JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 115, D17301, doi:10.1029/2009JD013207, 2010
Implementation of road and soil dust emission parameterizations
in the aerosol model CAMx: Applications over the greater Athens
urban area affected by natural sources
E. Athanasopoulou,1 M. Tombrou,1 A. G. Russell,2 A. Karanasiou,3 K. Eleftheriadis,3
and A. Dandou1
Received 16 September 2009; revised 12 February 2010; accepted 27 April 2010; published 1 September 2010.
[1] A detailed dust emission parameterization was developed for aerosol production by
human and natural activity. The road dust scheme includes tire wear, break wear, road
abrasion and vehicle‐induced re‐suspension. The natural dust scheme includes wind
(Aeolian) erosion from soil surfaces and land disturbances, and considers the effects of soil
and atmospheric parameters on dust productivity. Emission rates are chemically and size‐
resolved and are incorporated in the CAMx aerosol model coupled with the ISORROPIA
II inorganic module. Emissions and concentrations are predicted for five simulation
periods, using a domain covering Greece with a fine mesh over the greater Athens area.
Re‐suspended mass is the main dust component, calculated 4–5 times higher than exhaust
emissions. Soil dust emissions are much greater than road dust during high winds but are
of less importance inside the city with maximums located at the periphery of the urban
core. Comparison with observations suggests that the road dust component seems
adequately estimated in traffic‐affected areas and accounts for 15–40% of the total PM10.
Calcium, regarded as a soil dust tracer, is calculated to be 2–4 mg m−3 in areas with high
Aeolian emission rates, similar to measured values. Sodium and magnesium predictions
show their marine origin, and reasonably replicate the observed mass and size distribution
during well‐established onshore flows. Nitrates are predicted as measured during lower
winds, but are underestimated when stronger winds prevail from the west. This is caused
by an overestimated industrial influence of Athens. As a result, ammonia is bound to the
excess of sulfate, rather than reacting with nitric acid. Secondary species are influenced
slightly by heterogeneous chemistry on dust particles.
Citation: Athanasopoulou, E., M. Tombrou, A. G. Russell, A. Karanasiou, K. Eleftheriadis, and A. Dandou (2010),
Implementation of road and soil dust emission parameterizations in the aerosol model CAMx: Applications over the greater
Athens urban area affected by natural sources, J. Geophys. Res., 115, D17301, doi:10.1029/2009JD013207.
1. Introduction
[2] Dust is primary aerosol arising from mechanical processes and may be formed either anthropogenically or naturally. Anthropogenic sources include vehicle movement
(on‐ and off‐road, including construction), mining and
farming. Vehicle movement initiates both aerosol production
and re‐suspension mechanisms. Road dust production is a
result of the generation of shear forces by the relative
movement of the vehicles’ tires and the road surface, and also
by the brakes, during the deceleration process [European
Environment Agency (EEA), 2007]. A fraction of the total
1
Department of Applied Physics, National and Kapodistrian University
of Athens, Athens, Greece.
2
School of Civil and Environmental Engineering, Georgia Institute of
Technology, Atlanta, Georgia, USA.
3
Institute of Nuclear Technology and Radiation Protection, National
Center of Scientific Research “Demokritos,” Athens, Greece.
Copyright 2010 by the American Geophysical Union.
0148‐0227/10/2009JD013207
atmospheric aerosol mass produced by all processes is
deposited onto land, particularly road surfaces, where it can
be re‐emitted. Natural dust occurs when wind acts on
land surfaces, either producing or re‐suspending particles.
Aeolian erosion is the process of wind‐forced movement of
soil particles, which gives birth to soil “bursts” that mobilize
mineral aerosol [Shao, 2001; Kondratyev et al., 2005].
Although particle flux is naturally driven, it can be enhanced
by anthropogenic disturbances in the natural land surface
(e.g., earthwork sites, land use changes).
