- Email: [email protected]
Evaluation of the surface roughness effect on suspended particle deposition near unpaved roads
Atmospheric Environment 122 (2015) 541e551
Contents lists available at ScienceDirect
Atmospheric Environment journal homepage: www.elsevier.com/locate/atmosenv
Evaluation of the surface roughness effect on suspended particle deposition near unpaved roads Dongzi Zhu a, *, John A. Gillies a, Vicken Etyemezian b, George Nikolich b, William J. Shaw c a
Desert Research Institute, 2215 Raggio Pkwy, Reno, NV 89512, USA Desert Research Institute, 755 E. Flamingo Road, Las Vegas, NV 89119, USA c Pacific Northwest National Laboratory, 902 Battelle Blvd, Richland, WA 99354, USA b
h i g h l i g h t s Measurement road dust PM10 concentration evolution during near-source transport. Study PM horizontal & vertical transport by array of real-time particle counters. Estimate impacts of the roughness/vegetation on PM deposition under 5 surfaces. Rougher surface induced increased turbulence, thus increased PM deposition. Dense grass site has highest reduction by enhanced particle impaction/interception.
a r t i c l e i n f o
a b s t r a c t
Article history: Received 3 July 2015 Received in revised form 5 October 2015 Accepted 6 October 2015 Available online xxx
The downwind transport and deposition of suspended dust raised by a vehicle driving on unpaved roads was studied for four differently vegetated surfaces in the USA states of Kansas and Washington, and one barren surface in Nevada. A 10 m high tower adjacent to the source (z10 m downwind) and an array of multi-channel optical particle counters at three positions downwind of the source measured the flux of particles and the particle size distribution in the advecting dust plumes in the horizontal and vertical directions. Aerodynamic parameters such as friction velocity (u*) and surface roughness length (z0) were calculated from wind speed measurements made on the tower. Particle number concentration, PM10 mass exhibited an exponential decay along the direction of transport. Coarse particles accounted for z95% of the PM10 mass, at least to a downwind distance of 200 m from the source. PM10 removed by deposition was found to increase with increasing particle size and increasing surface roughness under similar moderate wind speed conditions. The surface of dense, long grass (1.2 m high and complete surface cover) had the greatest reduction of PM10 among the five surfaces tested due to deposition induced by turbulence effects created by the rougher surface and by enhanced particle impaction/ interception effects to the grass blades. © 2015 Elsevier Ltd. All rights reserved.
Keywords: Atmospheric deposition Fugitive dust Near-source deposition Road dust PM10
1. Introduction Unpaved roads are common in rural areas and large military and other installations in the U.S. Unpaved road dust due to surface disturbance by vehicle travel is the highest single emissions category within the non-point fugitive dust category, accounting for about one third of non-windblown fugitive dust emissions in the U.S.A (Pouliot et al., 2012). Williams et al. (2008) found that
* Corresponding author. E-mail address: [email protected] (D. Zhu). http://dx.doi.org/10.1016/j.atmosenv.2015.10.009 1352-2310/© 2015 Elsevier Ltd. All rights reserved.
suspended dust collected at 1.5 m, 4 m and 6 m heights near an unpaved road spanned the size (diameter) range 0.05 mme159 mm with 97.8%e99.6% of particles having diameters <50 mm. The fine part of suspended particulate matter with aerodynamic diameter 10 mm (PM10) is regulated as a primary pollutant in the U.S. due to its potential to adversely affect human health. Fugitive dust raised by vehicles traveling on unpaved surfaces impacts regional and local air quality and visibility (Pinnick et al., 1985). Typically the impact of dust from unpaved roads on air quality is evaluated using emission factors to estimate the mass of particulates that are introduced into the atmosphere by vehicular travel (e.g., US EPA, 1995; Gillies et al., 2010), but what is lacking in this approach is
542
D. Zhu et al. / Atmospheric Environment 122 (2015) 541e551
that all the mass that is estimated to have been ejected at the source is assumed to be available for near- and long-range transport. This neglects the effect of the interaction of the dust plume with the near field (<200 m) environment where it may be interacting with vegetation and roughness resulting in particle removal. Modeling particle deposition has concentrated on two types of flow environments: particle deposition in pipe flows and indoor environments, such as ventilation ducts, and from the atmosphere to the underlying surface. The latter is germane to the question of quantifying deposition of fugitive dust from unpaved roads. For supermicron size particles (diameter 1 mm), studies examining deposition and its variation with surface roughness and meteorological conditions are limited and highly uncertain due to the low concentration of particles detected in this size range in field studies (e.g., Nemitz et al., 2002), and it is typically examined within a framework of far-downwind transport where concentrations are relatively uniform with height and the surface is considered as a sink (e.g., Raupach et al., 2001). Dong et al. (2003) report that overprediction of coarse particles (2.5e10 mm diameter) concentrations of road dust contributions by as much as a factor of 1.7e11 in gridbased dispersion models is due to two factors: 1) the forced mixing of the plume to a minimum thickness of 20 m of height at the source, when it is known to be between 2 m and 6 m depending on vehicle height (Gillies et al., 2005), and 2) the dependency of coarse particle deposition on wind speed not being accounted for adequately. This lack of accounting for near source deposition is a shortcoming in evaluating the contributions of dust to the atmosphere for sources such as unpaved or off-road vehicle travel. Field measurements of dust emissions created by vehicles traveling on unpaved roads are limited (e.g., Johnson et al., 1992; Gillies et al., 2005; Williams et al., 2008) and have typically focused on measuring the total mass flux of particles as a function of vehicle operating parameters and surface conditions to quantify vehicle emission factors. The subsequent transport and deposition of the entrained road dust along the downwind direction over relatively short distances has not been as well studied. When dust particles are suspended by vehicles, they are transported downwind and dispersed by turbulent diffusion. Wilson (2000) using data from Hage (1961) and Walker (1965) demonstrated that particles >50 mm diameter released into the wind quickly settle out in a few hundred meters. When particles in suspension collide with surface roughness elements such as grass, rocks, or buildings, they can deposit (Pardyjak et al., 2008). Particles are brought to these depositional surfaces by Brownian diffusion, impaction, interception, sedimentation, or combinations of these processes. For submicron particles, diffusion dominates; as particle size increases, deposition by sedimentation (i.e., gravity force) and impaction become more important removal processes relative to Brownian diffusion and interception (Friedlander, 2000). Particle deposition depends not only on the particle diameter, but also on parameters like surface roughness and canopy morphology, wind friction velocity (u*), and atmospheric stability (Fowler et al., 2009). Field testing and direct measurement of dust-sized particle (<30 mm diameter) deposition rates is difficult and only a limited number of studies have been reported. Johnson et al. (1992) used light detection and ranging (LIDAR) instrumentation to measure the backscatter coefficient of a depositing dust plume over a relatively smooth surface (an unpaved road) and reported the deposition flux was proportional to the plume concentration as would be expected by the linearity of the advection diffusion equation. Johnson et al. (1992) observed that particles >10 mm almost completely settled out by the time their measured plumes traveled 250 m from the source, and the reduction of the initial mass scaled through time as e1, which under the conditions present during their study accounted for a reduction of initial mass to 36% in
10e40 s Etyemezian et al. (2004) reported that they could not measure depositional losses in vehicle-generated dust plumes traveling 100 m over a surface of sparse vegetation cover at Ft. Bliss near El Paso, TX, for a range of atmospheric stability conditions. Veranth et al. (2003), however, reported quite high removal rates (z85%) of fugitive dust at 95 m downwind of an array of 2.5 m high containers simulating an urban setting. Cowherd et al. (2006) reported different PM10 loss ratios (from <10% to 67%) at 20 m downwind for various vegetation types (cedar trees, tall grass) bordering unpaved road sources. Zhu et al. (2011) reported an exponential decay of PM10 as a function of downwind distance from a paved road source with leafless deciduous birch trees acting as a vegetation barrier under winter conditions. Mao et al. (2013) reported 51% of 20 mm and 61% of 50 mm suspended particles deposited within 60 m downwind of an unpaved road with the plume traveling over a surface composed of the agricultural crop peas and a passing through a shelterbelt of trees approximately 10 m high. These various studies do not reveal a consistent pattern to explain the range of observations of dust deposition that have been reported. Movement of dust laden wind into a vegetated canopy where the flows are generally characterized by strong flow heterogeneity, intermittency, and non-Gaussian flow statistics limits the applicability of simplified modeling approaches (Poggi et al., 2006). This also can create complex vertical profiles of PM that are not easily defined or well-predicted by models. Due to lack of experimental data to distinguish PM loss either from sedimentation or impaction with vegetation downwind, some researchers assume the PM transportable fraction (TF, the fraction of particles not captured by the surrounding land cover) can be used as an index to characterize the magnitude of particle removal from a plume. For example, a plume advecting over short distances could have a TF value of 1 (i.e., no discernable loss) for flat surfaces barren of macro-roughness elements (i.e., no vegetation or large solid elements). Pace (2005) proposed the following TFs for various surface types to improve emission inventory processing: 0.05 for forests, 0.5 for urban, 0.75 for grass, and 1 for bare surfaces. Pace's (2005) assumed TF values for different surfaces have not been validated by field experimentation. Therefore, there is a need to clarify near source dust deposition flux under different vegetative roughness configurations to quantify the deposition losses during the initial transport phase of fugitive dust. Better quantification of the magnitude of the near field depositional flux will improve regional air quality modeling using an emission inventory approach as well as dispersion-modeling approaches that quantify the spatial distribution and magnitude of PM at identified receptor sites. To contribute to improved understanding of the near field deposition process, a series of field measurement campaigns were undertaken to measure in real-time the change in suspended particulate matter (PM) concentration and size distribution in dust plumes advecting downwind following emission from unpaved road sources to infer particle loss downwind and close to the source (<200 m). The measurements were carried out over one nonvegetated and four differently vegetated surfaces and under a restricted range of wind speeds and atmospheric stability conditions. Use of near real-time instruments (sampling interval 1 Hz) allowed for the characterization of multiple, individual, transient dust plumes at each study location. These data can be aggregated into ensemble mean values with associated uncertainties (e.g., mean normalized PM10 and associated standard deviation) for further analyses. The objective of these field studies was to quantify the TF for PM10 as a function of downwind distance for different surface roughness and vegetation types (long grass, short grass, sagebrush, steppe grass, and no vegetation) to evaluate the attenuation of emissions by near-field deposition processes. With dust number concentration as a function of particle (aerodynamic)
D. Zhu et al. / Atmospheric Environment 122 (2015) 541e551
diameter measured at sequential downwind locations, the change in TF could be observed as a function of downwind distance for each vegetative surface cover. A refined box model (Zhu et al., 2011) was used to estimate PM deposition velocities (vd) of different sized particles. The estimates of vd provide a means to parameterize the effect roughness and plant cover type have on particle deposition for dispersion modeling. 2. Experimental methods 2.1. Instrument set up, inter-comparison and measurement uncertainty To evaluate the precision of the instruments used to measure particle (aerodynamic) diameter for multiple size classes, the instruments were first collocated near the roadside at each site prior to the formal field testing. Instruments used for measuring particle number concentration and mass concentration were encased in weatherproof enclosures. Instrumentation in each enclosure consisted of: two DustTraks (Model 8520, TSI, Shoreview, MN) equipped with PM10 and PM2.5 inlets, one Aerosol Profiler (AP) (Model 212-2, MetOne, Grants Pass, OR) for measuring the particle size distribution (PSD) of the suspended PM, one MiniVol sampler (Airmetric, Eugene, OR) collecting PM10 on a 47 mm Teflon filter for gravimetric and chemical analysis, and two deep cycle gel cell batteries (Fig. 1). Particle size cut points for the MetOne AP instrument were programmed at 0.3 mm, 0.5 mm, 0.7 mm, 1 mm, 2 mm, 2.5 mm, 5 mm and 10 mm. MetOne TSP inlets were placed on the sampling tubes for each of the five aerosol samplers. Each enclosure
Fig. 1. Deposition boxes and sampling equipment. Note the anemometer and wind vane are not at the height used during field measurements, which matched the height of the PM inlets.
