wind directions during the day (Fig. 5b). In comparison to Sept. 9, which had relatively lower O₃ levels, the wind direction was consistently from east-southeast with stronger wind speeds during the day (Fig. 5a). The progression of change in the wind conditions during week 4 at MT followed observed pollutant trends (i.e. OC, PM_2.5, and O₃ concentrations) where the week began with consistent wind direction/high wind speeds then transitioned to varying wind direction/low wind speeds (i.e. stagnant conditions on Sept. 25) and then back (Figure S2). The sea/bay breeze on Sept. 25 caused recirculation of air masses transporting pollutants from the HSC out to the Gulf and then back to Houston; this circulation pattern was considered an important factor in reaching peak O₃ concentrations during this period (Baier et al., 2015; Li et al., 2016). Additionally, on Sept. 25, measurement and analysis of direct O₃ production rates at MT and at Smith Point (mouth of the HSC) indicated high O₃ production around MT with advection to Smith Point and outlying areas in Houston (Baier et al., 2015). The recirculation pattern coupled with the high O₃ production makes Sept. 24–25 a time of high interest in understanding the major influences on aerosol formation across the Houston Metropolitan area.

Fig. 5.

Hourly averaged wind speed and wind direction on a) Sept. 9, 2013 and b) Sept. 25, 2013 at Moody Tower (MT) and Manvel Croix (MC).

3.3. Detailed analysis of two high PM weeks

To better characterize sources and potential atmospheric chemistry driving the buildup of PM in weeks 2 and 4 during DISCOVER-AQ, chemical and source apportionment analysis was conducted. Results of this analysis were compared to co-located measurements of O₃ and real-time OA by HR-ToF-AMS. The ¹⁴C-based analysis apportions TC either as contemporary or fossil carbon, while molecular marker analysis indicates specific emission sources for the two weeks with building periods of PM during DISCOVER-AQ. Carbonaceous aerosols at both sites had larger contributions of contemporary than of fossil carbon. The MT measurements provide a comparison between weeks 2 and 4 to understand drivers of high PM in the Houston urban core. Analysis of week 4 provided comparisons of sources at both sites during a period of high PM.

3.3.1. Week-to-week comparison of contemporary and fossil carbon at Moody Tower

The detailed analysis of MT provided comparison of apportioned fossil and contemporary TC during a week with high PM (week 2) and a week with high PM and a peak O₃ event (week 4). During week 2, contribution of contemporary carbon was relatively consistent until Sept. 14, when an increase in contemporary carbon contribution was observed (Fig. 6). During week 4, contribution of contemporary carbon had strong day-to-day variability with a sharp decrease in contemporary contribution on Sept. 25, the day of peak OC and O₃ concentrations (Figs. 4 and 6). A closer look at ambient concentrations of contemporary and fossil TC for week 4 reveals that the apparent decrease in contribution of contemporary TC was driven by an increase in the ambient fossil TC concentration (Fig. 6). Contribution of fossil carbon at MT for week 2 (average 45 ± 8%) and week 4 (average: 39 ± 10%) were not statistically different (Mann-Whitney Rank Sum test; p > 0.50).

Fig. 6.

Contemporary and fossil total carbon (TC) concentration from ¹⁴C-based source apportionment (left y-axis). Fraction contemporary contribution is included (right y-axis). *These are of 24 h (06:30–06:00) samples.

3.3.2. Comparison of contemporary and fossil carbon by site

Based on the ¹⁴C-based source apportionment, contemporary sources contributed a majority of the TC at both sites. During week 4, MT had smaller contribution and concentration of contemporary carbon than MC. Contemporary carbon contribution at MT ranged from 48 to 78% with an average of 61 ± 10%, while MC ranged from 60 to 86% with an average of 72 ± 9%. The average contemporary and fossil carbon concentration at MT was of 2.65 ± 1.01 and 1.73 ± 0.89 μg C m⁻³, respectively, while MC was 2.72 ± 0.99 and 1.11 ± 0.66 μg C m⁻³, respectively. Contribution from fossil carbon was significantly larger at MT than MC (t-test; p > 0.05). These results are expected as MT is located at the urban core of Houston and is often impacted by more traffic and industrial emission sources than MC.