[3] Anthropogenic dust leads to large quantities of carbon
and several metals (copper, zinc, iron, magnesium, calcium
etc.) in ambient aerosol and contributes not only to the
coarse but also to the fine sizes of aerosol [EEA, 2007;
Thorpe and Harrison, 2008]. Natural dust is mainly coarse
and dominated by mineral species (silicon, calcium, potassium, magnesium etc.) which may be reactive sites for
heterogeneous reactions of sulphur (SO2, H2SO4) and
nitrogen‐containing species (e.g., HNO3), leading to secondary aerosol formation (CaSO4, Ca(NO3)2 etc.) in the
D17301
1 of 21
D17301
ATHANASOPOULOU ET AL.: DUST INCORPORATION IN PM SIMULATIONS
atmosphere [Seinfeld and Pandis, 2006; Fountoukis and
Nenes, 2007]. Matsuki et al. [2004] found that calcite
(CaCO3) absorbs the largest amount of sulphur. Krueger et
al. [2004] also found high yields of nitrate formation on
CaCO3 and dolomite (MgCa(CO3)2). Mineral dust particles
can account from 55 to 70% for the formation of coarse
mode sulfate and for more than 90% of coarse mode nitrate
in non‐marine influenced sites [Hien et al., 2005].
[4] There are compelling reasons to include dust in
aerosol simulations, such as their aforementioned reactivity,
plus radiation, photochemistry and health effects [Myhre et
al., 2003; Tanré et al., 2003; Hien et al., 2005; Hodzic et
al., 2006; Fountoukis and Nenes, 2007; L. Perez et al.,
2008]. There is a potential health burden of dust particles,
mostly due to heavy metals in road dust [Schwarze et al.,
2007; Jalava et al., 2008; Ostro et al., 2009]. The paucity
of dust emission data, e.g., in inventories, leads to a lack of
knowledge about its air quality burden, effective controls
and impact on modeling of aerosol chemistry.
[5] Several authors have proposed parameterizations for
dust particle fluxes. A widely used vehicle‐induced dust
emission factor depends on vehicle weight and road surface
silt loading [U.S. Environmental Protection Agency, 1995,
2003]. A simpler dust emission factor calculation depends
on vehicle weight and speed [Gillies et al., 2005], which is
combined with the number of vehicles and kilometers
traveled [Allwine et al., 2006]. A detailed, analytical
parameterization for tire wear, brake wear and road abrasion
dust calculates the mass fluxes of each aerosol size fraction
and chemical species as a function of vehicle class, weight,
speed and load, number of axles (in the case of trucks),
number of vehicles and kilometers traveled [EEA, 2007].
Aerosol re‐suspension by vehicle movement can be indirectly estimated from measurements [Lohmeyer et al., 2002;
Gidhagen et al., 2004; Omstedt et al., 2005; Ketzel et al.,
2007; Thorpe et al., 2007; Patra et al., 2008]. This procedure is controlled by all the aforementioned factors plus the
surface and atmospheric humidity, as well as the number of
hours since last rainfall.
[6] Aeolian erosion dust flux is primarily a function of
friction velocity [White, 1979; Westphal et al., 1987; Shao et
al., 1993; Gillette and Passi, 1988; Alfaro et al., 1998;
Ginoux et al., 2001; Chin et al., 2002, 2004]. The process is
controlled by atmospheric conditions (precipitation, temperature and pressure), soil characteristics (texture, type and
water content) and land‐surface properties (aerodynamic
roughness length and vegetation cover). Instead of using
approximated constant values, physically based parameterizations of these parameters can be included in the
wind‐driven dust flux calculations [Bagnold, 1941; Tegen
and Fung, 1994; Marticorena and Bergametti, 1995;
Fécan et al., 1999; Liu and Westphal, 2001; Park and In,
2003; Alfaro et al., 2004; Shao et al., 1993]. The additional dust flux from disturbed land‐use practice is currently
partly parameterized either using information on heavy‐
vehicle movement around quarries [Moosmüller et al., 1998;
Abu‐Allaban et al., 2006] or by using average construction
expenses [Watson and Chow, 2000].
[7] A number of attempts has been made to incorporate
dust source functions in atmospheric models. Some studies
focus on simulations of the production of natural dust
[Alfaro and Gomes, 2001; Alfaro et al., 2004; Mansell et al.,
D17301
2004], particularly, desert dust mobilization and its impacts
on a global [Zender et al., 2003; Ginoux et al., 2004; Bauer
et al., 2004; Chin et al., 2007; Cheng et al., 2007; Lee et al.,
2009] or on regional scale [Nickovic et al., 2001; Alpert et
al., 2004; Vautard et al., 2005; Kallos et al., 2006; N.