543
also had one MetOne Model 014A anemometer with a 024A wind vane attached to it on an exterior mast. When an enclosure was resting on the ground, the wind sensors were located z2.2 m above ground level (a.g.l). Each deposition box is capable of operating for approximately 20 h on a battery charge and all environmental data were recorded by a laptop computer inside the enclosure. A deposition box can be configured with aerosol inlets at z2.2 m or z6.2 m a.g.l. Fig. 2 illustrates the eight-channel PM concentrations during a portion of a co-location test that consisted of 20 vehicle passes. Regression analyses were performed to reference all measurements to one unit (i.e., Unit 1). The standard error of the slope and intercept for each regression were calculated and propagated as the uncertainty for all particle count concentrations (Zhu et al., 2011). The slope and intercept correction factors were applied to the measurements from the other four APs (Units 2 to 5) to standardize the readings relative to Profiler Unit 1. The average standard error of the regression was applied to each reading from each particle size bin to infer the precision of the measurements. 2.2. Sampling array at the different test sites Plume-by-plume concentration measurements during vehicle passes at each test site were made using the measurement system described above, but configured differently than that used during the co-location tests. The array of samplers downwind of the source was placed along a line perpendicular to the road beginning with sampler 1 at a near source position (10 or 20 m) with a low level inlet (2.2 m a.g.l.), sampler 2 with a low level inlet (2.2 m a.g.l.) and sampler 3 with a high level inlet (6.2 m a.g.l.) at z110 m downwind of the road, sampler 4 with a low level inlet (2.2 m a.g.l.) and sampler 5 with a high level inlet (6.2 m a.g.l.) at z210 m downwind of the road (Fig. 4). The array of samplers was designed to resolve changes in the particle number concentration and size distribution of the suspended dust in the horizontal and vertical directions as the plume advected downwind. A 10 m high tower configured with five anemometers, one wind vane, five PM10 and three PM2.5 monitors was also placed 10 m downwind of the road to measure the vertical wind speed profile and calculate the mass flux of dust using the method of Gillies et al. (2005). At the Hanford, WA site, a nearby 120 m meteorological tower also provided data for estimating atmospheric stability. Measurement of PM particle number concentration for each surface type at each site was conducted over a minimum of two days. On each day, the test vehicle ran multiple passes (typically 50e70 per day) when the wind was within ±45 of perpendicular to the road. Testing was confined from mid-morning to mid-afternoon. The duration of time that an individual plume interacted with the samplers was dependent on wind speed, but typically lasted between one and 2 min. The surfaces at four sites used to generate the dust plumes were unpaved roads and in all cases the vehicle used to generate the plumes was a 2005 Dodge Sprinter van (weight z2250 kg). At the Ft. Riley, KS test sites (39140 3100 N, 96 560 5400 W) the vegetation downwind of the roadway was initially a continuous cover of dense long grass (z1.2 m high). Subsequent to the first set of tests the grass was mowed to a height of z0.2 m to provide a second, different vegetative cover. At the Hanford, WA site (46 340 0000 N, 119 360 900 W), PM measurements were made for a sagebrush community (z1.5 m high) and over a steppe grass community (z0.4 m high). At the Jean Lake, NV site (35 470 2000 N, 115150 5400 W) PM measurements were made of dust plumes as they advected over a smooth and flat dry lakebed surface devoid of vegetation. Only at the long grass site at Ft. Riley (the first site tested), one deposition box was placed upwind of the source (to measure background conditions) with four measurement locations
544
D. Zhu et al. / Atmospheric Environment 122 (2015) 541e551
Fig. 2. Eight-channel PM concentrations for five samplers during one co-location test.
D. Zhu et al. / Atmospheric Environment 122 (2015) 541e551
545
Fig. 3. Schematic of the near-field dust deposition box model. All C0 s represent dust concentrations generated by the source.
Fig. 4. Schematic diagram of the array of optical particle sizing instruments to measure changes in the particle size distribution of the emitted dust plumes in the horizontal and vertical dimensions.
downwind (10 m, 55 m, 105 m, and 155 m). To improve spatial resolution in the downwind direction the array was reconfigured for the other sites by moving the upwind sampler to a position downwind (z205 m) and calculating background conditions during periods when there were no dust emissions from the source. 2.3. Horizontal gradient method to infer deposition velocity If multiple dust plumes are averaged by particle size bin number concentration, the transient dust source can be approximated as a quasi-steady-state continuous release source and can be analyzed with a box model approach. For this study deposition velocities for particles with diameter >1 mm were calculated using the particle size concentration measurements (#particles m3) of the background air (background, Cbkg) in the periods prior to the presence of a generated dust plume, and then of the dust plume generated by 0 ) at a distance L downwind of the source (C 0 ). The the source (Csrc L two dimensional model (z is vertical, x is horizontal) shown in Fig. 3 depicts a road dust plume transported from left to right at wind speed u (m s1) through a control volume (CV) of height H (m) and length L (m). For the flux calculation, the vertical unit area (cross section of the CV) was assumed to be unit width times H and the horizontal unit area (cross section) was assumed to be unit width times L. For ease of notation, the concentrations solely attributable to the 0 road dust are denoted as: Csrc ¼ Csrc e Cbkg, CL ¼ CL0 e Cbkg, and 0 Cavg ¼ Cavg e Cbkg. The mass balance for the particle concentration in the control volume (i.e., the sum of the fluxes is 0) is:
0 ¼ Fin Fout Fup Fdown ¼ uHCsrc uHCL DL
dCavg vd LCavg dz
(1)
where Fup is the gross particle upward flux, D is the atmospheric
diffusion coefficient (m2 s1), and vd is the deposition velocity (m s1). The model is based on first principles but makes the following three assumptions: Particles less than 0.3 mm have a vd that is negligible with respect to particles >1 mm (i.e.,vd j < 0:3mm < < vd j > 1mm ). Submicron particles are slow to deposit by interception, impaction and/or gravitation compared to particles >1 mm and deposit primarily by Brownian diffusion (Seinfeld and Pandis, 1998). As observed by Zhu et al. (2011), the PM deposition can be approximated by C ¼ Co eekx. The average concentration for each size range in the control volume is determined by integrating Co eekx over the beginning and ending horizontal points of the CV (i.e., x1 and x2) and dividing by the distance (x1 x2) as follows: Cavg ¼ ekx1 ekx2 =kðx1 x2 Þ. Since the horizontal concentration change (dC/dx) is not constant, an exponential approximation is a better representation of a real world deposition scenario than the assumption Cavg ¼ (Csrc þ CL)/2 made previously by Zhu et al. (2011). The relative rate of vertical diffusion is invariant with particle size since the turbulent diffusion coefficient D is much greater than the Brownian diffusion coefficient, for dust particles between 0.3 mm and 20 mm (Seinfeld and Pandis, 1998). Eddy diffusivity can be taken as a linear function of height, D(z) ¼ 0.4 u*z in neutral conditions in the surface layer up to 30e100 m (Fisher, 1978) whereas the Brownian diffusion coefficient is small (of the order of 105 cm2 s1 or less). That is:
D dCavd D dCavd ¼ Cavg dz < 0:3mm Cavg dz > 1:0mm Solving Eq. (1) for vd for particles >1 mm yields:
(2)
546
D. Zhu et al. / Atmospheric Environment 122 (2015) 541e551
vd j > 1:0mm
uH Csrc CL ¼ L C avg
D dCavd Cavg dz > 1:0mm > 1:0mm
(3)
Similarly, solving Eq. (1) for vd for particles <0.3 mm and subtracting this from Eq. (1) yields:
! uH Csrc CL D dCavg vd j >1:0mm vd j <0:3mm ¼ L Cavg >1:0mm Cavg dz >1:0mm ! uH Csrc CL D dCavg L Cavg <0:3mm Cavg dz <0:3mm (4) since vd for particles >1 mm is much larger than particles <0.3 mm, Eq. (2) produces the following equation for vd as a function of particle diameter:
vd j > 1:0mm vd j < 0:3mm zvd ¼
uH L
Csrc CL C avg
> 1:0mm
> 1:0mm
! Csrc CL Cavg < 0:3mm
(5)
3. Results and discussion 3.1. PM10 attenuation downwind of the source u* and z0 were calculated from the vertical wind speed profile data from the near road tower calculated by:
U¼
u* z ln z0 k
(6)
rma n's constant where U is mean wind speed at height z, k is von Ka (0.4). The logarithmic wind profile is valid for the neutral atmosphere; however, for most applications close to the ground (i.e., z 10 m) the wind profile can be assumed almost always to be logarithmic even during non-neutral conditions (Kaimal and Finnigan, 1994). The stability function j(z/Lm) (Lm is MoninObukhov length) correction can be introduced into Eq. (6) for unstable conditions. In our study Lm values were 40 m for Ft. Riley, KS and 75 m for Hanford, WA, which represent unstable atmospheric conditions. The errors associated with the calculated z0 and u* values with the stability correction is within 10% of the log-linear approximation for neutral atmospheric conditions. Average z0, u* and atmospheric Pasquill stability classes for the testing periods and by location are listed in Table 1. The stability classes for the sagebrush and steppe grass at the Hanford site were derived from Richardson number estimates based on meteorological data from the nearby 120 m tower. The long grass vegetation had the highest z0 value at 0.206 m. Sagebrush had second highest value of 0.188 m, the mowed grass 0.025 m, the steppe grass site 0.027 m, and the bare soil 0.007 m. As not all winds came from the direction perpendicular to the road source, the angle of wind approach (AOA) was used to calculate the actual transport distances from source to the sampling positions. If the distance between two measurement positions is Lt, then the actual plume transport distance under an AOA (q) would be Lt/ cosq. Particles in all size classes generally show an exponential decay in particle number concentration with increasing downwind distance. A non-linear least squares regression fitted through the ensemble mean values for each size class from the multiple tests
(without forcing) was used to fit a function to the data and the coefficients of determination, R2 all >0.95 indicate the exponential regression is a strong fit. However, for the dense long grass site (Ft. Riley, KS), the decay is better fit by a power function due to the very rapid change of particle number concentration for all sizes measured in the initial z100 m downwind of the source. The largest fractional decrease in particle number concentration was observed at the 50e100 m position downwind of the source. The decay in number concentration is reflective of the removal of particles from the plume and proportional to plume concentration as previously observed by Johnson et al. (1992). Two causal, and complimentary, mechanisms are proposed that could explain the very rapid change of particle number concentration for the long grass surface. First, this surface has the highest aerodynamic roughness length, z0 (0.206 m). Numeric modeling and measurement of channel flow indicate roughness increases turbulence intensity (Bhaganagar et al., 2004; Kussin and Sommerfeld, 2002). Roth (2000) summarizes several mechanisms by which roughness elements influence turbulence. Roth (2000) notes an intense shear layer is formed near the top of the roughness elements, resulting in higher turbulence intensities; the mixing generated by turbulent wakes behind individual roughness elements, efficiently mix and diffuse momentum or any scalar quantity, the pressure difference across individual roughness elements augments the transport of momentum onto the surfaces. For particles in turbulent flow, turbulence will spread the particles and enhance the deposition rate (Li and Ahmadi, 1992). Pardyjak et al. (2013) report experimental evidence and modeling results that indicate that turbulence enhances particle deposition onto all surfaces of vegetation even for small particle sizes (e.g., PM10) in near source transport. A second mechanism that may be enhancing deposition in the long grass, more than the other vegetation types, is interception/impaction that affects deposition via contact of