On-Road Chemical Transformation as an Important Mechanism of NO₂Formation

Abstract

Nitrogen dioxide (NO₂) not only is linked to adverse effects on the respiratory system but also contributes to the formation of ground-level ozone (O₃) and fine particulate matter (PM_2.5). Our curbside monitoring data analysis in Detroit, MI, and Atlanta, GA, strongly suggests that a large fraction of NO₂ is produced during the “tailpipe-to-road” stage. To substantiate this finding, we designed and carried out a field campaign to measure the same exhaust plumes at the tailpipe-level by a portable emissions measurement system (PEMS) and at the on-road level by an electric vehicle-based mobile platform. Furthermore, we employed a turbulent reacting flow model, CTAG, to simulate the on-road chemistry behind a single vehicle. We found that a three-reaction (NO–NO₂–O₃) system can largely capture the rapid NO to NO₂ conversion (with time scale ≈ seconds) observed in the field studies. To distinguish the contributions from different mechanisms to near-road NO₂, we clearly defined a set of NO₂/NO_x ratios at different plume evolution stages, namely tailpipe, on-road, curbside, near-road, and ambient background. Our findings from curbside monitoring, on-road experiments, and simulations imply the on-road oxidation of NO by ambient O₃ is a significant, but so far ignored, contributor to curbside and near-road NO₂.

Graphical Abstract

Introduction

The transportation sector contributes about 30–40% of the emitted nitrogen oxides (NO_x),(1,2) and the health effects of nitrogen dioxide (NO₂) are a public health concern worldwide.(3–6) As part of the revised the National Ambient Air Quality Standards (NAAQS), the U.S. Environmental Protection Agency (USEPA) established monitoring requirements in urban areas to measure NO₂levels within 50 m of major roads as part of the NO₂ NAAQS. In Europe, NO₂ is one of the major contributors to urban air quality problems and has been routinely monitored in the near-road or street canyon environments.(3) The European Environment Agency estimated 71,000 premature deaths due to NO₂ exposure in Europe, higher than the estimated mortality caused by ground-level ozone (O₃).(3) From around 1990 onward, the total emissions of NO_x declined significantly in Europe by ~56%,(1) but roadside concentrations of NO₂ declined much less than expected.(7)Approximately 94% of exceedances of the European NO₂ annual limit were observed at traffic monitoring sites.(3) In China, ambient NO₂ concentrations in some megacities (e.g., Beijing, Shanghai) have persisted and exceeded the annual standard limit even though the concentrations of other gaseous pollutants have decreased significantly during the same period.(3,8,9) Even though China does not have a nationwide near-road air quality monitoring network, recently released near-road concentrations well exceed the current standard (e.g., an average exceedance of 20 to 30 μg m⁻³ for five near-road sites in Beijing).(10) Furthermore, besides being a criteria pollutant itself, NO₂ is also widely adopted as a marker for traffic-related air pollution in health studies.(11–13) For example, positive associations of NO₂ with mortality suggest that traffic pollution relates to premature death.(14)

In the complex roadway environments, direct tailpipe NO₂ emissions, secondary NO₂ through on-road and near-road NO_x titration and background NO₂ all lead to near-road NO₂. Therefore, a better understanding of the contributions from different pathways to near-road NO₂ is necessary for not only effective air quality management, but also evaluating the role of NO₂ as a marker in health studies, as near-road NO₂ could be dominated by different pathways at different conditions.

Moreover, the NO₂/NO_x ratio (typically by volume) is a widely used parameter in air quality management due to its applicability across the scales, from tailpipes to ambient environments. Both highway dispersion modeling (for NO₂) and regional air quality modeling (for O₃, PM_2.5, etc.) require inputs of NO₂/NO_x ratios, and previous studies have shown that both types of modeling are sensitive to those inputs.(15,16) Thus, another major motivation for our study is to elucidate the physical and chemical processes governing near-road NO₂ in order to provide appropriate NO₂/NO_x ratios for highway dispersion and regional air quality modeling efforts.