The apportioned contemporary carbon contribution from the current study can be compared to measurements made in southeastern U.S. cities from a previous decade, though it is important to note differences due to improved emissions control technology and geography. The contribution of contemporary carbon from DISCOVER-AQ are similar to Centreville, AL (85%), Nashville, TN (56–80%), and Tampa, FL (52–89%), but have a higher contribution than Birmingham, AL (37%) and Aldine, TX (another suburban city in the Houston metropolitan area: 34–68%) (Lemire et al., 2002; Lewis et al., 2004; Lewis and Stiles, 2006; Weber et al., 2007; Zheng et al., 2006). Studies have observed enhanced biogenic SOA production, particularly isoprene-derived SOA under the influence of urban air masses (Edney et al., 2005; Gunsch et al., 2018; Shilling et al., 2013; Surratt et al., 2010). Monoterpenes, with the influence of anthropogenic emissions, are also a significant contributor to summer SOA production in the southeastern U.S. region (Zhang et al., 2018).

3.3.3. Molecular marker source apportionment modeling

The CMB model is a receptor-based model which was used to apportion carbonaceous aerosols to primary sources including vegetative detritus, wood smoke, and MVE (sum of diesel exhaust, gasoline exhaust and lubricating oil contribution). The ¹⁴C-based apportionment measurements was combined with the CMB results to calculate an estimate of SOC from secondary fossil and biogenic carbon precursors (Fig. 7). This combined apportionment method provides an estimate of SOC due to potential exclusion of other, likely minor, primary carbon sources. On average, the CMB apportioned 33 ± 14% of the TC with the larger contribution from primary sources at MT (38 ± 13%) than MC (21 ± 7%). When comparing weeks 2 and 4, the contributions of fossil and biogenic SOC were higher during week 4 when the peak O₃ event occurred (Fig. 7).

Fig. 7.

Apportionment of TC using the CMB and ¹⁴C analysis. Primary sources include vegetative detritus, wood smoke, and motor vehicle exhaust and secondary sources include fossil and biogenic precursors. Apportioned a) concentration and daily maximum 8-h average O₃ concentrations (right-axis) and b) fractional contribution. *These are of 24 h (06:30–06:00) samples.

3.3.3.1. Week-to-week comparison of Moody Tower by CMB + ¹⁴C.

Weeks 2 and 4 are of particular interest, as both weeks exhibited build up and peak concentrations of OC and peak EC concentration, with different atmospheric oxidation conditions. Week 2 at MT was significantly more impacted by primary sources (t-test; p < 0.001), with an average TC apportionment to primary sources of 48 ± 10% compared to 28 ± 6% in week 4. For both weeks, the majority of the apportioned primary TC at MT was from MVE, followed by wood smoke, then vegetative detritus. The average ambient concentration of MVE during the weeks 2 and 4 was 1.59 ± 0.50 and 1.02 ± 0.49 μg C m⁻³, respectively, while average ambient concentration of wood smoke was 0.28 ± 0.22 and 0.19 ± 0.19 μg C m⁻³, respectively. Although the relative contribution of MVE to total apportioned primary TC during each week was the same, the ambient concentration varied. The differences in ambient concentrations was likely more influenced by meteorological/environmental conditions than by large changes in local MVE emissions. This is in agreement with Glen et al. (1996), in which meteorological factors were a large driver of variability in MVE emissions in urban atmospheres.

Contribution of secondary biogenic and fossil carbon to TC were significantly larger (t-test; p < 0.05) during week 4 (56 ± 8% and 16 ± 10%, respectively) compared to the week 2 (48 ± 6% and 5 ± 6%, respectively). However, the ambient concentration between the two weeks for both secondary biogenic and fossil carbon at MT were not statistically different (t-test; p > 0.05), due to the high standard deviation; the average ambient concentration for secondary biogenic carbon was 2.00 ± 0.23 and 2.42 ± 0.70 μg C m⁻³, while secondary fossil carbon was 0.25 ± 0.37 and 0.70 ± 0.56 μg C m⁻³ for weeks 2 and 4, respectively. Secondary fossil carbon and O₃ concentrations peaked on Sept. 14 and Sept. 25, which were also peak days for OC and PM_2.5.