the particles in the dust plume with the grass blades. Belot and Gauthier (1975) report from wind tunnel experimental results, roughness elements with more surface area have better particle collection efficiency, e.g., blades of grass collect more particles than small leaf shrub types. This also can be invoked to explain why the long grass site shows enhanced particle losses compared to the sagebrush site even though their aerodynamic roughness lengths are similar (Table 1). With particle density assumed to be 2.65 g cm3 (with minerals like silica abundant in soil dust, Tegen et al., 2002), the particle number concentration was converted to a PM10 mass concentration by summing all particles in the size bins less than 10 mm. PM10 mass concentrations were found to decrease at faster rates as function of downwind distance with increased surface roughness for the five surface types tested in the first z100 m from source, again taking into account the actual fetch length depending on AOA q (Fig. 5). For the long grass site at Ft. Riley, KS, PM10 decayed to 13% of the near road concentration at 84 m downwind. For the short grass (mowed long grass), PM10 diminished to 48% of the near road concentration at 130 m downwind. For bare soil without vegetation at the Jean Lake, NV site, PM10 reduced to 67% of the near road concentration at 136 m downwind. Further downwind, the rate of decrease (dC/dx) reversed (decreased) with increased surface roughness, reflecting the proportionally larger decrease in the concentration along the downwind (x) direction. Due to increased turbulence caused by the higher roughness and increased impaction losses associated with the grass blades, the long grass site has the lowest TF for PM10 after 100 m transport. At distances >100 m the rate of decrease of particle number concentration for the long grass site was the lowest observed, while dC/ dx was highest for the non-vegetated lakebed surface.
D. Zhu et al. / Atmospheric Environment 122 (2015) 541e551
547
Table 1 Mean surface aerodynamic roughness length, friction velocity, and stability class for the five different surface sites during testing. The standard deviation of the mean value is provided in parentheses. Vegetation type
z0(m)
No vegetation (Jean Lake, NV) Cut grass (Ft Riley, KS) Steppe grass (Hanford, WA) Sagebrush (Hanford, WA) Dense long grass (Ft Riley, KS)
0.007 0.025 0.027 0.188 0.206
(±0.003) (±0.007) (±0.014) (±0.045) (±0.022)
u* (ms1)
Avg. U at 2.2 m (m s1)
Avg. U at 10 m (m s1)
Stability class
0.31 0.37 0.31 0.34 0.33
2.94 4.19 2.71 3.07 3.21
4.08 4.85 3.15 3.38 3.64
Neutral Mostly neutral Unstable Unstable Unstable to slightly unstable
(±0.03) (±0.03) (±0.06) (±0.04) (±0.04)
(±1.08) (±0.57) (±0.53) (±1.32) (±0.49)
(±1.04) (±0.51) (±0.75) (±1.50) (±0.61)
Fig. 5. Change in downwind PM10 concentration to near source PM10 concentration to for different test surfaces. The y axis is PM10 concentration normalized to the sampler nearest the source. Vertical error bars represent the standard deviation of the mean PM10 ratio for multiple plume passes. The horizontal error bars represent the standard deviation of the mean distance from source to receptor due to plume to plume changes in the AOA.
Fig. 6. Particle deposition velocity based on box model analysis for the five surface types. The vertical error bars are the standard deviation of the mean vd based on multiple plume passes. Values on the x-axis are particle geometric mean diameters of each size bin.
548
D. Zhu et al. / Atmospheric Environment 122 (2015) 541e551
Fig. 7. Ratio of vd over wind speed the for five surface types. The vertical error bars are the standard deviation of the mean vd/U based on multiple plume passes.
3.2. Size-specific PM deposition velocities Size-specific mean PM deposition velocities inferred from box model (Eq. (5)) calculations are presented in Fig. 6. For comparison, the terminal settling velocity vt was also calculated for the same sized particles using the Stokes Eq.:
vt ¼
2 1 Dp rp gCc 18 m
(7)
where Dp is the particle diameter, rp is the particle density, g is the gravitational constant (9.8 m s2), Cc is the Cunningham Slip Correction factor, and m is the dynamic viscosity of the atmosphere (1.7 105 kg m1 s1). Fig. 6 illustrates that the mean particle deposition velocity increases with particle size (geometric mean of the particle bin end sizes), as expected, and ranges from z1 cm s1 for particles <1 mm to z10 cm s1 for particles >10 mm. The deposition velocity is generally higher for long grass vegetation with greater roughness even though the average wind speed was higher for the short grass tests. This is due to higher surface
Fig. 8. Deposition velocities for particles measured during a single plume passage with similar wind speed for five surface types.
D. Zhu et al. / Atmospheric Environment 122 (2015) 541e551
roughness causing greater reduction in the number (or mass) concentration of PM, which then results in higher vd. A plot of mean vd/U versus particle diameter is shown in Fig. 7. It indicates that after the wind speed normalization, the ratio vd/U is highest for the aerodynamically roughest surface. This is due to increased turbulence induced by the larger and more-dense cover of roughness elements. Another factor enhancing the magnitude of vd on this surface type is demonstrated in Fig. 8, which compares vd values from a
549
single pass for the five vegetation types tested under similar mean wind speeds. vd calculated for particles in the turbulent surface layer, as expected, is higher than the Stokes setting velocities for same-sized particles According to Pardyjak et al. (2013), this can be explained with a combination of the classical Stokes number (Stk) and the Taylor-microscale Reynolds number (Rel), where particle impaction on to the vegetative elements is enhanced by increased turbulence intensity created by the roughness that transfers additional momentum to the particles resulting in an enhanced ability
Fig. 9. An example of suspended PM number size distribution (left column) and mass size distribution (right column) from samplers at 10 m, 100 m, and 200 m downwind of the roadway source.
550
D. Zhu et al. / Atmospheric Environment 122 (2015) 541e551
to deposit. In addition, plant morphology may also be playing a role as the surface area of the long grass blades and their increased density compared to the other vegetation types offers greater opportunities for the critical interactions to occur. 3.3. Changes in particle size with downwind distance Examples of the downwind particle number and mass size distribution for a single test run on the long grass surface are shown in Fig. 9. The coarse particles (between 2.5 and 10 mm) accounted for z20% of the PM10 number concentration at the three downwind measurement locations at Ft. Riley, but accounted for z95% of the PM10 mass concentration at the three downwind distances for the other sites. The coarse PM mass dominance was observed for all surface types, which indicates coarse particles dominate resuspended dust mass created by the vehicle travel on all the unpaved test roads. The concentration measurements made at the lower inlet height (2.2 m a.g.l.) at the z100 m and z200 m downwind positions, compared with number concentrations at the same size bin from the sampler with the low elevation inlet at the near road position (i.e., 10 m downwind of the source), show that larger particles decrease at a faster rate than smaller particles not only in horizontal direction, but also in the vertical direction (Table 2). For example, under the short grass condition at Ft. Riley, KS, the number concentration measured at 6.2 m a.g.l. at the 105 m and 205 m downwind sites normalized to the corresponding size-bin number concentration at the 2.2 m a.g.l. sampling height 20 m downwind measurement position, show a greater fractional reduction in 10 mm particles at 205 m (0.01) than at 105 m (0.042). This pattern is produced across all sites (Table 2). Similar trends of change in PM mass concentration in the vertical direction and with increasing plume travel distance are reported by other researchers (e.g., Chan €ll et al., 2003). and Kwok, 2000; Janha The mass median diameter (MMD) is a useful statisticallyderived parameter for characterizing a distribution of suspended particles to compare and contrast differences between the distributions resulting from changes due to aging of the plume mass as it is transported. For example, an MMD of 10 mm means that 50% of the total sample mass will be present in particles having diameters <10 mm, and 50% of the total sample mass will be present in particles having a diameter >10 mm. For the Ft. Riley short grass surface, the MMDs of the dust plume were 11 mm, 9.6 mm, and 9.0 mm measured at the low sampler inlets at 20 m, 105 m, and 205 m downwind, respectively, and 9 mm and 7.5 mm at the high sampler inlets at 105 m and 205 m downwind. At the bare, smooth Jean Lake site the MMDs were 16 mm, 16 mm, and 15.2 mm at the low sampler inlets at 10 m, 110 m, and 210 m downwind, respectively, and 13.2 mm and 13 mm from the high inlets at 110 m and 210 m
downwind. The difference of MMD at the first downwind sampler between the two sites (Ft. Riley vs. Jean Lake) is due to the different particle size distribution of the source material at each site. The trend of MMD decreasing horizontally and vertically as a function of downwind distance however is similar for the different surface conditions, which as expected indicates the larger particles were deposited and removed from the dust plume at a faster rate than small particles during downwind transport, which provides confidence that our data set has captured the process of interest. 4. Conclusions Near source PM reduction for different surface roughness conditions was characterized by the TF of PM10 and inferred vd using particle number concentration measurements and application of a horizontal gradient box model (for vd). The dust plumes were created by wheeled vehicles traveling at a constant speed on unpaved roads at test sites in Kansas and Washington, and a playa surface in Nevada. These three locations provided five different surfaces, one smooth and four with various types and covers of vegetation to estimate the impact of the roughness/vegetation on PM deposition close to the source (0e200 m). Coarse particles accounted for z95% of the PM10 for four of the sites (short grass, steppe grass, sagebrush, and barren) at three downwind locations between z100 and z200 m downwind, with the exception being the tall grass site where coarse particles comprised z20% of the PM10 over this same distance range, which was due to the rapid change between z10 m and z50 m. PM10 removal increased with increasing surface and aerodynamic roughness. The dense, high grass surface had the highest reduction of PM10, which is attributed to increased deposition due to increased turbulence induced by the rougher surface and by enhanced particle impaction/interception effects with the blades of grass. These data also suggest that the type of vegetation exerts some control on the depositional process as the long grass site had higher PM losses than the sagebrush site even though their aerodynamic roughness, and hence drag characteristics, were not that dissimilar. Estimated deposition velocities ranged from z1 cm s1 for particles <1 mm to z10 cm s1 for particles >10 mm. The estimated vd values were found to increase with particle size and surface roughness under similar moderate wind speed conditions. Compared to the Stokes settling velocity vd was increased by a factor of z2 (steppe grass) to z4 (long grass) for the largest measured particle size depending on the surface type. Large particle mass number concentrations were observed to decrease at a faster rate not only in horizontal direction, but also in the vertical dimension as the dust plume was transported downwind. The mass median diameter of the dust plume decreased along both the horizontal and vertical dimensions, at rates greater than can be
Table 2 Size-segregated survival rate of PM number concentration after transport normalized to 20 m downwind 2.2 m a.g.l. inlet readings for the short grass surface at Ft. Riley, KS. Survival rates along the horizontal direction (same 2.2 m a.g.l. inlet readings at 105 m and 205 m)
Kansas short grass site Nevada No vegetation site Washington Sagebrush site Washington Steppe grass site
PM size bins 0.5e0.7 mm Downwind distances 105 m 0.85 205 m 0.12 105 m 0.84 205 m 0.39 105 m 0.29 205 m 0.041 105 m 0.72 205 m 0.114
Note: No high inlets were set for the long grass site, Ft Riley, KS.