To facilitate the subsequent discussions, it is necessary to clearly define NO₂/NO_x ratios at different stages of plume evolution, listed in Table 1. The tailpipe NO₂/NO_x ratio (R_TP), also called the primary NO₂/NO_x emission ratio, is defined as the NO₂/NO_x ratio at the vehicle tailpipe exit. R_TP for gasoline vehicles is typically around 5% or lower(17,18) but can be higher for diesel vehicles equipped with diesel particulate filters (DPF).(19–21) The curbside NO₂/NO_x ratio, R_CS, is defined as the ratio at the edge of roadways. We further define the ratio between the tailpipe exit and the curbside as the on-road NO₂/NO_x ratio, R_OR. The ratio in the near-road environments (from the curbside to ~100 m) is defined as near-road NO₂/NO_x ratio, R_NR. Lastly, the ratio in the ambient environments is defined as background (or ambient) NO₂/NO_x ratio, R_BG.

Table 1.

Defined NO₂/NO_x Ratio at Different Locations in a Traffic Environment

location	NO2/NOx ratio	description
tailpipe exit	RTP	NO2/NOx ratio at the tailpipe exit, also called the primary NO2/NOx ratio, which can be measured by emission testing.
on-road	ROR	NO2/NOx ratios on the roadway, from the tailpipe to the curbside
curbside	RCS	NO2/NOx ratio at the curbsides
near-road	RNR	NO2/NOx ratio from the curbside to about 100 m away from the roadways
background	RBG	NO2/NOx ratio from the upwind direction or in the ambient environments

It is well recognized that NO_x transforms chemically from NO to NO₂ through titration of O₃ in the near-road environment, resulting in R_NR differing from R_CS. The near-road chemical transformation has been observed in several field studies.(22–25) In our previous study, we estimated R_CS to be ~20–40% based on near-road monitoring data collected at several Texas roadways in the summer of 2007.(16) The roadways studied had low diesel traffic, and the penetration of diesel particulate filters was very low in 2007. Therefore, the relatively high R_CSmust be explained by mechanisms in addition to the presence of diesel traffic.

This study aims to investigate the physiochemical mechanisms that lead to enhanced curbside NO₂/NO_x ratios (R_CS). We posit that on-road (i.e., “tailpipe-to-road”(26)) chemical reactions are a major contributor to NO₂ production, resulting in an evolution from R_TP to R_OR and to R_CS. To our knowledge, on-road chemical transformation of NO_x has not been explicitly studied. However, several recent studies have implicitly indicated the significant role of on-road NO₂ production. The curbside measurements of NO, NO₂ and O₃ in street-canyon environments at two German cities showed that secondary NO₂ production was responsible for a major fraction of measured NO₂.(27) One can deduce from the study that the referred production of NO₂ most likely took place on the streets. Another study of near-road monitoring data at Las Vegas, NV, indicated the R_CSranging from 0.25 to 0.35, substantially higher than the anticipated R_TP,(28) which we argue can potentially be explained by the on-road production of NO₂. Even though the term “on-road” NO₂/NO_x ratio was used in the Las Vegas study,(28) a close examination of the reported methodology suggested that the term was referred to R_CS according to our definition.

In this study, we present the supporting evidence from (1) curbside monitoring near two major highways in Detroit, MI, and Atlanta, GA, respectively, (2) an on-road field experiment in Research Triangle Park (RTP), NC, and (3) on-road simulations using a turbulent reacting flow model. This work is intended to improve future modeling of near-road NO₂ concentrations, air quality and transportation management, and health studies on traffic-related pollutants.