3.3.3.2. Comparison of the sites by CMB + ¹⁴C.

Direct comparison of the two sites is possible from the detailed chemical analysis and source apportionment of aerosol samples collected during the week 4 of the campaign. Both sites observed a building of fossil TC (i.e. primary MVE and secondary fossil) concentration with the increasing peak O₃ concentration (Fig. 7). EC was strongly correlated with MVE at MT (r² = 0.90; slope = 0.37; intercept = − 0.03) and at MC (r² = 0.95; slope = 0.42; intercept = 0.06). The MVE concentration at MT (average: 1.02 ± 0.49 μg C m⁻³) was not significantly larger than at MC (average: 0.55 ± 0.34 μg C m⁻³) (t-test; p < 0.05). The ambient concentration of wood smoke at MT and MC was 0.19 ± 0.19 and 0.22 ± 0.12 μg C m⁻³, respectively. The ambient concentration of vegetative detritus was 0.04 ± 0.02 and ±0.03 μg C m⁻³ at MT and MC, respectively. Both primary TC concentrations from wood smoke and vegetative detritus at MT and MC were minimal.

As mentioned previously, the CMB apportioned 31 ± 13% of the TC, with larger contribution from primary sources at MT (weeks 2 and 4; 38 ± 13%), than MC (21 ± 7%). These values are comparable to the results of Dunker et al. (2019) using the CAMx model, which showed the primary to total OA ratios at Park Place (C416; most similar to the MT site) 40 ± 9%. However, for MC the CAMx value (35 ± 8%) was larger than the CMB estimate for MC. Differences in the values can be due to the source apportionment of either TC (CMB + ¹⁴C) or OA (CAMx) where the TC accounts for just carbon mass while OA includes oxygen, hydrogen, nitrogen, and other elements associated with organic matter (OM). The OM/OC ratio measured previously range from 1.1 to 2.1 while contribution of secondary species are at the higher end of the range (El-Zanan et al., 2005; Russell, 2003). For sites more impacted by secondary processes (e.g. MC), the difference between apportioned TC and OA will likely be greater. Another potential reason for the difference in apportioned TC and OA can be due to the sampling period, as the CAMx modeled values were 24-h averages, while the measured values at MT and MC were averages during the daytime (6:30 to 20:00). The exclusion of the nighttime in the measured data at MC can also contribute to driving a lower primary to total TC compared to the modeled 24-h average primary to total OA.

For both sites, the largest contribution of TC was unapportioned, likely from SOC (average all sites: 67 ± 14%) or aged emissions (oxidized POC). With improvement in MVE technology, studies have observed a decrease of primary emissions (Gentner et al., 2017; Gordon et al., 2014). Secondary processes of carbonaceous aerosols previously have been demonstrated to be important during DISCOVER-AQ 2013 Houston studies (Bean et al., 2016; Dunker et al., 2019; Leong et al., 2017). The largest fraction of TC at both sites was from secondary biogenic carbon, which could also include secondary or aged BB, with an average contribution of 65 ± 10% at MC and 52 ± 8% (week 2 and 4) at MT. During the week with a peak O₃ event (i.e. week 4), the average ambient concentration of secondary biogenic TC at both sites was 2.29 ± 0.86 μg C m⁻³ which is higher than other southeastern U.S. cities including Atlanta, GA (0.95 μg C m⁻³); Pensacola, FL (1.15 μg C m⁻³); and Birmingham (2.03 μg C m⁻³) and Centreville (2.04 μg C m⁻³), AL (Lewandowski et al., 2013). The biogenic SOC concentrations of these four southeastern cities were summer averages, while the current study reflects an average during a high pollution period for O₃ and PM. A previous study using molecular-tracer based source apportionment of biogenic SOC in Research Triangle Park, NC in 2003 included high particulate pollution days (Kleindienst et al., 2007). The September average (1.48 ± 0.27 μg C m⁻³) at Research Triangle Park was lower than the current study, while average biogenic SOC concentrations during August (2.75 ± 0.67 μg C m⁻³) was more similar to the average concentration at the Houston sites (2.42 ± 0.89 μg C m⁻³) (Kleindienst et al., 2007).

