James Kacera,
Ralph Altmaier
a,
David M. Cwiertny
bc and
Patrick T. O'Shaughnessy
*a
aDepartment of Occupational and Environmental Health, College of Public Health, University of Iowa, USA. E-mail: patrick-oshaughnessy@uiowa.edu
bDepartment of Civil and Environmental Engineering, College of Engineering, University of Iowa, USA
cDirector, Center for Health Effects of Environmental Contamination, University of Iowa, W195 Chemistry Building, Iowa City, Iowa 52242, USA
First published on 19th August 2024
Particulate matter (PM) emitted from unpaved rural roads presents a potential inhalation hazard to people living and working near them. In the absence of site-specific exposure data, plume dispersion modeling can be used to predict ambient particulate concentrations in the vicinity of the unpaved roads. Hourly averaged PM10 concentrations were measured near a gravel road using an EPA reference method resulting in a geometric mean of 50 μg m−3. With these ambient concentrations, the AERMOD plume dispersion model was used to derive a PM emission factor of 444 g/VKT (grams per vehicle kilometer travelled). This result was lower than the emission factor calculated using the EPA's AP-42 guidance for unpaved roads (795 g/VKT). The transient nature of the plume of PM concentrations due to road traffic was also evaluated using a direct-reading instrument. Vehicle speed and wind speed were found to be significant determinants of PM concentration, average PM concentration, and total PM mass for each plume. Each vehicle produced an average concentration of 4096 μg m−3 over the duration of the plume. Therefore, residents near the road are potentially exposed to substantially higher short-term concentrations from individual plumes than would be indicated by hourly averages.
Environmental significanceGiven their worldwide ubiquity, unpaved roads have the potential to be a large-scale health hazard as a source of exposure to ambient particulate matter (PM) in rural areas. However, their length scale causes PM exposures from unpaved roads to be difficult to assess. This study determined a generalizable PM emission rate using the inverse-modeling approach for application in a plume dispersion model. Notably, 18% of measured concentrations exceeded the 24 hour PM standard. Since the ambient PM is created by passing vehicles, the resulting airborne dust plumes were also investigated. Roadside average concentrations during a plume event were typically 30 times over the PM standard and peak concentrations were over 100 times the standard, therefore stressing the importance of PM from unpaved roads as a potential health hazard. |
Although PM concentrations near unpaved roads can be measured using a number of stationary monitoring methods, their length scale diminishes the usefulness of measurements at any given nearby location to assess community exposure risk. To augment that problem, studies have been conducted using portable, direct reading instruments (DRIs) to measure PM concentrations near unpaved roads. Many of these studies were conducted to compare concentrations measured with DRIs relative to measurements made with federal reference methods (FRMs12) or other methods13–16 for assessment of risk. For example, Zhao et al.14 used an aerosol photometer and an optical particle spectrometer alongside several FRMs to compare these DRIs to FRM instruments. The results from the DRIs indicated that the FRMs under-sampled smaller particles based on wind direction. Zhao et al.14 was the only study found that used DRIs to characterize PM plumes developed from passing vehicles on a time scale less than 1 h. They compared the response from several DRIs operating at 1–5 s sample rates to plumes generated by passing vehicles but did not further characterize the individual plumes relative to, for example, vehicle speed or peak PM concentrations generated.
To supplement and enhance PM exposure information from monitors, plume dispersion modeling has been used to estimate PM concentrations parallel and perpendicular to unpaved roads. The plume dispersion model, AERMOD, has been used for this purpose in several studies involving paved and unpaved roads.17,18 AERMOD is the EPA's preferred model for near-field (up to 50 km from the source) analyses of pollutant concentrations19 developed for use with steady-state emission sources20,21 and is capable of evaluating several types of sources including roads. Askariyeh et al.17 compared modeled and tracer concentrations of SF6 using volume sources to represent roads in the model and found that AERMOD more accurately predicted tracer concentrations at low elevations (0.5 m) than at high elevations (3.5 m and 9.5 m) under low wind speed conditions. Huertas et al.18 used AERMOD to aid in developing a function to predict PM concentrations emanating perpendicular to the edge of unpaved roads.