Curbside Monitoring

Site Description and Field Measurements

Detroit, MI

Field measurements were performed at several monitoring sites in Detroit, MI, including the Eliza Howell near-road sites, starting fall 2010 and concluded in late summer 2011. For this study, we used data collected between September 2010 and June 2011, when the traffic and meteorological conditions were both well characterized. These sites were originally deployed as part of the Federal Highway Administration (FHWA) and USEPA National Near-Road Study for a one-year period, then continued by the Michigan Department of Environmental Quality (MDEQ) to the present. Data from these sites have being used in the Near-Road Exposures and Effects of Urban Air Pollutants Study (NEXUS), which examined the relationship between near-roadway exposures to air pollutants and respiratory outcomes in a cohort of asthmatic children who live close to major roadways in Detroit, Michigan.(29)

The Eliza Howell 1 monitoring site (EH1; 42°23′09.6″N, 83°15′58.9″W) is placed at the curbside of Interstate I-96/Jeffries Freeway (Figure S1a in the Supporting Information (SI)). On the opposite side, site EH4 is located ~100 m south of I-96, which served as an upwind site when EH1 was downwind of I-96, and allowed us to quantify the contribution of background NO and NO₂ to those measured at EH1. The immediate area is mostly open parkland with some trees and playing fields; surrounding areas are mostly low-rise commercial and residential buildings.

NO/NO₂/NO_x were measured at a height of 3 m using chemiluminescence (Teledyne ML9841B Nitrogen Oxides Analyzer),(30) which reports NO₂ as the difference between NO_x and NO. Wind speed and wind direction were measured at a height of 10 m. Both NO_x and surface meteorology were also measured at EH4. Hourly ozone (O₃) measurements from the Allen Park (AP) site (AQS Site ID: 261630001; 42°13′43.0″N, 83°12′29.9″W), located ~18.5 km south of EH1 and ~200 m southeast from the closest major roadway (I-75), were obtained from the USEPA Air Quality System (www.epa.gov/ttn/airs/airsaqs/). A closer O₃ measurement site, E7Mile (AQS Site ID: 261630019; 42°25′50.9″N, 83°00′01.0″W), was not selected due to more than three months missing data between September 2010 and June 2011.

Traffic on I-96 is normally free-flowing with an established speed of 70 mph for cars and 65 mph for trucks. The annual average daily traffic (AADT) on this I-96 segment are 155,000.(30) Based on vehicle classification data measured during the study period at the EH1 site (using radar to classify vehicle size), the percentage of heavy-duty diesel trucks (HDDT) for every 5 min averaged ~3.0%.

Atlanta, GA

Continuous air quality monitoring data of NO, NO₂, and O₃ were obtained at two sites in Atlanta, GA: a curbside site off Interstate I-85 near Georgia Institute of Technology (NRGT; AQS Site ID: 13–121-0056; 33°46′41.934″ N, 84°23′29.1048″ W) and a background site in South DeKalb (SDK; AQS Site ID: 13–089-0002; 33°41′16.692″ N, 84°17′25.7280″ W), about 13.5 km southeast from the NRGT site. Figure S1b in the SI depicts the locations of the two sites. The monitoring at NRGT started near the end of May 2014. We analyzed data collected between January 2015 and March 2017 since minute-level data of of NO, NO₂, and O₃ from both NRGT and SDK are available during this period.

I-85 is a major interstate freeway that runs northeast–southwest in Georgia. The AADT on this segment is ~332,000 in 2015, and ~353,700 in 2016, accordingly to the closest traffic monitoring station (ID: 1215481; 33°46′23.8800″ N, 84°23′24.0000″ W) (http://trafficserver.transmetric.com/gdot-prod/tcdb.jsp?siteid=1215481). The truck percentage is ~3.7% based on another nearby station (ID: 1216118; 33°46′32.88″ N, 84°23′24.72″ W) (http://trafficserver.transmetric.com/gdot-prod/tcdb.jsp?siteid=1216118).