Contribution from secondary fossil sources was quite variable at both sites, ranging from 3 to 27% at MT and 9–18% at MC. The largest day-to-day variability and largest average contribution of secondary fossil carbon was observed at MT with an average contribution and concentration of 16 ± 10% and 0.70 ± 0.56 μg C m⁻³, respectively. Contribution of gas-phase SOA precursors from MVE emissions would contribute to the secondary fossil carbon in the Houston metropolitan area (Gentner et al., 2017). MT also is influenced by emissions from the heavily industrialized HSC (Bahreini et al., 2009; Wallace et al., 2018). Dechapanya et al. (2004), utilizing emission inventory data for 2000, estimated that 53% of projected anthropogenic SOA in the Houston area was from industrial precursors, i.e. aromatics and terpenes from pulp and paper processing. These industrially produced terpene emissions, if biologically derived, would be apportioned as contemporary carbon and/or biogenic SOC; while petroleum-derived terpenes would be apportioned to fossil SOC. Since the metropolitan area is greatly impacted by biogenic sources of terpenes from the piney woods just north east of the city (Fig. 1), it would take significant emissions to supplant this natural source (Nowak et al., 2017). Current studies of these point sources are needed to understand the impact of the industrially produced VOCs. Dunker et al. (2019) suggested that point sources could enhance both the yield of biogenic SOA and the production of O₃ in the HGB area via co-emission of NO_x. The current study observed peak secondary fossil and biogenic carbon concentrations during peak PM and O₃ days at MT. Thus, Houston industrial point sources may be the drivers for the observed peak in both SOC and O₃ concentrations during DISCOVER-AQ. The recirculation pattern was observed on Sept. 25 where air masses passed through HSC out to the Gulf and then back onshore. This circulation pattern would allow the air mass to pick up potential anthropogenic emissions including heavy duty diesel, industrial, and ship emissions (Schulze et al., 2018). While the more stagnant conditions during the daytime (i.e. lower wind speeds) (Figure S1b) allowed the emissions to build during periods of high photochemical activity ultimately contributing to increased levels of SOC and O₃.

3.3.3.3. Positive matrix factorization of PM₁ organic aerosols.

The HR-ToF-AMS operated in the MAQL with mobile measurements on-road transects near the MC site during DISCOVER-AQ. During week 4, the MAQL operated at or near MC, which provided opportunity to compare with the filter-based TC source apportionment. The PMF analysis of the HR-ToF-AMS data resolved four factors: biomass burning organic aerosol (BBOA), hydrocarbon-like organic aerosol (HOA), less oxidized OOA (LO-OOA), and more-oxidized (MO-OOA) (Fig. 8). Determination of these factors is detailed in S1. The HOA, LO-OOA and MO-OOA factors are largely distinguished by their oxidation state (Leong, 2015). The HOA had the lowest oxidation state of the three factors based on a low oxygen to carbon (O:C) but high hydrogen to carbon (H:C) ratio. This HOA is most representative of POA (Bean et al., 2016; Zhang and Ying, 2011), which would be similar to the MVE in the CMB analysis. The LO-OOA and MO-OOA have higher oxidation state than HOA (e.g. high O:C and low H:C ratios), which is more representative of SOA and can be compared to the SOC biogenic and SOC fossil from the CMB analysis (Zhang and Ying, 2011). The oxidized aerosol dominated the HR-ToF-AMS dataset; MO-OOA was the dominant constituent with 68% followed by the LO-OOA with 15% (Fig. 8a). The MO-OOA has the highest oxidation state of the factors, implicating highly aged and less volatile OA. Similarly, the SOC dominated the filter-based CMB+¹⁴C analysis: SOC biogenic was the dominate constituent with 71% followed by SOC fossil with 15%. Overall, the PMF factors compare well to the CMB+¹⁴C sources: biogenic and fossil SOC (combined 86%) closely matches the MO-OOA and LO-OOA (combined: 83%) (Fig. 8a and b) while HOA (9%) closely matches primary OC from MVE and vegetative detritus (combined 11%) (Fig. 8a and b). The CMB – wood smoke contribution at MC had an average contribution of 4%, which is comparable to contribution of BBOA (8%) (Fig. 8a and b). The CMB – wood smoke apportions the primary contribution of wood smoke based on the ratio of levoglucosan to OC. The PMF – BBOA includes OA (both primary and secondary) produced from biomass burning events. The difference in contribution and concentration of CMB – wood smoke and BBOA measurements are due to the difference in source apportionment methods including exclusion of possible SOC from wood smoke which would be apportioned as secondary biogenic contribution in the CMB + ¹⁴C analysis.