In addition to local meteorological information, a primary dispersion model input that influences modeled concentrations is the emission rate of the source. The term “emission rate” is defined for use in AERMOD as the mass per second (units of grams per second) emitted from the source, whereas an “emission factor” for unpaved roads is defined in the EPA AP-42 emission estimation method22 as mass per vehicle distance travelled. Studies have been conducted to measure emission rates directly with the use of DRIs.13,16,23,24 Kavouras et al.13 used a vehicle-mounted photometer to estimate PM10 emission factors and found statistically significant spatial and temporal variability in emission factors, based on differing road surfaces and vehicle traffic characteristics between sample sites. Edvardsson et al.16 used aerosol photometers mounted downwind on vehicles to estimate dust emissions and to estimate PM decay with distance from roads to which different dust suppressants had been applied. They found that PM concentrations were dependent on the vehicle speed and type of vehicle. Gillies et al.24 also used an aerosol photometer to evaluate the differences in emissions from unpaved roads based on vehicle types. Fitz et al.23 used several DRI types mounted on a trailer to measure vehicle emissions on paved roads and found a close comparison between their results and those derived from the AP-42 methods for paved roads. However, a similar study comparing either measured or estimated emission rates of PM from unpaved roads to estimates using the AP-42 method could not be found in the literature.
Inverse dispersion modeling is an approach that has also been used to estimate particulate emissions from unpaved roads and to identify factors that influence emission factors and airborne particulate concentrations.25 When applying inverse modeling, the emission rate is adjusted in the model so that the model output for a receptor location matches a field measurement at that location. Meteorological conditions and other model inputs are typically those observed during field activities for a given sampling duration. For example, the Industrial Source Complex model19 was used to estimate the particulate emission rates from unpaved roads in Riverside County, California26 using the line source option. Another study used tracers and the line source model to back-calculate emission factors for an unpaved road with low traffic (average 5 vehicles per hour) that reported emission factors ranging from 75 to 298 g/VKT.27
Despite the millions of miles of unpaved roads in the United States, there have been relatively few studies conducted to monitor PM concentrations near them. Furthermore, no studies could be found that compared PM emissions from unpaved roads to the AP-42 method for calculating those emissions, nor has a study been performed that systematically characterized the intermittent plumes of PM generated by passing vehicles. Therefore, the primary aim of this study was to determine a fugitive PM10 emission rate for rural unpaved roads located in the Midwest United States using the inverse-modeling approach. A sub-aim was to compare the average emission factor estimated by inverse modeling with the emission factor estimated using EPA AP-42 methodology.22 A second sub-aim was to evaluate the nature and magnitude of transient PM concentrations developed as short-term plumes by individual moving vehicles.
Road rock in this area was crushed calcite and/or dolomite.29 The crushed rock on these unpaved roads was supplemented with electric arc furnace (EAF) slag from a nearby steel foundry. There was no significant difference in PM10 concentrations measured near roads supplemented with slag compared to those without supplemental slag when corrected for traffic volume.29 Therefore, the four sample sites were considered representative of rural gravel roads in the Midwest. None of the locations had been treated to suppress dust emissions from the road surface prior to sampling.
Wind direction and velocity data for the sampling period were obtained from a meteorological station (Vantage Vue, Davis Instruments, Hayward, California) located on a neighboring property. Other meteorological data required for the model, such as ceiling height and visibility, were obtained for the National Weather Service facility in Moline, Illinois located 63 km east of Site 1.
Two trail cameras (Model H45, Apeman, Shenzhen, China) were mounted on a post alongside the road at a height of approximately 1 m to obtain traffic counts. The photographs were time- and date-stamped so that traffic counts could be obtained for each one-hour sampling period.
At both locations (Sites 3 and 4), the photometer was positioned approximately 1.5 m above ground level on a tripod and approximately 15 m from the centerline of the adjacent road. Measurements were taken under a variety of wind conditions. A handheld anemometer (Model BTMETER BT-876, Zhuhai, Guangdong, China) and a windsock located near the photometer were used to measure wind velocity and direction during sampling events. Photometer measurements were collected when a project team member drove past the sample location at known speeds of 53, 80, and 113 km h−1 (33, 50, and 70 mph) in a mid-sized sedan.
The line volume source option for haul roads was chosen to represent the unpaved road near Site 1. In effect, the line volume source type is a series of adjacent volume sources with a total length equivalent to the road length. This source option was chosen because it includes a meander algorithm which accounts for changes in concentrations due to non-diffusing eddies, and because it better approximated monitoring data in a limited number of studies.33 However, ambient concentrations cannot be calculated within the volume source exclusion zone, which is defined in AERMOD as the effective radius of the source (2.15 times the lateral plume dimension plus one meter).34 Therefore, the Site 1 sample location (15 m north of the centerline of the road) was placed outside of the exclusion zone (9 m from the centerline of the road).