Curbside Data Analysis

To investigate the role of on-road chemical transformation of NO_x, we screened the Detroit data using the following criteria: First, we selected daytime data (6AM-6PM) under south-dominant wind conditions (101.25 < θ < 258.75, θ is the wind direction in degrees) so that EH1 was downwind of I-96. Next, we subtracted the upwind NO₂ and NO_x concentrations (i.e., at EH4) from the corresponding values at EH1, thus reflecting the net contributions of highway emissions. Finally, we set a net NO_x concentration threshold to ensure the curbside data reflect the highway influence. In other words, the curbside NO_x concentrations are expected to be higher than the upwind NO_x concentrations. The higher the threshold value, the more confident we are of the highway influence, given the uncertainties and detection limits in the NO_x measurements. But the same time, a higher threshold value may also exclude some low-traffic conditions. We tested the sensitivity of our results to the net NO_x concentration threshold. We followed similar steps to analyze the Atlanta data. First, we selected daytime data (8AM-5PM) under east-dominant wind conditions (11.25 < θ < 168.75) so that NRGT was downwind of I-85. Next, we subtracted the ambient NO₂ and NO_x concentrations (i.e., at SDK) from the corresponding values at NRGT to reflect the net contributions of highway emissions. We also tested the sensitivity of net NO_xconcentration thresholds.

Figure 1 summarizes the main findings from our curbside analysis of the Detroit and Atlanta data sets. The relationships between net curbside NO₂/NO_x ratios (R_CS) and ambient O₃concentrations (with net NO_x concentrations represented by color scales) are illustrated in Figure 1a,b for Detroit and Atlanta, respectively. The overall linear relationships between O₃concentrations and net R_CS from these two curbside measurements support the significance of on-road chemical transformation. In other words, higher ambient O₃ concentrations, once mixed with highway emissions, convert NO to NO₂ during the tailpipe-to-road process. Conversely, if the role of on-road chemical transformation was insignificant, then net R_CS would not show a relationship with O₃ concentrations.

Figure 1.

Curbside NO2/NOx ratio (RCS) vs ambient O3 concentrations (ppb) with net NOx threshold 40 ppb at (a) Detroit, MI (I-96), during the daytime and south dominated wind (101.25 < θ < 258.75) and; (b) Atlanta, GA (I-85), during the daytime and east dominated wind (11.25 < θ < 168.75). Linear regression parameters (slope, intercept and R2) of the RCS vs O3 relationships and the amount of data included (as percentage of the total data available) as functions of net NOx concentration thresholds for (c) Detroit, MI (I-96), and (d) Atlanta, GA (I-85).

Parts c and d of Figure 1 depict the sensitivities of linear regression parameters (slope, intercept, and R² as shown in Figure 1a,b) and the amount of data included (as percentage of the total data available) to the different net NO_x concentration thresholds, ranging from 10 to 50 ppb. As expected, the percentage of data included decreases with the net NO_x threshold for both highways. At the 10 ppb threshold, around 76% data are included in Detroit, MI, and 94% data were kept in Atlanta, GA. At the 50 ppb threshold, both of the two data sets dropped to around 50%. However, the slopes from both data sets appear not sensitive to the net NO_x threshold, only changing from 0.71 to 0.81 for Detroit and from 0.66 to 0.73 for Atlanta, respectively. The R²values for Detroit, varying from 0.73 to 0.76, were not sensitive to the NO_x threshold, while those for Atlanta were more sensitive to the threshold, increasing from 0.57 to 0.70 as the NO_x threshold increases from 10 to 50 ppb. Parts a and b of Figures 1 were both plotted at 40 ppb net NO_xthreshold to achieve a balance of high percentage data inclusion and high R².

The intercepts for Detroit and Atlanta both appear to be sensitive to the NO_x threshold, but with opposite directions. Physically, the intercepts represent the tailpipe NO₂/NO_x ratios, R_TP. The intercepts for Detroit are negative (albeit small in magnitude), approaching positive values as the NO_x threshold increases. The intercepts for Atlanta are positive, starting with relative large magnitude, e.g., 15% when the NO_x threshold is set to 10 ppb and decreasing to ~11% at the 50 ppb net NO_x threshold. One contributing factor to the negative intercepts in Detroit was found to be associated with the traffic patterns. Figure 1a indicates high NO_x conditions predominately appear when ambient O₃ concentrations are low, which represented the early morning and late afternoon traffic peaks. Those high-NO_x and low-O₃ conditions lead to a small fraction of NO being converted and thus low NO₂/NO_x ratios. The clustering of data points of low NO₂/NO_x ratios and low O₃ concentrations pushes the linear regression toward the negative intercept side. By contrast, high NO_x conditions are evenly distributed at the NGRT site across the whole range of O₃ concentrations, suggesting a weak diurnal traffic pattern. It is worth noting that I-85 segment near NRGT is one of the most trafficked highways in the U.S.