Pérez et al., 2008; Shaw et al., 2008]. High resolution
regional simulations have found improvements in total mass
levels, but without investigating the impact of local dust on
atmospheric chemistry [Choi and Fernando, 2008]. The
chemical atmospheric impact of natural dust is investigated
by a few regional, though coarse resolution, applications
that focus on nitrate species [Hodzic et al., 2006; Astitha
and Kallos, 2008]. Both studies introduce an uptake coefficient of nitric acid on dust particles, which leads to a
reduction of model biases.
[8] The greater Athens area is a challenging region for
dust modeling. While the area is dominated by human
activities, it is occasionally subjected to Saharan dust outbreaks and suffers from low precipitation rates, leading to
increased dust accumulation and wind erosion [Kallos et al.,
2006, 2007; Rodríguez et al., 2004]. Indeed, Saharan, road
and soil dust are found as important aerosol contributors
[Scheff and Valiozis, 1990; Eleftheriadis et al., 1998; Siskos
et al., 2001; Dall’Osto and Harrison, 2006; Mitsakou et al.,
2008; Karanasiou et al., 2009]. Recent studies indicate the
dust contribution to PM10 measurements in Athens is 7–
20% Saharan [Mitsakou et al., 2008], 25% road and 13%
soil dust [Karanasiou et al., 2009], normalized over the
year. Previous aerosol simulations indicated that dust
incorporation could improve evaluation results of aerosol
over the area [Athanasopoulou et al., 2005, 2008].
[9] The aim of this study is to develop an approach to
quantify dust emissions by source and assess atmospheric
pollution, including its impact on urban chemical environments affected by both natural and anthropogenic pollution
sources. To this end, parameterizations of vehicles’ tire
wear, brake wear and road abrasion, vehicle‐induced re‐
suspension and Aeolian erosion of soil, earthwork sites and
quarries are developed and described. These algorithms are
implemented in a research version of the CAMx aerosol
model [ENVIRON, 2003] that incorporates analytical
equations for the equilibrium of the K+–Ca2+–Mg2+–NH+4 –
−
−
Na+–SO2−
4 –NO3 –Cl –H2O aerosol system. The impact of
dust aerosol on atmospheric chemical constituents over
Athens is examined by means of simulations. The results of
the modeling study are compared to simultaneous analytical
aerosol measurements.
2. Measurements
2.1. Mass Size Distributions and Selected Water
Soluble Ions
[10] Data from size distribution measurements of aerosol
mass and concentrations determined for water soluble
inorganic species [Eleftheriadis et al., 2006a] were utilized
in this study. Samples were obtained by means of an 11
stage Berner cascade impactor at the “Demokritos” urban
background site (DEM). Measurements took place during
May 25 and 26, 2005 and April 10–13, 2006 and data refer
to average 24‐h values from 10:00a.m. until 09:00a.m. next
day (LST). Aerosol was classified in 10 fractions with 50%
cut off sizes given at 0.026, 0.062, 0.110, 0.173, 0.262,
2 of 21
D17301
ATHANASOPOULOU ET AL.: DUST INCORPORATION IN PM SIMULATIONS
D17301
Figure 1. Application domain and monitoring sites: (a) the parent domain covers Greece, and (b) the
nested domain (shown as an embedded rectangle in Figure 1a) covers the Attica peninsula. Outside ticks
at 10 km intervals. Topography contours at 300 m intervals.
0.46, 0.89, 1.77, 3.4, and 6.8 mm. Samples were collected on
tedlar foils and were equilibrated in a relative humidity (RH)
and temperature (T) controlled weighing room for at least 24
h before weighing. Weighing of the foils before and after
sampling was conducted on a Sartorius BP211D balance
(10−5 gr resolution).