Each modeling run consisted of a one-hour period corresponding to a one-hour period of BAM data collection. Although road dust emissions are not steady-state at the study site, they were assumed to be constant during the one-hour sampling period of the BAMs at Sites 1 and 2. A receptor height of 4 m was used in the model since that was the height of the air intake of the BAMs. Other model assumptions included a vehicle height of 2 m, a vehicle width of 2 m, and a single-lane road. Applying the AERMOD default factors to the vehicle height and width, the release height was set at 1.7 m with a plume height of 3.4 m and a sigma z (initial vertical dilution of the emissions) of 1.58 m; the plume width was set at 8.0 m with a sigma y (initial lateral dilution of the emissions) of 3.72 m.35
![]() | (1) |
The geometric mean (GM) of the emission rates determined by inverse modeling in units of g s−1 m−1 was divided by the number of vehicles (V) traveling on the unpaved road during a sampled hour (g s−1 m−1 V−1) to determine an emission factor in units of mass per vehicle kilometer traveled (g/VKT) and converted to mass per vehicle mile traveled (lb/VMT) for comparison with the AP-42 method. The EPA AP-42 equation developed for emissions from light-duty vehicles traveling on publicly accessible unpaved roads is defined as:22
![]() | (2) |
There are several relevant differences in these methods for determining an emission factor. The AP-42 method includes ranges for vehicle weights while AERMOD requires vehicle height and width as inputs. The AP-42 method includes site-specific data for silt content and moisture of the unpaved road surface, whereas AERMOD ignores these road characteristics. Furthermore, the AP-42 method directly accounts for vehicle speed while inverse modeling only implicitly accounts for vehicle speed as it contributes to the resulting emission factor.
Vehicle characteristics have been found to have a significant effect on the calculation of emission factors.37 Therefore, the sensitivity of the inverse-modeled per-vehicle emission rate to variations in vehicle height and width were also calculated using a range of values for these variables that are representative of the expected conditions at the sample site. Changes were made to vehicle height and width while holding the meteorological conditions applied to the model constant and representative of a warm, summer day near noon.
A sensitivity analysis was also undertaken to investigate the relative effect of different wind speeds and directions on the value of the inverse-modeled emission rate. A single hour of meteorological data was applied to AERMOD representing a typical summer afternoon and the concentration at the receptor (Site 1 sample location) was set at 50 μg m−3. The source descriptors applied were the same as those described in “Dispersion Modeling.” Inverse modeling was applied to determine the emission rate while varying wind direction between 180° to 270° by every 22.5°, and wind speed was set at 1.6, 8, 12, and 16 km h−1 (1, 5, 7.5, 10 mph).
Multiple regression analysis was applied to determine whether these plume characteristics were significantly affected by the continuous independent variables of vehicle speed, wind speed, and wind direction. Minitab® version 21.2 (Minitab, LLC, State College, PA) was used for all statistical analyses. Significance was tested at the α = 0.05 level.
During the entire sampling campaign, there were 34 24 h periods starting and ending at midnight for which 24 h averages could be calculated. These daily averages had a GM of 77 μg m−3. The EPA NAAQS PM10 24 h limit of 150 μg m−3 is not to be exceeded more than once per year. However, 6 (18%) of the 24 h averages were over the limit with a maximum of 330 μg m−3.
There were 365 hours during which the BAMs at Sites 1 and 2 were both operational during the same hours to enable background correction of the Site 1 measurements. However, those used for inverse modelling were limited to hours when the wind was from 180° to 270°; that is, when a south-to-southwest wind was blowing from the road toward the Site 1 sampler. PM measurements when south-easterly winds were blowing from 90° to 180° were not used because a north-south unpaved road to the southeast of the sampler at Site 2 could have contributed to particulate concentrations above background levels (Fig. S1†). Hours with no measurable wind were also not used because AERMOD will not provide results for a wind speed of zero.31,32 A total of 64 paired BAM concentration measurements from Sites 1 and 2 met the wind-related selection criteria and were used for the inverse-modeling process. The GM for the emission rate and related emission factor determined by inverse modeling are listed in Table 1. Additional descriptive statistics are provided in Table S2.†
Emission rate (g s−1) | Emission rate/vehicle (g s−1 V−1) | Emission factor (g/VKT) |
---|---|---|
1.09 | 0.26 | 444 |
The GM emission factor predicted by inverse modeling, 444 g/VKT, is approximately half the emission factor calculated using AP-42 methodology, 795 g/VKT. The GSD of the modeled emission factor (4.33) was high, indicating a large amount of variability in inverse-modeled emission rates and indicates the potential for much higher actual values than the GM value. Some of the variability in emission rates may be explained by the short-term variations in wind speed that can be observed under low wind conditions, which may not be adequately accounted for in the AERMOD simulations.38 Many of the observations in this study were recorded under relatively low wind conditions. Furthermore, a prior study estimating the particulate emission rates from paved and unpaved roads by inverse modeling26 found little correlation between the emission factors determined using the model and the emission factors estimated using AP-42 methodology across several study sites, which may have been associated with the high variability found in the current study. O'Shaughnessy and Altmaier36 also obtained an inverse-modeled emission rate for hydrogen sulfide emanating from swine facilities that was lower than those determined from direct emission measurement studies such as the study by Cowherd et al.38 that was used as the basis of the AP-42 method applied in this study. Given a choice between the different emission rates obtained using the two methods, O'Shaughnessy and Altmaier36 suggest that the emission rate developed from inverse modeling using AERMOD may be the most accurate rate to apply when also using AERMOD to model the dispersion of the same contaminant under similar meteorological and terrain conditions.