Figure 1a raises the question over the contribution of on-road chemistry to curbside NO₂concentrations in Detroit since conditions with high net R_CS and high ambient O₃ appear when net NO_x concentrations are relatively low. Figure S2 in the SI illustrates the relationship between net NO₂ concentrations (i.e., from both primary NO₂ emissions and on-road chemical transformation) and ambient O₃ from the Detroit data set, which presents a positive linear trend and implies a significant contribution from on-road chemical transformation to curbside NO₂ concentrations. Furthermore, Figure 1b (based on the Atlanta data) corroborates this conclusion as conditions with high net R_CS and high ambient O₃ appear with high net NO_x concentrations (thus high net NO₂concentrations). Moreover, Figure S3 in the SI depicted the relationships between R_CS and solar radiation at different ambient O₃ concentrations. The results suggest that R_CS is not very sensitive to the NO₂ photolysis rates (with solar radiation as a surrogate), which in turn indicates that the titration of O₃ by NO is the dominant reaction. Future studies are needed to conduct detailed analysis of the contributions from different pathways (such as ambient background, tailpipe or primary NO₂ emissions, on-road chemistry, near-road chemistry) to curbside/near-road NO₂concentrations.

We acknowledge the uncertainties in the curbside data analysis. For example, the NRGT does not have a near-road upwind site for us to reliably estimate net NO_x contributions, which at least partially explain why the regression parameters are much more sensitive to the net NO_x threshold in Atlanta than Detroit. In addition, we did not have near-road upwind O₃ measurements for either Detroit or Atlanta. Several studies reported spatial variations of O₃ concentrations in urban areas.(31–33) The O₃ concentrations measured at the ambient sites may not accurately represent the near-road upwind conditions. Nevertheless, the curbside data from two different highways showed very consistent trends. Together, the curbside analysis provides strong evidence for the significant role of on-road NO_x transformation.

On-Road Experiment

Field Measurements

Motivated by the finding from the curbside data analysis described earlier, we designed an on-road field campaign to gather further evidence on the on-road NO_x chemical transformation. The overall concept of the on-road experiment was to measure the same exhaust plumes both at the tailpipe-level by Portable Emissions Monitoring System (PEMS) and at the on-road level (after initial dilution) by a mobile chasing platform. Then we would expect to observe noticeable differences in the tailpipe-level and on-road level NO₂/NO_x ratios with significant on-road NO_xchemical transformation. However, it is technically challenging to directly measure on-road NO₂/NO_x ratios (R_OR) because reliable, fast-responding instruments for NO (or NO_x) are not readily available. Wild and co-workers recently developed a fast-responding instrument via Cavity Ring-Down Spectroscopy.(34) But we did not have access to that instrument. As an alternative, our analysis relied on CO₂ and NO₂ by defining CO₂-normalized NO₂ concentration as ΔNO₂/ΔCO₂, essentially using CO₂ as a dilution indictor, for PEMS and mobile platform measurements, respectively. Details for this analysis are provided later.

In our experiment, we employed an electric vehicle-based mobile platform equipped with air quality analyzers at 1 s sampling intervals to conduct the on-road chasing measurements, capturing the on-road concentrations of CO₂ (Quantum Cascade Laser, Aerodyne Research, Inc.) and NO₂ (Cavity Attenuation Phase Shift, Aerodyne Research, Inc.). The sampling probe was positioned at the bottom front of the mobile platform, around the same height as the exhaust tailpipe of the vehicle being followed. We detected a six-second delay in response due to the sampling line, and this delay was accounted for in the subsequent data analysis. The chasing routes were recorded using a high-precision GPS unit (Crescent R100, Hemisphere GPS). A webcam was operated in the front seat of the mobile monitoring vehicle and recorded the traffic conditions continuously during the on-road measurements.