2.2. PM10 and PM2.5 Mass Concentration
[11] Additional data include hourly and daily concentrations of PM2.5 and PM10 measured by b‐gauge attenuation
monitors operating for the National Air Pollution Monitoring Network. Aerosol PM concentrations are provided for
the Aristotelous, Marousi, Piraeus and Goudi urban traffic
sites (ARI, MAR, PIR, GOU), the Agia Paraskeui, Zografou
and Thrakomakedones urban background sites (AGP, ZOG,
THR), the Lykovrisi suburban site (LYK), the Colegio and
Iraklio highway sites (COL, IRA) and the Heraklion, Patra
and Volos Greek periphery sites (HER, PAT, VOL) (map as
in Figure 1). It has to be noted that b‐gauge instruments
often introduce systematic bias due to liquid water, gas to
particle disequilibrium during sampling and trace gas
adsorption on filter media. The bias is RH and T dependent,
or in other words, is site and season specific [Chang and
Tsai, 2003; Takahashi et al., 2008]. As a result, an uncertainty in the range of 30% is assumed reasonable for the
measurements reported here. On the other hand, size distribution measurements at DEM, yield gravimetric PM10
concentrations, which in general suffer from minor artifact
formation and their uncertainly is estimated in the range of
15%.
2.3. Atmospheric Conditions and Dates for
Comparison With Model Results
[12] The comparison for PM10 at the sites described above
can show whether aerosol modeling can reproduce the
general aerosol regime over Attica (section 5.2). Hourly data
for 2005 are processed for developing the approach to calculate aerosol re‐suspension (section 4.2). Hourly data for
2006 are used for evaluation of the simulations (section 5.2).
Size distributions (for 2006) give detailed information for
the particulate mass (section 4). The ratio PM2.5/PM10
(DEM, AGP, LYK, PIR) can be an indicator of anthropogenic to total aerosol loading (section 5.2).
[13] The dates selected for numerical simulations include
May 25 and 26, 2005, April 10–13, 2006 (covering the
above measurements), April 25 and 26 and May 17 and 18,
2007. Meteorological information for the simulated days is
taken from the data set of the National Observatory of
Athens and from “Demokritos” (Table 3).
[14] Detailed study of the size distribution and chemical
composition of aerosol is confined to the days where such
information is available. Caution should be placed when
examining the days of 2005, as rain poses another modeling
uncertainty and minimizes dust production and effects. The
Saharan dust intrusion during the 12th of April, 2006
(section 3.2) is not captured in the boundary conditions, thus
modeling underestimations are expected for that period.
3. Model Details
3.1. Air Quality Model
[15] PMCAMx is the research version of the publicly
available 3‐D, Eulerian chemical transport model CAMx
[ENVIRON, 2003]. The CMU scheme designed to estimate size resolved and speciated aerosol concentrations
[ENVIRON, 2004], involves ten discrete and internally mixed
size sections (cut‐off diameters: 0.04, 0.08, 0.1, 0.3, 0.6, 1.2,
2.5, 5, 10, 20, 40 mm), each consisting of sixteen chemical
aerosol components (primary: organic and elemental carbon,
chloride, sodium, sulfate, calcium, magnesium, potassium
3 of 21
D17301
ATHANASOPOULOU ET AL.: DUST INCORPORATION IN PM SIMULATIONS
ions and the rest crustal material; secondary: four organics,
water, ammonium, nitrate and sulfate ions). The aerosol‐
related dynamical processes considered in PMCAMx include
primary emissions, new particle formation by nucleation,
condensation, evaporation, wet and dry deposition and
chemistry. Coagulation has a low impact for aerosol mass and
is currently ignored.
[16] The model used in this study (version 4.02) uses the
Carbon Bond gas‐phase chemical mechanism (CB) IV
[Gery et al., 1989], the RADM algorithm for cloud chemistry [Chang et al., 1987], the SOAP scheme for the organic
species partitioning [Strader et al., 1998] and ISORROPIA
II for the thermodynamic equilibrium of inorganic aerosol
[Fountoukis and Nenes, 2007]. The system treated by
ISORROPIA II includes sodium (Na+), ammonium (NH+4 ),
−
2+
chloride (Cl−), sulfate (SO2−
4 ), nitrate (NO3 ), calcium (Ca ),
2+
+
magnesium (Mg ), potassium (K ) and water (H2O), which
are partitioned between gas, and condensed (liquid and
solid) phases. Sulfate, sodium and crustal species completely reside in the aerosol phase, and their relative abundance determines the potential aerosol regime. For the rest
aerosol species, mass transfer between gas and condensed
phases is calculated using the hybrid approach [Capaldo
et al., 2000], which assumes instantaneous equilibrium
between gases and fine aerosols (equilibrium approach) and
solves the mass transfer rate equations for the coarse mass
(dynamic approach). The threshold diameter between the
two approaches is set at 2.5 mm instead of 1 mm [Capaldo
et al., 2000; Koo et al., 2003; Gaydos et al., 2003]. The
computational burden imposed by the dynamic treatment of
crustal species imposes the restricted use of the dynamic
approach. A sensitivity test by applying the lower cut‐off
(1 mm) resulted in small changes in average predictions
[Athanasopoulou et al., 2008].