Survival rates along the vertical direction (6.2 m a.g.l. inlet readings at 105 m and 205 m)
2.5e5 mm
5e10 mm
>10 mm
0.5e0.7 mm
2.5e5 mm
5e10 mm
>10 mm
0.47 0.08 0.43 0.36 0.26 0.037 0.62 0.059
0.47 0.065 0.37 0.31 0.22 0.036 0.51 0.049
0.22 0.025 0.13 0.07 0.11 0.030 0.35 0.042
0.80 0.05 0.74 0.30 0.18 0.028 0.37 0.086
0.35 0.03 0.32 0.24 0.14 0.025 0.29 0.038
0.116 0.02 0.19 0.17 0.12 0.023 0.26 0.029
0.042 0.01 0.10 0.05 0.08 0.014 0.20 0.02
D. Zhu et al. / Atmospheric Environment 122 (2015) 541e551
attributed to simply gravitational sedimentation modifying the plume PSD. These results also suggest a reason for the poor reconciliation between emission inventory estimates of unpaved road dust and the amount of mineral dust observed in samples collected in air quality sampling networks. The inventory approach may be overestimating contributions because it is based on only near source emission fluxes that do not take into account the depositional losses in the near field that were clearly affected to a significant degree by the type of surface the dust plumes interacted with as they advected downwind. Acknowledgments This work was sponsored by the Strategic Environmental Research and Development Program (Project RC-1729). Many thanks to those who participated in the field work from the Pacific Northwest National Laboratory, Richland WA. We would also like to thank U.S. Army, Ft. Riley, KS for their logistical support during the field measurement campaign there. References Belot, Y., Gauthier, D., 1975. Transport of micronic particles from atmosphere to foliar surfaces. In: Heat Mass Transf. Biosphere, 2 (4). Scripta Book, Co., Washington DC, p. 31. Bhaganagar, K., Kim, J., Coleman, G., 2004. Effect of roughness on wall-bounded turbulence. Flow Turbul. Combust. 72 (2e4), 463e492. Chan, L.Y., Kwok, W.S., 2000. Vertical dispersion of suspended particulates in urban area of Hong Kong. Atmos. Environ. 34, 4403e4412. Cowherd, C., Muleski, G., Gebhart, D.L., 2006. Development of an emission reduction term for near-source dust depletion. In: Conference Proceedings, 15th International Emission Inventory Conference: “Reinventing Inventories e New Ideas in New Orleans”, New Orleans, LA, May, 2006. Dong, Y., Hardy, R., McGown, M., 2003. Why road dust concentrations are overestimated in Eulerian grid models. In: Appendix K, Northern Ada County (Idaho) PM10 Maintenance Plan. Idaho Department of Environmental Quality. June 2003. http://www.cmascenter.org/conference/2004/abstracts/poster/ dong_abstract.DOC. Etyemezian, V., Gillies, J., Kuhns, H., Gillette, D., Ahonen, S., Nikolic, D., Veranth, J., 2004. Deposition and removal of fugitive dust in the arid southwestern United States: measurements and model results. J. Air & Waste Manage. Assoc. 54, 1099e1111. Fisher, B.E.A., 1978. The calculation of long term sulphur deposition in Europe. Atmos. Environ. 12, 489e501. Fowler, D., Pilegaard, K., Sutton, M.A., Ambus, P., Raivonen, M., Duyzer, J., Simpson, D., et al., 2009. Atmospheric composition change: ecosystemseatmosphere interactions. Atmos. Environ. 43, 5193e5267. Friedlander, S.K., 2000. Smoke, Dust, and Haze, vol. 198. Oxford University Press, New York. Gillies, J.A., Etyemezian, V., Kuhns, H., Nikolic, D., Gillette, D.A., 2005. Effect of vehicle characteristics on unpaved road dust emissions. Atmos. Environ. 39, 2341e2347. Gillies, J.A., Etyemezian, V., Kuhns, H., McAlpine, J.D., King, J., Uppapalli, S., Nikolich, G., Engelbrecht, J., 2010. Dust emissions created by low-level rotarywinged aircraft flight over desert surfaces. Atmos. Environ. 44, 1043e1053.
551
Hage, K.D., 1961. On the dispersion of large particles from a 15-m source in the atmosphere. J. Meteor. 18, 534e539. €ll, S., Molna r, P., Hallquist, M., 2003. Vertical distribution of air pollutants at Janha €teborg, Sweden. Atmos. Environ. 37 (2), 209e217. the Gustavii Cathedral in Go Johnson, T.C., Gillette, D.A., Schwiesow, R.L., 1992. Fate of dust particles from unpaved road under various atmospheric conditions. In: Precipitation Scavenging and Atmosphere-surface Exchange, pp. 933e948. Kaimal, J.C., Finnigan, J.J., 1994. Atmospheric Boundary Layer Flows: Their Structure and Measurement. Oxford University Press, New York, NY. Kussin, J., Sommerfeld, M., 2002. Experimental studies on particle behaviour and turbulence modification in horizontal channel flow with different wall roughness. Exp. Fluids 33 (1), 143e159. Li, A., Ahmadi, G., 1992. Dispersion and deposition of spherical particles from point sources in a turbulent channel flow. Aerosol Sci. Technol. 16 (4), 209e226. Mao, Y., Wilson, J.D., Kort, J., 2013. Effects of a shelterbelt on road dust dispersion. Atmos. Environ. 79, 590e598. Nemitz, E., Gallagher, M.W., Duyzer, J.H., Fowler, D., 2002. Micrometeorological measurements of particle deposition velocities to moorland vegetation. Quart. J. Roy. Meteor. Soc. 128 (585), 2281e2300. Pace, T.G., 2005. Methodology to Estimate the Transportable Fraction (TF) of Fugitive Dust Emissions for Regional and Urban Scale Air Quality Analyses. U.S. EPA, Research Triangle Park. August 2005. http://www.epa.gov/ttnchie1/emch/ dustfractions/transportable_fraction_080305_rev.pdf. Pardyjak, E.R., Speckart, S.O., Yin, F., Veranth, J.M., 2008. Near source deposition of vehicle generated fugitive dust on vegetation and buildings: model development and theory. Atmos. Environ. 42, 6442e6452. Pardyjak, E., Veranth, J., Speckart, S., Moran, S., Price, T., 2013. Development of a Windbreak Dust Predictive Model and Mitigation Planning Tool. Final Report. SERDP Project RC-1730. University of Utah, Salt Lake City, UT. Pinnick, R.G., Fernandez, G., Hinds, B.D., Bruce, C.W., Schaefer, R.W., Pendleton, J.D., 1985. Dust generated by vehicular traffic on unpaved roadways: sizes and infrared extinction characteristics. Aerosol Sci. Technol. 4, 99e121. Poggi, D., Katul, G.G., Albertson, J.D., 2006. Scalar dispersion within a model canopy: measurements and three-dimensional Lagrangian models. Adv. Water Res. 29, 326e335. Pouliot, G., Simon, H., Bhave, P., Tong, D., Mobley, D., Pace, T., Pierce, T., 2012. Assessing the anthropogenic fugitive dust emission inventory and temporal allocation using an updated speciation of particulate matter. Air Pollut. Model. Appl. XXI 4, 585e589. January 01, 2012. Raupach, M.R., Woods, N., Dorr, G., Leys, J.F., Cleugh, H.A., 2001. The entrapment of particles by windbreaks. Atmos. Environ. 35, 3373e3383. Roth, M., 2000. Review of atmospheric turbulence over cities. Quart. J. Roy. Meteor. Soc. 126 (564), 941e990. Seinfeld, J.H., Pandis, S.N., 1998. Atmospheric Chemistry and Physics: from Air Pollution to Climate Change. Wiley, Hoboken, NJ, p. 1203. Tegen, I., Harrison, S.P., Kohfeld, K., Prentice, I.C., Coe, M., Heimann, M., 2002. Impact of vegetation and preferential source areas on global dust aerosol: results from a model study. J. Geophys. Res. 107 (D21), 4576. US EPA, 1995. Compilation of Air Pollutant Emission Factors Volume I: Stationary, Point and Area Sources. US EPA Office of Air Quality Planning and Standards, Research Triangle Park, NC. Veranth, J.M., Pardyjak, E., Seshadri, G., 2003. Vehicle-generated fugitive dust transport: analytic models and field study. Atmos. Environ. 37, 2295e2303. Walker, E.R., 1965. A particulate diffusion experiment. J. Appl. Meterol. 4, 614e621. Williams, D.S., Shukla, M.K., Ross, J., 2008. Particulate matter emission by a vehicle running on unpaved road. Atmos. Environ. 42 (16), 3899e3905. Wilson, J.D., 2000. Trajectory models for heavy particles in atmospheric turbulence: comparison with observations. J. Appl. Meterol. 39 (11), 1894e1912. Zhu, D., Kuhns, H., Gillies, J.A., Etyemezian, V., Gertler, A., Brown, S., 2011. Inferring deposition velocities from changes in aerosol size distributions downwind of a roadway. Atmos. Environ. 45, 957e966.