The vehicle being chased (referred to as “target vehicle”) was a 2003 Ford F350 diesel powered truck (engine displacement 6.0 L, 8 cylinders, ~ 53,000 miles, 7940 lbs actual weight). Tailpipe-level CO₂ and NO₂ concentrations were measured using a Portable Emissions Monitoring System (PEMS) (SEMTECH Ecostar; Sensors, Inc.). The PEMS samples directly from the subject vehicle’s raw exhaust and is capable of providing on-board CFR 1065 compliant emissions determinations at a rate of 1 Hz as the subject vehicle is being operated on the highway. The PEMS utilizes a nondispersive infrared (NDIR) sensor for CO₂ measurements and a nondispersive ultraviolet (NDUV) sensor for NO₂ measurements.

The on-road field campaign took place from February 22nd until March 4th, 2016 in Research Triangle Park (RTP), NC. The measurements conducted in the afternoon of February 29 (FEB29PM), the morning of March 1 (MAR01AM), and the afternoon of March 1 (MAR01PM) focused on comparing PEMS and on-road chasing results.

On-Road Data Analysis

We started the data analysis by screening the videos recorded during the experiment to select the time segments without the presence of heavy duty vehicles around the target vehicle. Emissions from heavy duty vehicles may distort the on-road chasing results. For the selected time segments, we processed the chasing data using 2 s average to remove noise in the raw 1 s data.

Then a two-step synchronization was adopted to further process the data. The first-step was to synchronize the chasing CO₂ and NO₂ signals, which usually required 1-s or 2-s time shift in time series, to identify synchronized CO₂ and NO₂ spikes. At the end of the first step, we obtained a series of synchronized CO₂ and NO₂ spikes that captured on-road plume signals with high confidence. The second step was to synchronize chasing plume signals with PEMS data. The distance between the chasing object and the mobile platform was usually within 20 m. Given the exit velocity ~20 m s^–1, the time delay between the PEMS and chasing signals should be ~1 s. In practice, the acceleration of the truck usually caused a large spike in both PEMS and chasing signals, which helped us to determine the time delay. At the end of the second step, we acquired a list of paired PEMS and chasing signals, representing the same exhaust plumes measured by the two methods.

The CO₂-normalized NO₂ concentration (ΔNO₂/ΔCO₂) can be visualized by the slope of NO₂ vs CO₂ scatter plots. Figure 2 illustrates how we calculated CO₂-normalized NO₂ concentration using one pair of synchronized plume signals as an example.

Figure 2.

One pair of synchronized plume signals of NO2 vs CO2 concentrations based on (a) on-road chasing measurement and (b) tailpipe-level PEMS measurement for deriving CO2-normalized NO2 concentrations.

We identified a total of 109 synchronized plume signals, and calculated the ratio of CO₂-normalized NO₂ concentrations from on-road chasing over tailpipe PEMS measurements for each plume signal, referred to as Ratio_Chase/PEMS, i.e., a Ratio_Chase/PEMS > 1 implies on-road NO₂production. Figure 3 shows the distributions of this ratio and coefficients of determination (R²) for the linear regressions (as indicated in Figure 2). Ratio_Chase/PEMS > 1 for all 109 plume signals with the median value as 1.87, and R² values for both PEMS and on-road chasing measurements are high (with the median values over 0.8), suggesting the results are reasonably robust.

Figure 3.

Distributions of (a) RatioChase/PEMS (i.e., the ratio of CO2-normalized NO2 concentrations from on-road chasing over tailpipe PEMS measurements) and (b) R2 values for deriving CO2-normalized NO2 concentrations from chasing and PEMS measurements. In both (a) and (b), the box represents the middle 50% of the data, extending from the 25th to the 75th percentiles; the horizontal line through the center of the box is the median; the whiskers represent 10th and 90th percentiles.