Fig. 8.

Apportioned a) OA of PM₁ and b) TC of PM_2.5 at Manvel Croix from September 21–23. PM₁ OA factors include hydrocarbon-like OA (HOA), biomass burning OA (BBOA), less-oxidized oxygenated OA (LO-OOA), and more-oxidized oxygenated OA (MO-OOA). PM_2.5 TC apportioned sources include vegetative detritus (Veg Det), motor vehicle exhaust (MVE), primary biomass burning (BB), secondary organic carbon (SOC) fossil and biogenic. SOC sources and oxidized aerosol sections are striped while primary organic carbon (MVE and Veg Det) and aerosol (HOA) sections are solid fill. Biomass burning sources (primary BB and BBOA) are both solid red. Below each pie chart includes average ambient concentration.

3.4. Carbonaceous aerosols and ozone

Ozone and PM_2.5 (including SOA) within the Houston metropolitan region is a result of a complex mix of emission sources. Although it has been demonstrated previously that O₃ and secondary PM_2.5 are interrelated in urban areas, it is not always apparent which emission sources drive peak days. This is relevant for developing effective multipollutant mitigation strategies that can reduce local community exposure to both O₃ and PM_2.5 (Fann et al., 2011; Liao and Hou, 2015; Sofia et al., 2020). Recent detailed analysis of the O₃ peak on Sept 25 (Mazzuca et al., 2016) demonstrated the impact of industrial emissions on this event. Similarly, recent air quality modeling analysis also reported the strong influence of industrial emissions on peak O₃ and PM_2.5 (Dunker et al., 2019). With the addition of the combined CMB+¹⁴C source apportionment, this study constrains the fossil vs contemporary (e.g. biogenic or biomass burning) contributions to TC and confirms that the industrial influence on O₃ also had significant contribution to the high PM_2.5. Though it has been previously demonstrated that O₃ and PM_2.5 are often correlated, this study allows for a source-based correlation analysis to define whether fossil or contemporary sources have the most impact of PM_2.5 during peak O₃ events.

During the current study period, O₃ daily maximum and 24 h average PM_2.5 concentrations demonstrated similar trends across the HGB region (Fig. 3), with r² values ranging from 0.52 to 0.68 (Figure S3) for the different TCEQ monitoring sites. The sites nearest to Texas City (Galveston and Seabrook), a port city by the Gulf with heavy petroleum-refining and petrochemical-manufacturing activities, had the highest correlations (r² = 0.57–0.68) for O₃ and PM_2.5 (Figure S3). A strong correlation between PM_2.5 and OC, for both 24 h and daytime concentration, was observed at MT. However, when the OC concentrations are compared to O₃ daily maximum concentrations, a stronger correlation was observed with the daytime (r² = 0.56) compared to the 24 h average (r² = 0.37) measurement (Figure S4c – e).

Ozone was significantly more correlated to SOC (r² = 0.56) than to primary OC (r² = 0.10). Although PM chemistry is complex, the same key species (i.e. NO_x and VOCs) which enhance oxidation processes to form SOA are also precursors for O₃ formation (Mazzuca et al., 2016). Amongst the apportioned fossil and contemporary TC at both sites, O₃ was more strongly correlated to fossil carbon (r² = 0.45) than contemporary carbon (r² = 0.34) and comparable to fossil SOC (r² = 0.53) (Fig. 9a and b). The intercept on the relationship of contemporary TC with O₃ indicates 0.8 μg that is independent of O₃. This intercept is only slightly lower for the biogenic SOC relationship with O₃ (0.6). Interestingly, the fossil TC intercept is nearly zero. While there is a stronger correlation with O₃ for fossil SOC than for biogenic SOC, the fossil SOC regression has a lower slope and a negative intercept. This seems to indicate that although fossil SOC has a stronger relationship with O₃, the fossil SOC associated with O₃ is not present in high concentrations and is only evident at higher daily O₃ mixing ratios. However, it should be noted that these interpretations are limited by the lack of O₃ data below 20 ppm. In contrast, the relationship of biogenic SOC with O₃ is weaker, but biogenic SOC increases at a higher ratio with increasing O₃.