Contents lists available at ScienceDirect
Atmospheric Environment journal homepage: www.elsevier.com/locate/atmosenv
Evaluation of the surface roughness effect on suspended particle deposition near unpaved roads Dongzi Zhu a, *, John A. Gillies a, Vicken Etyemezian b, George Nikolich b, William J. Shaw c a
Desert Research Institute, 2215 Raggio Pkwy, Reno, NV 89512, USA Desert Research Institute, 755 E. Flamingo Road, Las Vegas, NV 89119, USA c Pacific Northwest National Laboratory, 902 Battelle Blvd, Richland, WA 99354, USA b
h i g h l i g h t s Measurement road dust PM10 concentration evolution during near-source transport. Study PM horizontal & vertical transport by array of real-time particle counters. Estimate impacts of the roughness/vegetation on PM deposition under 5 surfaces. Rougher surface induced increased turbulence, thus increased PM deposition. Dense grass site has highest reduction by enhanced particle impaction/interception.
a r t i c l e i n f o
a b s t r a c t
Article history: Received 3 July 2015 Received in revised form 5 October 2015 Accepted 6 October 2015 Available online xxx
The downwind transport and deposition of suspended dust raised by a vehicle driving on unpaved roads was studied for four differently vegetated surfaces in the USA states of Kansas and Washington, and one barren surface in Nevada. A 10 m high tower adjacent to the source (z10 m downwind) and an array of multi-channel optical particle counters at three positions downwind of the source measured the flux of particles and the particle size distribution in the advecting dust plumes in the horizontal and vertical directions. Aerodynamic parameters such as friction velocity (u*) and surface roughness length (z0) were calculated from wind speed measurements made on the tower. Particle number concentration, PM10 mass exhibited an exponential decay along the direction of transport. Coarse particles accounted for z95% of the PM10 mass, at least to a downwind distance of 200 m from the source. PM10 removed by deposition was found to increase with increasing particle size and increasing surface roughness under similar moderate wind speed conditions. The surface of dense, long grass (1.2 m high and complete surface cover) had the greatest reduction of PM10 among the five surfaces tested due to deposition induced by turbulence effects created by the rougher surface and by enhanced particle impaction/ interception effects to the grass blades. © 2015 Elsevier Ltd. All rights reserved.
Keywords: Atmospheric deposition Fugitive dust Near-source deposition Road dust PM10
1. Introduction Unpaved roads are common in rural areas and large military and other installations in the U.S. Unpaved road dust due to surface disturbance by vehicle travel is the highest single emissions category within the non-point fugitive dust category, accounting for about one third of non-windblown fugitive dust emissions in the U.S.A (Pouliot et al., 2012). Williams et al. (2008) found that
* Corresponding author. E-mail address: [email protected] (D. Zhu). http://dx.doi.org/10.1016/j.atmosenv.2015.10.009 1352-2310/© 2015 Elsevier Ltd. All rights reserved.
suspended dust collected at 1.5 m, 4 m and 6 m heights near an unpaved road spanned the size (diameter) range 0.05 mme159 mm with 97.8%e99.6% of particles having diameters <50 mm. The fine part of suspended particulate matter with aerodynamic diameter 10 mm (PM10) is regulated as a primary pollutant in the U.S. due to its potential to adversely affect human health. Fugitive dust raised by vehicles traveling on unpaved surfaces impacts regional and local air quality and visibility (Pinnick et al., 1985). Typically the impact of dust from unpaved roads on air quality is evaluated using emission factors to estimate the mass of particulates that are introduced into the atmosphere by vehicular travel (e.g., US EPA, 1995; Gillies et al., 2010), but what is lacking in this approach is
542
D. Zhu et al. / Atmospheric Environment 122 (2015) 541e551
that all the mass that is estimated to have been ejected at the source is assumed to be available for near- and long-range transport. This neglects the effect of the interaction of the dust plume with the near field (<200 m) environment where it may be interacting with vegetation and roughness resulting in particle removal. Modeling particle deposition has concentrated on two types of flow environments: particle deposition in pipe flows and indoor environments, such as ventilation ducts, and from the atmosphere to the underlying surface. The latter is germane to the question of quantifying deposition of fugitive dust from unpaved roads. For supermicron size particles (diameter 1 mm), studies examining deposition and its variation with surface roughness and meteorological conditions are limited and highly uncertain due to the low concentration of particles detected in this size range in field studies (e.g., Nemitz et al., 2002), and it is typically examined within a framework of far-downwind transport where concentrations are relatively uniform with height and the surface is considered as a sink (e.g., Raupach et al., 2001). Dong et al. (2003) report that overprediction of coarse particles (2.5e10 mm diameter) concentrations of road dust contributions by as much as a factor of 1.7e11 in gridbased dispersion models is due to two factors: 1) the forced mixing of the plume to a minimum thickness of 20 m of height at the source, when it is known to be between 2 m and 6 m depending on vehicle height (Gillies et al., 2005), and 2) the dependency of coarse particle deposition on wind speed not being accounted for adequately. This lack of accounting for near source deposition is a shortcoming in evaluating the contributions of dust to the atmosphere for sources such as unpaved or off-road vehicle travel. Field measurements of dust emissions created by vehicles traveling on unpaved roads are limited (e.g., Johnson et al., 1992; Gillies et al., 2005; Williams et al., 2008) and have typically focused on measuring the total mass flux of particles as a function of vehicle operating parameters and surface conditions to quantify vehicle emission factors. The subsequent transport and deposition of the entrained road dust along the downwind direction over relatively short distances has not been as well studied. When dust particles are suspended by vehicles, they are transported downwind and dispersed by turbulent diffusion. Wilson (2000) using data from Hage (1961) and Walker (1965) demonstrated that particles >50 mm diameter released into the wind quickly settle out in a few hundred meters. When particles in suspension collide with surface roughness elements such as grass, rocks, or buildings, they can deposit (Pardyjak et al., 2008). Particles are brought to these depositional surfaces by Brownian diffusion, impaction, interception, sedimentation, or combinations of these processes. For submicron particles, diffusion dominates; as particle size increases, deposition by sedimentation (i.e., gravity force) and impaction become more important removal processes relative to Brownian diffusion and interception (Friedlander, 2000). Particle deposition depends not only on the particle diameter, but also on parameters like surface roughness and canopy morphology, wind friction velocity (u*), and atmospheric stability (Fowler et al., 2009). Field testing and direct measurement of dust-sized particle (<30 mm diameter) deposition rates is difficult and only a limited number of studies have been reported. Johnson et al. (1992) used light detection and ranging (LIDAR) instrumentation to measure the backscatter coefficient of a depositing dust plume over a relatively smooth surface (an unpaved road) and reported the deposition flux was proportional to the plume concentration as would be expected by the linearity of the advection diffusion equation. Johnson et al. (1992) observed that particles >10 mm almost completely settled out by the time their measured plumes traveled 250 m from the source, and the reduction of the initial mass scaled through time as e1, which under the conditions present during their study accounted for a reduction of initial mass to 36% in
10e40 s Etyemezian et al. (2004) reported that they could not measure depositional losses in vehicle-generated dust plumes traveling 100 m over a surface of sparse vegetation cover at Ft. Bliss near El Paso, TX, for a range of atmospheric stability conditions. Veranth et al. (2003), however, reported quite high removal rates (z85%) of fugitive dust at 95 m downwind of an array of 2.5 m high containers simulating an urban setting. Cowherd et al. (2006) reported different PM10 loss ratios (from <10% to 67%) at 20 m downwind for various vegetation types (cedar trees, tall grass) bordering unpaved road sources. Zhu et al. (2011) reported an exponential decay of PM10 as a function of downwind distance from a paved road source with leafless deciduous birch trees acting as a vegetation barrier under winter conditions. Mao et al. (2013) reported 51% of 20 mm and 61% of 50 mm suspended particles deposited within 60 m downwind of an unpaved road with the plume traveling over a surface composed of the agricultural crop peas and a passing through a shelterbelt of trees approximately 10 m high. These various studies do not reveal a consistent pattern to explain the range of observations of dust deposition that have been reported. Movement of dust laden wind into a vegetated canopy where the flows are generally characterized by strong flow heterogeneity, intermittency, and non-Gaussian flow statistics limits the applicability of simplified modeling approaches (Poggi et al., 2006). This also can create complex vertical profiles of PM that are not easily defined or well-predicted by models. Due to lack of experimental data to distinguish PM loss either from sedimentation or impaction with vegetation downwind, some researchers assume the PM transportable fraction (TF, the fraction of particles not captured by the surrounding land cover) can be used as an index to characterize the magnitude of particle removal from a plume. For example, a plume advecting over short distances could have a TF value of 1 (i.e., no discernable loss) for flat surfaces barren of macro-roughness elements (i.e., no vegetation or large solid elements). Pace (2005) proposed the following TFs for various surface types to improve emission inventory processing: 0.05 for forests, 0.5 for urban, 0.75 for grass, and 1 for bare surfaces. Pace's (2005) assumed TF values for different surfaces have not been validated by field experimentation. Therefore, there is a need to clarify near source dust deposition flux under different vegetative roughness configurations to quantify the deposition losses during the initial transport phase of fugitive dust. Better quantification of the magnitude of the near field depositional flux will improve regional air quality modeling using an emission inventory approach as well as dispersion-modeling approaches that quantify the spatial distribution and magnitude of PM at identified receptor sites. To contribute to improved understanding of the near field deposition process, a series of field measurement campaigns were undertaken to measure in real-time the change in suspended particulate matter (PM) concentration and size distribution in dust plumes advecting downwind following emission from unpaved road sources to infer particle loss downwind and close to the source (<200 m). The measurements were carried out over one nonvegetated and four differently vegetated surfaces and under a restricted range of wind speeds and atmospheric stability conditions. Use of near real-time instruments (sampling interval 1 Hz) allowed for the characterization of multiple, individual, transient dust plumes at each study location. These data can be aggregated into ensemble mean values with associated uncertainties (e.g., mean normalized PM10 and associated standard deviation) for further analyses. The objective of these field studies was to quantify the TF for PM10 as a function of downwind distance for different surface roughness and vegetation types (long grass, short grass, sagebrush, steppe grass, and no vegetation) to evaluate the attenuation of emissions by near-field deposition processes. With dust number concentration as a function of particle (aerodynamic)
D. Zhu et al. / Atmospheric Environment 122 (2015) 541e551
diameter measured at sequential downwind locations, the change in TF could be observed as a function of downwind distance for each vegetative surface cover. A refined box model (Zhu et al., 2011) was used to estimate PM deposition velocities (vd) of different sized particles. The estimates of vd provide a means to parameterize the effect roughness and plant cover type have on particle deposition for dispersion modeling. 2. Experimental methods 2.1. Instrument set up, inter-comparison and measurement uncertainty To evaluate the precision of the instruments used to measure particle (aerodynamic) diameter for multiple size classes, the instruments were first collocated near the roadside at each site prior to the formal field testing. Instruments used for measuring particle number concentration and mass concentration were encased in weatherproof enclosures. Instrumentation in each enclosure consisted of: two DustTraks (Model 8520, TSI, Shoreview, MN) equipped with PM10 and PM2.5 inlets, one Aerosol Profiler (AP) (Model 212-2, MetOne, Grants Pass, OR) for measuring the particle size distribution (PSD) of the suspended PM, one MiniVol sampler (Airmetric, Eugene, OR) collecting PM10 on a 47 mm Teflon filter for gravimetric and chemical analysis, and two deep cycle gel cell batteries (Fig. 1). Particle size cut points for the MetOne AP instrument were programmed at 0.3 mm, 0.5 mm, 0.7 mm, 1 mm, 2 mm, 2.5 mm, 5 mm and 10 mm. MetOne TSP inlets were placed on the sampling tubes for each of the five aerosol samplers. Each enclosure
Fig. 1. Deposition boxes and sampling equipment. Note the anemometer and wind vane are not at the height used during field measurements, which matched the height of the PM inlets.