In summary, the results from the on-road experiment further indicate the significant role of on-road NO_x chemical transformation.

Coupled On-Road Turbulence and Chemistry Modeling

Methodology for On-Road Simulations

As the third part of our study, we employed the Comprehensive Turbulent Aerosol Dynamics and Gas Chemistry (CTAG) model to simulate the coupled on-road turbulent mixing and chemistry of individual exhaust plumes to test whether the rapid transformation observed is kinetically feasible under typical on-road conditions. CTAG is an appropriate model for this purpose as it was designed to simulate transport and transformation of multiple air pollutants, e.g., from emission sources to ambient background.(35–40)

Shown in Figure 4, the computational domain was constructed based on a single lane with an individual passenger car. The domain dimensions are 30 m (L) × 15 m (W) × 10 m (H). The positive Y direction is the downwind crosswind direction. The geometry of the passenger car was modeled using a realistic shape rather than block shapes.

Figure 4.

Computational domain and mesh around the car.

While full-scale simulations with detailed site-specific conditions (e.g., capturing multiple vehicles corresponding to traffic mix, highway geometry, etc.) are beyond the scope of this paper, we utilized the environmental conditions captured in the Detroit data set to set up the simulations. The velocity inlet boundary was specified on the left and top sides of the domain to simulate the ambient south crosswind speed. The periodic boundary conditions with the flow rate based on vehicle speed (70 mph or 31.3 m s^–1 as in the I-96 segment), specified on both ends of the computational domain, was adopted to represent the continuous traffic flow. The passenger car’s tailpipe, 60 mm in diameter, is parallel to the road and specified as a mass flow inlet with a total mass flow rate of 0.055 kg s^–1 and exhaust temperature of 480 K.(41) The tailpipe-level NO_xemission rate was estimated by using EPA’s MOtor Vehicle Emission Simulator (MOVES). A NO₂/NO_x volume ratio of 5% was assigned at the tailpipe level (i.e., R_TP = 5%). The road surface was modeled as a moving wall at the vehicle speed, thus representing the relative motion between the vehicle and the road. Ambient background NO_x and NO₂ concentrations for each case are specified based on the Detroit data (EH4).

We modeled the dispersion and chemical reactions using steady-state Reynolds averaged Navier–Stokes (RANS) turbulence model to simulate the flow fields and species transport. The finite rate (FR) chemical reaction model(42) was coupled with the eddy dissipation method (EDM),(43) referred to as the EDM/FR model, in which turbulence–chemistry interaction is accounted for through the representation of the source term the species transport equations. Both an Arrhenius rate based on the global chemistry mechanism, and a turbulent mixing rate based on the Magnussen–Hjertager expression are estimated.(43) The smaller of the two rates is then used as the source term. We adopted a simplified chemical mechanism using a four reaction (NO–NO₂–O₃) system shown in Table S1 with reaction constants taken from the CB05 mechanism.(44)These reactions have been shown to be dominant in near-source environments.(16,45) The photolysis rate of NO₂ was determined based on our previous work(16) (also described in section S4 of the SI) and average solar radiation at the near-road site in Detroit obtained from the U.S. National Solar Radiation Database. The governing equations of the Realizable k-ε model (one type of RANS models) and the species transport equation are presented in section S2. The mesh resolution sensitivity study is illustrated in Figures S4 and S5.

Results from On-Road Simulations

Figure 5a illustrates the contour of R_OR on a horizontal plane at the tailpipe height in the simulation domain (Figure 4) with the ambient ozone concentration set to be 40 ppb. Some representative R_OR values are marked in along the tailpipe centerline with the corresponding residence time, estimated by dividing the distance to the tailpipe center by the driving velocity. Figure 5b shows R_OR values along the tailpipe centerline as a function of resident time at levels of ambient O₃ concentrations (20, 40, and 60 ppb).

Figure 5.