The correlation analysis suggests anthropogenic sources and subsequent chemistry are important for both the promotion of O₃ and PM. Previous studies have found that co-emission of NO_x and light olefins from petrochemical facilities led to high O₃ production in Houston (Kleinman et al., 2005). During DISCOVER-AQ, the O₃ production rate was greatest near areas with high emissions of NO_x and VOCs which included Houston’s urban regions with hotspots measured over the HSC (Mazzuca et al., 2016). A strong correlation of O₃ to fossil carbon was observed at the both the suburban (r² = 0.88 at MC) and the urban (r² = 0.66 at MT) site. The Sept. 25 day both O₃ and PM concentration peak across the Houston metropolitan area during a period of observed atmospheric recirculation (Mazzuca et al., 2016). Stagnant wind conditions during this period in Houston (Fig. 5 and S2), allowing for the buildup of atmospheric pollutants. During this day, MT and MC observed relatively large contribution and concentration of fossil carbon and fossil SOC (Fig. 6). These results confirm the model conclusions from Dunker et al. (2019), which reported that the industrial emissions were increasing SOA. The Sept. 25 increase in fossil carbon and the correlation with O₃ suggests that it may not be only industrial NO_x which enhances SOA production, but fossil industrial precursors are also increasing SOA during this peak O₃ event. Additional study is needed to better understand the mechanisms of the potential SOA production associated with fossil point sources; however, these results strongly indicate the importance of these sources to O₃ production in the Houston metropolitan area.

4. Conclusion

The current study provided the first in-depth analysis of source apportioned daytime TC and its relation to O₃ during the DISCOVER-AQ Houston campaign. A novel source apportionment method combining molecular marker CMB and ¹⁴C analysis proved to be adept at apportioning both primary and secondary carbon sources. Based on these results, the carbonaceous aerosols in the Houston metropolitan area were dominated by biogenic SOC during DISCOVER-AQ. This finding supports previous DISCOVER-AQ studies which observed highly oxygenated OA (Dunker et al., 2019; Leong et al., 2017).

Though the composition of aerosol had consistent contributions from biogenic SOC and MVE, the fossil SOC had the greatest association with peak O₃. While fossil SOC concentrations were quite variable, they were enhanced at all sites during the high PM and peak O₃ event. Based on wind conditions and apportionment results, point source emissions from the HSC in conjunction with on- and off-shore wind cycling contributed to this peak event. This finding is in support of Dunker et al. (2019), who concluded that point source emissions were major contributors to OA and O₃ formation in the Houston metropolitan area.

The MT site (urban core) exhibited the largest and most consistent contribution from MVE and secondary fossil sources compared to the suburban MC site. Overall findings from the current study improve understanding of spatial trends (i.e. urban core and suburbs of the Houston metropolitan area) and identify key sources and factors that drive both O₃ and PM_2.5 production in the HGB region. Results from the current study highlight the need for continued high-intensity, multi-site studies to better understand impact of point sources from the HSC on air quality in the urban, suburb, and exurb regions of Houston. This will assist in effective regulation as the metropolitan area is still marginal nonattainment for O₃ amidst continual growth and spread of population into its suburbs and exurbs.

HIGHLIGHTS.

• Contemporary carbon was largest contributor to total organic carbon in Houston.

• Secondary organic carbon contributed an average of 67% of total organic carbon.

• Secondary fossil carbon was variable ranging from 3 to 27% of total organic carbon.

• During peak ozone event, secondary fossil carbon was well correlated with ozone.

• Houston was impacted by sea breeze re-circulation during high pollutant day.