543
also had one MetOne Model 014A anemometer with a 024A wind vane attached to it on an exterior mast. When an enclosure was resting on the ground, the wind sensors were located z2.2 m above ground level (a.g.l). Each deposition box is capable of operating for approximately 20 h on a battery charge and all environmental data were recorded by a laptop computer inside the enclosure. A deposition box can be configured with aerosol inlets at z2.2 m or z6.2 m a.g.l. Fig. 2 illustrates the eight-channel PM concentrations during a portion of a co-location test that consisted of 20 vehicle passes. Regression analyses were performed to reference all measurements to one unit (i.e., Unit 1). The standard error of the slope and intercept for each regression were calculated and propagated as the uncertainty for all particle count concentrations (Zhu et al., 2011). The slope and intercept correction factors were applied to the measurements from the other four APs (Units 2 to 5) to standardize the readings relative to Profiler Unit 1. The average standard error of the regression was applied to each reading from each particle size bin to infer the precision of the measurements. 2.2. Sampling array at the different test sites Plume-by-plume concentration measurements during vehicle passes at each test site were made using the measurement system described above, but configured differently than that used during the co-location tests. The array of samplers downwind of the source was placed along a line perpendicular to the road beginning with sampler 1 at a near source position (10 or 20 m) with a low level inlet (2.2 m a.g.l.), sampler 2 with a low level inlet (2.2 m a.g.l.) and sampler 3 with a high level inlet (6.2 m a.g.l.) at z110 m downwind of the road, sampler 4 with a low level inlet (2.2 m a.g.l.) and sampler 5 with a high level inlet (6.2 m a.g.l.) at z210 m downwind of the road (Fig. 4). The array of samplers was designed to resolve changes in the particle number concentration and size distribution of the suspended dust in the horizontal and vertical directions as the plume advected downwind. A 10 m high tower configured with five anemometers, one wind vane, five PM10 and three PM2.5 monitors was also placed 10 m downwind of the road to measure the vertical wind speed profile and calculate the mass flux of dust using the method of Gillies et al. (2005). At the Hanford, WA site, a nearby 120 m meteorological tower also provided data for estimating atmospheric stability. Measurement of PM particle number concentration for each surface type at each site was conducted over a minimum of two days. On each day, the test vehicle ran multiple passes (typically 50e70 per day) when the wind was within ±45 of perpendicular to the road. Testing was confined from mid-morning to mid-afternoon. The duration of time that an individual plume interacted with the samplers was dependent on wind speed, but typically lasted between one and 2 min. The surfaces at four sites used to generate the dust plumes were unpaved roads and in all cases the vehicle used to generate the plumes was a 2005 Dodge Sprinter van (weight z2250 kg). At the Ft. Riley, KS test sites (39140 3100 N, 96 560 5400 W) the vegetation downwind of the roadway was initially a continuous cover of dense long grass (z1.2 m high). Subsequent to the first set of tests the grass was mowed to a height of z0.2 m to provide a second, different vegetative cover. At the Hanford, WA site (46 340 0000 N, 119 360 900 W), PM measurements were made for a sagebrush community (z1.5 m high) and over a steppe grass community (z0.4 m high). At the Jean Lake, NV site (35 470 2000 N, 115150 5400 W) PM measurements were made of dust plumes as they advected over a smooth and flat dry lakebed surface devoid of vegetation. Only at the long grass site at Ft. Riley (the first site tested), one deposition box was placed upwind of the source (to measure background conditions) with four measurement locations
544
D. Zhu et al. / Atmospheric Environment 122 (2015) 541e551
Fig. 2. Eight-channel PM concentrations for five samplers during one co-location test.
D. Zhu et al. / Atmospheric Environment 122 (2015) 541e551
545
Fig. 3. Schematic of the near-field dust deposition box model. All C0 s represent dust concentrations generated by the source.
Fig. 4. Schematic diagram of the array of optical particle sizing instruments to measure changes in the particle size distribution of the emitted dust plumes in the horizontal and vertical dimensions.
downwind (10 m, 55 m, 105 m, and 155 m). To improve spatial resolution in the downwind direction the array was reconfigured for the other sites by moving the upwind sampler to a position downwind (z205 m) and calculating background conditions during periods when there were no dust emissions from the source. 2.3. Horizontal gradient method to infer deposition velocity If multiple dust plumes are averaged by particle size bin number concentration, the transient dust source can be approximated as a quasi-steady-state continuous release source and can be analyzed with a box model approach. For this study deposition velocities for particles with diameter >1 mm were calculated using the particle size concentration measurements (#particles m3) of the background air (background, Cbkg) in the periods prior to the presence of a generated dust plume, and then of the dust plume generated by 0 ) at a distance L downwind of the source (C 0 ). The the source (Csrc L two dimensional model (z is vertical, x is horizontal) shown in Fig. 3 depicts a road dust plume transported from left to right at wind speed u (m s1) through a control volume (CV) of height H (m) and length L (m). For the flux calculation, the vertical unit area (cross section of the CV) was assumed to be unit width times H and the horizontal unit area (cross section) was assumed to be unit width times L. For ease of notation, the concentrations solely attributable to the 0 road dust are denoted as: Csrc ¼ Csrc e Cbkg, CL ¼ CL0 e Cbkg, and 0 Cavg ¼ Cavg e Cbkg. The mass balance for the particle concentration in the control volume (i.e., the sum of the fluxes is 0) is:
0 ¼ Fin Fout Fup Fdown ¼ uHCsrc uHCL DL
dCavg vd LCavg dz
(1)
where Fup is the gross particle upward flux, D is the atmospheric
diffusion coefficient (m2 s1), and vd is the deposition velocity (m s1). The model is based on first principles but makes the following three assumptions: Particles less than 0.3 mm have a vd that is negligible with respect to particles >1 mm (i.e.,vd j < 0:3mm < < vd j > 1mm ). Submicron particles are slow to deposit by interception, impaction and/or gravitation compared to particles >1 mm and deposit primarily by Brownian diffusion (Seinfeld and Pandis, 1998). As observed by Zhu et al. (2011), the PM deposition can be approximated by C ¼ Co eekx. The average concentration for each size range in the control volume is determined by integrating Co eekx over the beginning and ending horizontal points of the CV (i.e., x1 and x2) and dividing by the distance (x1 x2) as follows: Cavg ¼ ekx1 ekx2 =kðx1 x2 Þ. Since the horizontal concentration change (dC/dx) is not constant, an exponential approximation is a better representation of a real world deposition scenario than the assumption Cavg ¼ (Csrc þ CL)/2 made previously by Zhu et al. (2011). The relative rate of vertical diffusion is invariant with particle size since the turbulent diffusion coefficient D is much greater than the Brownian diffusion coefficient, for dust particles between 0.3 mm and 20 mm (Seinfeld and Pandis, 1998). Eddy diffusivity can be taken as a linear function of height, D(z) ¼ 0.4 u*z in neutral conditions in the surface layer up to 30e100 m (Fisher, 1978) whereas the Brownian diffusion coefficient is small (of the order of 105 cm2 s1 or less). That is:
D dCavd D dCavd ¼ Cavg dz < 0:3mm Cavg dz > 1:0mm Solving Eq. (1) for vd for particles >1 mm yields:
(2)
546
D. Zhu et al. / Atmospheric Environment 122 (2015) 541e551
vd j > 1:0mm
uH Csrc CL ¼ L C avg
D dCavd Cavg dz > 1:0mm > 1:0mm
(3)
Similarly, solving Eq. (1) for vd for particles <0.3 mm and subtracting this from Eq. (1) yields:
! uH Csrc CL D dCavg vd j >1:0mm vd j <0:3mm ¼ L Cavg >1:0mm Cavg dz >1:0mm ! uH Csrc CL D dCavg L Cavg <0:3mm Cavg dz <0:3mm (4) since vd for particles >1 mm is much larger than particles <0.3 mm, Eq. (2) produces the following equation for vd as a function of particle diameter:
vd j > 1:0mm vd j < 0:3mm zvd ¼
uH L
Csrc CL C avg
> 1:0mm
> 1:0mm
! Csrc CL Cavg < 0:3mm
(5)
3. Results and discussion 3.1. PM10 attenuation downwind of the source u* and z0 were calculated from the vertical wind speed profile data from the near road tower calculated by:
U¼
u* z ln z0 k
(6)
rma n's constant where U is mean wind speed at height z, k is von Ka (0.4). The logarithmic wind profile is valid for the neutral atmosphere; however, for most applications close to the ground (i.e., z 10 m) the wind profile can be assumed almost always to be logarithmic even during non-neutral conditions (Kaimal and Finnigan, 1994). The stability function j(z/Lm) (Lm is MoninObukhov length) correction can be introduced into Eq. (6) for unstable conditions. In our study Lm values were 40 m for Ft. Riley, KS and 75 m for Hanford, WA, which represent unstable atmospheric conditions. The errors associated with the calculated z0 and u* values with the stability correction is within 10% of the log-linear approximation for neutral atmospheric conditions. Average z0, u* and atmospheric Pasquill stability classes for the testing periods and by location are listed in Table 1. The stability classes for the sagebrush and steppe grass at the Hanford site were derived from Richardson number estimates based on meteorological data from the nearby 120 m tower. The long grass vegetation had the highest z0 value at 0.206 m. Sagebrush had second highest value of 0.188 m, the mowed grass 0.025 m, the steppe grass site 0.027 m, and the bare soil 0.007 m. As not all winds came from the direction perpendicular to the road source, the angle of wind approach (AOA) was used to calculate the actual transport distances from source to the sampling positions. If the distance between two measurement positions is Lt, then the actual plume transport distance under an AOA (q) would be Lt/ cosq. Particles in all size classes generally show an exponential decay in particle number concentration with increasing downwind distance. A non-linear least squares regression fitted through the ensemble mean values for each size class from the multiple tests
(without forcing) was used to fit a function to the data and the coefficients of determination, R2 all >0.95 indicate the exponential regression is a strong fit. However, for the dense long grass site (Ft. Riley, KS), the decay is better fit by a power function due to the very rapid change of particle number concentration for all sizes measured in the initial z100 m downwind of the source. The largest fractional decrease in particle number concentration was observed at the 50e100 m position downwind of the source. The decay in number concentration is reflective of the removal of particles from the plume and proportional to plume concentration as previously observed by Johnson et al. (1992). Two causal, and complimentary, mechanisms are proposed that could explain the very rapid change of particle number concentration for the long grass surface. First, this surface has the highest aerodynamic roughness length, z0 (0.206 m). Numeric modeling and measurement of channel flow indicate roughness increases turbulence intensity (Bhaganagar et al., 2004; Kussin and Sommerfeld, 2002). Roth (2000) summarizes several mechanisms by which roughness elements influence turbulence. Roth (2000) notes an intense shear layer is formed near the top of the roughness elements, resulting in higher turbulence intensities; the mixing generated by turbulent wakes behind individual roughness elements, efficiently mix and diffuse momentum or any scalar quantity, the pressure difference across individual roughness elements augments the transport of momentum onto the surfaces. For particles in turbulent flow, turbulence will spread the particles and enhance the deposition rate (Li and Ahmadi, 1992). Pardyjak et al. (2013) report experimental evidence and modeling results that indicate that turbulence enhances particle deposition onto all surfaces of vegetation even for small particle sizes (e.g., PM10) in near source transport. A second mechanism that may be enhancing deposition in the long grass, more than the other vegetation types, is interception/impaction that affects deposition via contact of the particles in the dust plume with the grass blades. Belot and Gauthier (1975) report from wind tunnel experimental results, roughness elements with more surface area have better particle collection efficiency, e.g., blades of grass collect more particles than small leaf shrub types. This also can be invoked to explain why the long grass site shows enhanced particle losses compared to the sagebrush site even though their aerodynamic roughness lengths are similar (Table 1). With particle density assumed to be 2.65 g cm3 (with minerals like silica abundant in soil dust, Tegen et al., 2002), the particle number concentration was converted to a PM10 mass concentration by summing all particles in the size bins less than 10 mm. PM10 mass concentrations were found to decrease at faster rates as function of downwind distance with increased surface roughness for the five surface types tested in the first z100 m from source, again taking into account the actual fetch length depending on AOA q (Fig. 5). For the long grass site at Ft. Riley, KS, PM10 decayed to 13% of the near road concentration at 84 m downwind. For the short grass (mowed long grass), PM10 diminished to 48% of the near road concentration at 130 m downwind. For bare soil without vegetation at the Jean Lake, NV site, PM10 reduced to 67% of the near road concentration at 136 m downwind. Further downwind, the rate of decrease (dC/dx) reversed (decreased) with increased surface roughness, reflecting the proportionally larger decrease in the concentration along the downwind (x) direction. Due to increased turbulence caused by the higher roughness and increased impaction losses associated with the grass blades, the long grass site has the lowest TF for PM10 after 100 m transport. At distances >100 m the rate of decrease of particle number concentration for the long grass site was the lowest observed, while dC/ dx was highest for the non-vegetated lakebed surface.