Table S4† summarizes the sensitivity of the inverse-modeled emission rate to variations in vehicle heights and widths representative of the type of vehicles expected on roads where PM10 samples were collected. Whereas vehicle widths ranging from 1.73 m to 5.03 m had no discernible effect on the resulting emission rate, changes in vehicle height from 1.42 m to 3.81 m increased the emission rate 1.3-fold. The model is therefore more sensitive to changes in vehicle height than changes in width.
Results for the sensitivity analysis of the effects of wind direction and wind speed on the inverse-modeled emission rate are given in Fig. S7.† The emission rate was not sensitive to changes in the wind direction at wind speeds of 1.6 km h−1 and 8 km h−1, which may be explained by the initiation of plume meander when wind speeds are less than 2 m s−1 (7.2 km h−1).38 At 12 km h−1 and 16 km h−1, the emission rate changes with wind direction between 180° to 270°, with a minimum observed at approximately 250° and a maximum at 270°.
For wind blowing at 16 km h−1 directly from the south (Fig. 1B) and southwest (Fig. 1C), PM10 concentrations were higher downwind of the source, as expected, with the concentrations decaying faster at the higher wind speed, also as expected. For wind blowing directly from the west and, therefore, parallel to the road (Fig. 1D), the concentration profiles are nearly symmetrical and centered on the volume line source representing the gravel road. This scenario represents limited dispersion of the PM as the plume follows the axis of the road rather than dispersing predominantly to one side of the road. Of the three wind directions, the highest concentration at the receptor is produced at 225°. In effect, this wind angle moves the high concentrations developed by a nearly parallel wind towards one side of the road, in this case, towards the receptor. Furthermore, the inverse-modeled emission rate for a given measured concentration is lowest near 225° as shown in Fig. S7.† This result implies that, when the same emission rate is applied to the model, the spatial concentrations obtained with a 225° wind are higher than for any other wind angle as shown in Fig. 1.
Average conc. (μg m−3) | Max conc. (μg m−3) | Total mass (μg) | Time to peak (s) | Residence time (s) |
---|---|---|---|---|
4278 | 18743 | 3.4 | 8.2 | 7.4 |
Fig. 2 provides a comparison of the maximum (peak) concentrations and the plume residence duration for two plumes with similar vehicle speeds and wind directions but different wind speeds. The two plots in Fig. 2 demonstrate that the plume generally had a longer decay residence time at low wind speeds, as expected, which results in a longer PM exposure period. Fig. 3 provides probability plots of the average plume concentrations and corresponding peak concentrations, which demonstrate that peak concentrations were typically 4- to 5-fold higher than average concentrations.
![]() | ||
Fig. 2 Comparison of PM plume profiles for two different wind speeds, with approximately the same wind directions, using a direct-reading instrument with one-second logging. |
The wind speed and direction varied somewhat during trial periods with each trial generally lasting 20 to 35 minutes. The mean of the differences between the beginning and ending wind speeds for each trial was 2.3 km h−1, and the mean of the differences between the beginning and ending wind directions for each trial (sample day) was 46.1°.
Vehicle speed was found to be a significant predictor of both maximum and average PM concentration and the total mass, but not for the time-related variables of time to the peak concentration and plume-decay residence time. This suggests that as a vehicle passes the sampler, the time to the maximum concentration and plume residence time will stay relatively constant under constant meteorological conditions with different vehicle speeds, while the average concentration and total PM mass will increase with vehicle speed, therefore, the PM emission rate was also enhanced by vehicle speed. Wind speed was a significant predictor for all dependent variables. However, wind direction was significant only for dust plume residence time, consistent with the finding in our previous research29 This finding suggests the intuitive explanation that the long plume of airborne dust generated by a passing car will be pushed down the road when the wind is parallel to it, thus extending its decay residence time relative to a plume that is pushed laterally by a wind perpendicular to the road.
Results obtained when monitoring the plume generated by a passing vehicle indicated that wind speed and vehicle speed had significant effects on maximum PM and average concentrations, and the PM total mass collected. The direct reading aerosol monitor results demonstrated that the peak concentration of PM10 resulting from the passage of a single vehicle can be over 10000 μg m−3. A short-term concentration level of that magnitude could result in adverse health effects for sensitive individuals.
Footnote |
† Electronic supplementary information (ESI) available. See DOI: https://doi.org/10.1039/d4ea00055b |
This journal is © The Royal Society of Chemistry 2024 |