(a) On-road NO2/NOx ratio (ROR) contour plot on a horizontal plane at the tailpipe height in the simulation domain illustrated in Figure 4. The ambient ozone concentration is set to be 40 ppb. The black solid line is the tailpipe centerline. Values along the solid line represent the resident time and corresponding ROR. (b) ROR along the tailpipe centerline for different ambient ozone concentrations.

It is revealed from Figure 5 that higher the ambient O₃ concentration leads to higher R_OR. As expected, R_OR remain nearly a constant if chemical reactions are turned off. Furthermore, the coupled turbulent mixing and simplified NO–NO₂–O₃ chemistry can achieve the NO₂ conversion rate similar to the order of magnitude of the observed values, though a direct comparison is not feasible. The values of Ratio_Chase/PEMS derived from the on-road experiment are roughly equivalent of changes in NO₂/NO_x ratios. Given the ~1 s transport from the tailpipe of the target vehicle to the probe of the mobile platform, a Ratio_Chase/PEMS of 2 suggests doubling of NO₂/NO_xratios in less than 1 s at ambient O₃ concentration ~40 ppb. The modeling results show that even at relatively low ambient O₃ level (e.g., 20 ppb), R_OR can double of R_TP in less than 1 s. Figure S6 in the SI illustrates the R_OR values along the tailpipe centerline as a function of resident time at different NO₂ photolysis rates with ambient O₃ concentration at 40 ppb. The results suggest that R_OR is not sensitive to NO₂ photolysis rates, which further supports NO+O₃ as the dominant reaction during the “tailpipe-to-road” stage. As mentioned previously, R_OR also depends on the relative amounts of NO_x and O₃. More in-depth studies are needed to explore how different traffic and meteorology conditions affect on-road NO_x transformation.

Implications

Our study has provided experimental evidence, through curbside data analysis and on-road field measurements, that substantial amounts of freshly emitted NO are oxidized to NO₂ during the “tailpipe-to-road” stage. The subsequent modeling analysis indicates that chemical conversion is rapid, and that NO–NO₂–O₃ chemistry significantly increases NO₂/NO_x ratios in the on-road environment from the relatively low ratio at the tailpipe. The findings have raised the importance of differentiating NO₂/NO_x ratios at various stages of exhaust plume evolution in order to account for the contributions of primary emissions, on-road chemistry, near-road chemistry and ambient background to near-road NO₂ concentrations. We have shown that on-road chemistry is a major contributor to curbside NO₂, and likely near-road NO₂. In general, R_TP < R_OR < R_CS < R_NR. The study results have important implications for transportation and air quality management. First, the on-road distributions of NO₂/NO_x ratios resulting from chemistry and turbulent mixing challenge a common assumption in all major highway dispersion models that pollutants (as well as NO₂/NOx ratios) are uniformly distributed on and over the highways.(46–48) In practice, this assumption essentially treats R_TP = R_OR = R_CS, which has been shown to be invalid by our study and can lead to significant underestimation of near-road NO₂ as assumed R_TP are usually taken from tailpipe measurements without accounting for on-road chemistry. Our analysis can potentially lead to a parametrization scheme that accounts for the effects of on-road NO_x chemistry, which provides more accurate inputs for near-road NO₂ modeling and enhance the capability of modeling near-road NO₂ concentrations for regulatory purposes. Second, considering the significant production of NO₂ on roadways, strategies for attaining near-road NO₂ standards need to account for the effects of on-road NO_x chemistry, i.e., not only local tailpipe emissions, but also regional ozone concentrations. Finally, properly representing the effects of on-road chemistry on NO₂ production will also improve the regional O₃ and NO₂ modeling. A regional modeling study in Southeast Texas showed compared to the traditional 5% NO₂/NO_x emission ratio, a higher NO₂/NO_x ratio (29%) could lead to a 6 ppb increase in the 8 h ozone.(15) Another regional air quality study in Europe indicated an up to 8 μg m³ increase in the annual average NO₂ concentration when the NO₂/NO_x emission ratio is increased from 20 to 70%.(49) Future studies of the joint impact of chemical transformations through on-road chemistry and catalytic chemistry in diesel after treatment devices are needed.