D. Zhu et al. / Atmospheric Environment 122 (2015) 541e551
547
Table 1 Mean surface aerodynamic roughness length, friction velocity, and stability class for the five different surface sites during testing. The standard deviation of the mean value is provided in parentheses. Vegetation type
z0(m)
No vegetation (Jean Lake, NV) Cut grass (Ft Riley, KS) Steppe grass (Hanford, WA) Sagebrush (Hanford, WA) Dense long grass (Ft Riley, KS)
0.007 0.025 0.027 0.188 0.206
(±0.003) (±0.007) (±0.014) (±0.045) (±0.022)
u* (ms1)
Avg. U at 2.2 m (m s1)
Avg. U at 10 m (m s1)
Stability class
0.31 0.37 0.31 0.34 0.33
2.94 4.19 2.71 3.07 3.21
4.08 4.85 3.15 3.38 3.64
Neutral Mostly neutral Unstable Unstable Unstable to slightly unstable
(±0.03) (±0.03) (±0.06) (±0.04) (±0.04)
(±1.08) (±0.57) (±0.53) (±1.32) (±0.49)
(±1.04) (±0.51) (±0.75) (±1.50) (±0.61)
Fig. 5. Change in downwind PM10 concentration to near source PM10 concentration to for different test surfaces. The y axis is PM10 concentration normalized to the sampler nearest the source. Vertical error bars represent the standard deviation of the mean PM10 ratio for multiple plume passes. The horizontal error bars represent the standard deviation of the mean distance from source to receptor due to plume to plume changes in the AOA.
Fig. 6. Particle deposition velocity based on box model analysis for the five surface types. The vertical error bars are the standard deviation of the mean vd based on multiple plume passes. Values on the x-axis are particle geometric mean diameters of each size bin.
548
D. Zhu et al. / Atmospheric Environment 122 (2015) 541e551
Fig. 7. Ratio of vd over wind speed the for five surface types. The vertical error bars are the standard deviation of the mean vd/U based on multiple plume passes.
3.2. Size-specific PM deposition velocities Size-specific mean PM deposition velocities inferred from box model (Eq. (5)) calculations are presented in Fig. 6. For comparison, the terminal settling velocity vt was also calculated for the same sized particles using the Stokes Eq.:
vt ¼
2 1 Dp rp gCc 18 m
(7)
where Dp is the particle diameter, rp is the particle density, g is the gravitational constant (9.8 m s2), Cc is the Cunningham Slip Correction factor, and m is the dynamic viscosity of the atmosphere (1.7 105 kg m1 s1). Fig. 6 illustrates that the mean particle deposition velocity increases with particle size (geometric mean of the particle bin end sizes), as expected, and ranges from z1 cm s1 for particles <1 mm to z10 cm s1 for particles >10 mm. The deposition velocity is generally higher for long grass vegetation with greater roughness even though the average wind speed was higher for the short grass tests. This is due to higher surface
Fig. 8. Deposition velocities for particles measured during a single plume passage with similar wind speed for five surface types.
D. Zhu et al. / Atmospheric Environment 122 (2015) 541e551
roughness causing greater reduction in the number (or mass) concentration of PM, which then results in higher vd. A plot of mean vd/U versus particle diameter is shown in Fig. 7. It indicates that after the wind speed normalization, the ratio vd/U is highest for the aerodynamically roughest surface. This is due to increased turbulence induced by the larger and more-dense cover of roughness elements. Another factor enhancing the magnitude of vd on this surface type is demonstrated in Fig. 8, which compares vd values from a
549
single pass for the five vegetation types tested under similar mean wind speeds. vd calculated for particles in the turbulent surface layer, as expected, is higher than the Stokes setting velocities for same-sized particles According to Pardyjak et al. (2013), this can be explained with a combination of the classical Stokes number (Stk) and the Taylor-microscale Reynolds number (Rel), where particle impaction on to the vegetative elements is enhanced by increased turbulence intensity created by the roughness that transfers additional momentum to the particles resulting in an enhanced ability
Fig. 9. An example of suspended PM number size distribution (left column) and mass size distribution (right column) from samplers at 10 m, 100 m, and 200 m downwind of the roadway source.
550
D. Zhu et al. / Atmospheric Environment 122 (2015) 541e551
to deposit. In addition, plant morphology may also be playing a role as the surface area of the long grass blades and their increased density compared to the other vegetation types offers greater opportunities for the critical interactions to occur. 3.3. Changes in particle size with downwind distance Examples of the downwind particle number and mass size distribution for a single test run on the long grass surface are shown in Fig. 9. The coarse particles (between 2.5 and 10 mm) accounted for z20% of the PM10 number concentration at the three downwind measurement locations at Ft. Riley, but accounted for z95% of the PM10 mass concentration at the three downwind distances for the other sites. The coarse PM mass dominance was observed for all surface types, which indicates coarse particles dominate resuspended dust mass created by the vehicle travel on all the unpaved test roads. The concentration measurements made at the lower inlet height (2.2 m a.g.l.) at the z100 m and z200 m downwind positions, compared with number concentrations at the same size bin from the sampler with the low elevation inlet at the near road position (i.e., 10 m downwind of the source), show that larger particles decrease at a faster rate than smaller particles not only in horizontal direction, but also in the vertical direction (Table 2). For example, under the short grass condition at Ft. Riley, KS, the number concentration measured at 6.2 m a.g.l. at the 105 m and 205 m downwind sites normalized to the corresponding size-bin number concentration at the 2.2 m a.g.l. sampling height 20 m downwind measurement position, show a greater fractional reduction in 10 mm particles at 205 m (0.01) than at 105 m (0.042). This pattern is produced across all sites (Table 2). Similar trends of change in PM mass concentration in the vertical direction and with increasing plume travel distance are reported by other researchers (e.g., Chan €ll et al., 2003). and Kwok, 2000; Janha The mass median diameter (MMD) is a useful statisticallyderived parameter for characterizing a distribution of suspended particles to compare and contrast differences between the distributions resulting from changes due to aging of the plume mass as it is transported. For example, an MMD of 10 mm means that 50% of the total sample mass will be present in particles having diameters <10 mm, and 50% of the total sample mass will be present in particles having a diameter >10 mm. For the Ft. Riley short grass surface, the MMDs of the dust plume were 11 mm, 9.6 mm, and 9.0 mm measured at the low sampler inlets at 20 m, 105 m, and 205 m downwind, respectively, and 9 mm and 7.5 mm at the high sampler inlets at 105 m and 205 m downwind. At the bare, smooth Jean Lake site the MMDs were 16 mm, 16 mm, and 15.2 mm at the low sampler inlets at 10 m, 110 m, and 210 m downwind, respectively, and 13.2 mm and 13 mm from the high inlets at 110 m and 210 m
downwind. The difference of MMD at the first downwind sampler between the two sites (Ft. Riley vs. Jean Lake) is due to the different particle size distribution of the source material at each site. The trend of MMD decreasing horizontally and vertically as a function of downwind distance however is similar for the different surface conditions, which as expected indicates the larger particles were deposited and removed from the dust plume at a faster rate than small particles during downwind transport, which provides confidence that our data set has captured the process of interest. 4. Conclusions Near source PM reduction for different surface roughness conditions was characterized by the TF of PM10 and inferred vd using particle number concentration measurements and application of a horizontal gradient box model (for vd). The dust plumes were created by wheeled vehicles traveling at a constant speed on unpaved roads at test sites in Kansas and Washington, and a playa surface in Nevada. These three locations provided five different surfaces, one smooth and four with various types and covers of vegetation to estimate the impact of the roughness/vegetation on PM deposition close to the source (0e200 m). Coarse particles accounted for z95% of the PM10 for four of the sites (short grass, steppe grass, sagebrush, and barren) at three downwind locations between z100 and z200 m downwind, with the exception being the tall grass site where coarse particles comprised z20% of the PM10 over this same distance range, which was due to the rapid change between z10 m and z50 m. PM10 removal increased with increasing surface and aerodynamic roughness. The dense, high grass surface had the highest reduction of PM10, which is attributed to increased deposition due to increased turbulence induced by the rougher surface and by enhanced particle impaction/interception effects with the blades of grass. These data also suggest that the type of vegetation exerts some control on the depositional process as the long grass site had higher PM losses than the sagebrush site even though their aerodynamic roughness, and hence drag characteristics, were not that dissimilar. Estimated deposition velocities ranged from z1 cm s1 for particles <1 mm to z10 cm s1 for particles >10 mm. The estimated vd values were found to increase with particle size and surface roughness under similar moderate wind speed conditions. Compared to the Stokes settling velocity vd was increased by a factor of z2 (steppe grass) to z4 (long grass) for the largest measured particle size depending on the surface type. Large particle mass number concentrations were observed to decrease at a faster rate not only in horizontal direction, but also in the vertical dimension as the dust plume was transported downwind. The mass median diameter of the dust plume decreased along both the horizontal and vertical dimensions, at rates greater than can be
Table 2 Size-segregated survival rate of PM number concentration after transport normalized to 20 m downwind 2.2 m a.g.l. inlet readings for the short grass surface at Ft. Riley, KS. Survival rates along the horizontal direction (same 2.2 m a.g.l. inlet readings at 105 m and 205 m)
Kansas short grass site Nevada No vegetation site Washington Sagebrush site Washington Steppe grass site
PM size bins 0.5e0.7 mm Downwind distances 105 m 0.85 205 m 0.12 105 m 0.84 205 m 0.39 105 m 0.29 205 m 0.041 105 m 0.72 205 m 0.114
Note: No high inlets were set for the long grass site, Ft Riley, KS.
Survival rates along the vertical direction (6.2 m a.g.l. inlet readings at 105 m and 205 m)
2.5e5 mm
5e10 mm
>10 mm
0.5e0.7 mm
2.5e5 mm
5e10 mm
>10 mm
0.47 0.08 0.43 0.36 0.26 0.037 0.62 0.059
0.47 0.065 0.37 0.31 0.22 0.036 0.51 0.049
0.22 0.025 0.13 0.07 0.11 0.030 0.35 0.042
0.80 0.05 0.74 0.30 0.18 0.028 0.37 0.086
0.35 0.03 0.32 0.24 0.14 0.025 0.29 0.038
0.116 0.02 0.19 0.17 0.12 0.023 0.26 0.029
0.042 0.01 0.10 0.05 0.08 0.014 0.20 0.02
D. Zhu et al. / Atmospheric Environment 122 (2015) 541e551
attributed to simply gravitational sedimentation modifying the plume PSD. These results also suggest a reason for the poor reconciliation between emission inventory estimates of unpaved road dust and the amount of mineral dust observed in samples collected in air quality sampling networks. The inventory approach may be overestimating contributions because it is based on only near source emission fluxes that do not take into account the depositional losses in the near field that were clearly affected to a significant degree by the type of surface the dust plumes interacted with as they advected downwind. Acknowledgments This work was sponsored by the Strategic Environmental Research and Development Program (Project RC-1729). Many thanks to those who participated in the field work from the Pacific Northwest National Laboratory, Richland WA. We would also like to thank U.S. Army, Ft. Riley, KS for their logistical support during the field measurement campaign there. References Belot, Y., Gauthier, D., 1975. Transport of micronic particles from atmosphere to foliar surfaces. In: Heat Mass Transf. Biosphere, 2 (4). Scripta Book, Co., Washington DC, p. 31. Bhaganagar, K., Kim, J., Coleman, G., 2004. Effect of roughness on wall-bounded turbulence. Flow Turbul. Combust. 72 (2e4), 463e492. Chan, L.Y., Kwok, W.S., 2000. Vertical dispersion of suspended particulates in urban area of Hong Kong. Atmos. Environ. 34, 4403e4412. Cowherd, C., Muleski, G., Gebhart, D.L., 2006. Development of an emission reduction term for near-source dust depletion. In: Conference Proceedings, 15th International Emission Inventory Conference: “Reinventing Inventories e New Ideas in New Orleans”, New Orleans, LA, May, 2006. Dong, Y., Hardy, R., McGown, M., 2003. Why road dust concentrations are overestimated in Eulerian grid models. In: Appendix K, Northern Ada County (Idaho) PM10 Maintenance Plan. Idaho Department of Environmental Quality. June 2003. http://www.cmascenter.org/conference/2004/abstracts/poster/ dong_abstract.DOC. Etyemezian, V., Gillies, J., Kuhns, H., Gillette, D., Ahonen, S., Nikolic, D., Veranth, J., 2004. Deposition and removal of fugitive dust in the arid southwestern United States: measurements and model results. J. Air & Waste Manage. Assoc. 54, 1099e1111. Fisher, B.E.A., 1978. The calculation of long term sulphur deposition in Europe. Atmos. Environ. 12, 489e501. Fowler, D., Pilegaard, K., Sutton, M.A., Ambus, P., Raivonen, M., Duyzer, J., Simpson, D., et al., 2009. Atmospheric composition change: ecosystemseatmosphere interactions. Atmos. Environ. 43, 5193e5267. Friedlander, S.K., 2000. Smoke, Dust, and Haze, vol. 198. Oxford University Press, New York. Gillies, J.A., Etyemezian, V., Kuhns, H., Nikolic, D., Gillette, D.A., 2005. Effect of vehicle characteristics on unpaved road dust emissions. Atmos. Environ. 39, 2341e2347. Gillies, J.A., Etyemezian, V., Kuhns, H., McAlpine, J.D., King, J., Uppapalli, S., Nikolich, G., Engelbrecht, J., 2010. Dust emissions created by low-level rotarywinged aircraft flight over desert surfaces. Atmos. Environ. 44, 1043e1053.
551
Hage, K.D., 1961. On the dispersion of large particles from a 15-m source in the atmosphere. J. Meteor. 18, 534e539. €ll, S., Molna r, P., Hallquist, M., 2003. Vertical distribution of air pollutants at Janha €teborg, Sweden. Atmos. Environ. 37 (2), 209e217. the Gustavii Cathedral in Go Johnson, T.C., Gillette, D.A., Schwiesow, R.L., 1992. Fate of dust particles from unpaved road under various atmospheric conditions. In: Precipitation Scavenging and Atmosphere-surface Exchange, pp. 933e948. Kaimal, J.C., Finnigan, J.J., 1994. Atmospheric Boundary Layer Flows: Their Structure and Measurement. Oxford University Press, New York, NY. Kussin, J., Sommerfeld, M., 2002. Experimental studies on particle behaviour and turbulence modification in horizontal channel flow with different wall roughness. Exp. Fluids 33 (1), 143e159. Li, A., Ahmadi, G., 1992. Dispersion and deposition of spherical particles from point sources in a turbulent channel flow. Aerosol Sci. Technol. 16 (4), 209e226. Mao, Y., Wilson, J.D., Kort, J., 2013. Effects of a shelterbelt on road dust dispersion. Atmos. Environ. 79, 590e598. Nemitz, E., Gallagher, M.W., Duyzer, J.H., Fowler, D., 2002. Micrometeorological measurements of particle deposition velocities to moorland vegetation. Quart. J. Roy. Meteor. Soc. 128 (585), 2281e2300. Pace, T.G., 2005. Methodology to Estimate the Transportable Fraction (TF) of Fugitive Dust Emissions for Regional and Urban Scale Air Quality Analyses. U.S. EPA, Research Triangle Park. August 2005. http://www.epa.gov/ttnchie1/emch/ dustfractions/transportable_fraction_080305_rev.pdf. Pardyjak, E.R., Speckart, S.O., Yin, F., Veranth, J.M., 2008. Near source deposition of vehicle generated fugitive dust on vegetation and buildings: model development and theory. Atmos. Environ. 42, 6442e6452. Pardyjak, E., Veranth, J., Speckart, S., Moran, S., Price, T., 2013. Development of a Windbreak Dust Predictive Model and Mitigation Planning Tool. Final Report. SERDP Project RC-1730. University of Utah, Salt Lake City, UT. Pinnick, R.G., Fernandez, G., Hinds, B.D., Bruce, C.W., Schaefer, R.W., Pendleton, J.D., 1985. Dust generated by vehicular traffic on unpaved roadways: sizes and infrared extinction characteristics. Aerosol Sci. Technol. 4, 99e121. Poggi, D., Katul, G.G., Albertson, J.D., 2006. Scalar dispersion within a model canopy: measurements and three-dimensional Lagrangian models. Adv. Water Res. 29, 326e335. Pouliot, G., Simon, H., Bhave, P., Tong, D., Mobley, D., Pace, T., Pierce, T., 2012. Assessing the anthropogenic fugitive dust emission inventory and temporal allocation using an updated speciation of particulate matter. Air Pollut. Model. Appl. XXI 4, 585e589. January 01, 2012. Raupach, M.R., Woods, N., Dorr, G., Leys, J.F., Cleugh, H.A., 2001. The entrapment of particles by windbreaks. Atmos. Environ. 35, 3373e3383. Roth, M., 2000. Review of atmospheric turbulence over cities. Quart. J. Roy. Meteor. Soc. 126 (564), 941e990. Seinfeld, J.H., Pandis, S.N., 1998. Atmospheric Chemistry and Physics: from Air Pollution to Climate Change. Wiley, Hoboken, NJ, p. 1203. Tegen, I., Harrison, S.P., Kohfeld, K., Prentice, I.C., Coe, M., Heimann, M., 2002. Impact of vegetation and preferential source areas on global dust aerosol: results from a model study. J. Geophys. Res. 107 (D21), 4576. US EPA, 1995. Compilation of Air Pollutant Emission Factors Volume I: Stationary, Point and Area Sources. US EPA Office of Air Quality Planning and Standards, Research Triangle Park, NC. Veranth, J.M., Pardyjak, E., Seshadri, G., 2003. Vehicle-generated fugitive dust transport: analytic models and field study. Atmos. Environ. 37, 2295e2303. Walker, E.R., 1965. A particulate diffusion experiment. J. Appl. Meterol. 4, 614e621. Williams, D.S., Shukla, M.K., Ross, J., 2008. Particulate matter emission by a vehicle running on unpaved road. Atmos. Environ. 42 (16), 3899e3905. Wilson, J.D., 2000. Trajectory models for heavy particles in atmospheric turbulence: comparison with observations. J. Appl. Meterol. 39 (11), 1894e1912. Zhu, D., Kuhns, H., Gillies, J.A., Etyemezian, V., Gertler, A., Brown, S., 2011. Inferring deposition velocities from changes in aerosol size distributions downwind of a roadway. Atmos. Environ. 45, 957e966.