[17] Meteorological fields used in PMCAMx are provided
by the fifth generation Penn State/NCAR Mesoscale Model,
MM5 [Grell et al., 1994], processed using “mm5camx‐v4.2”
(http://www.camx.com/down/support.php). It should be
noticed that the diffusion coefficients are not calculated by
“mm5camx‐v4.2” off‐line (by considering the calculated
temperature, wind speed, friction velocity etc) but they are
provided directly by the MM5 model. Furthermore,
“mm5camx‐v4.2” was extended to treat frictional velocity,
monthly vegetation fraction, fractional land use / vegetation
type, gravimetric soil moisture and soil texture type, so as to
provide dust emission fields. Land use fractions for the
Athens urban core are replaced by information from a recent
high resolution (30 × 10−3 km) satellite image [Dandou et al.,
2005].
3.2. Modeling Domain, Initial and Boundary
Conditions
[18] The simulation area covers the greater Greece area
(parent domain) using a spatial resolution of 6 km, while a
finer resolution of 2 km is used in the nested domain centered on the Attica peninsula (Figure 1). The areas of the
parent and the nested domain are 777924 km2 (147 × 147
cells) and 21600 km2 (75 × 72 cells), respectively. Vertically, there are 14 layers extending to 4.5 km, with the first
layer being approximately 30 × 10−3 km thick. The specific
model top is sufficient for studying of local dust emissions.
Inclusion of long range transport of dust would demand a
D17301
higher model top (i.e., 7–8 km), in order to capture the
Saharan typical boundary layer heights.
[19] The MM5 numerical simulations also use the above
mentioned domains, which are nested into the extended
European area (7193124 km2) using a spatial resolution of
18 km. All domains have their top layer at about 13 km agl.
The meteorological conditions assumed to initialize simulations are provided by the European Centre for Medium
range Weather Forecast (ECMWF) analysis fields (0.5 deg ×
0.5 deg) and sea‐surface temperature (SST) data (1 deg ×
1 deg).
[20] Initial and boundary conditions for pollutant concentrations during the days examined are assumed to be
spatially and temporally invariant for the parent domain.
The concentrations for the main gaseous pollutants are
typical values for the area during warm periods, i.e., 50 ppb
CO, 1 ppb NOx, 1 ppb SO2 and 40 ppb O3. For the rest of
the gaseous species, lower bound concentrations are automatically set by the model (10−6 ppb). Simulations discussed above cover the five 2‐day episodes with 3‐day spin‐
up periods to dampen the effect of initial conditions.
[21] Almost all simulated periods are not associated with
African dust transport in the Mediterranean atmosphere, and
this was confirmed by applications of the trajectory model
HYSPLIT (http://www.ready.noaa.gov/ready/open/hysplit4.
html) and by OMI Aerosol Index (AI) satellite imagery
(http://toms.gsfc.nasa.gov/eptoms/ep_v8.html). Thus, the
default aerosol boundary concentrations are used for all size
bins in this study (10−9 mg m−3). This is important when
evaluating coarse aerosol predictions for the period of April
12 and 13, 2006, which is affected by Saharan dust.
4. Emissions
[22] Conventional anthropogenic emission data (both gas
and condensable phases), as well as the sea‐salt aerosol
emission parameterization are described in a preceding
study [Athanasopoulou et al., 2008]. Advances in the existing emission database developed here, involve recent
changes in the Athens road network, improvements in the
ammonia emission rates, changes in the chemical profiles
and size distribution of aerosol emissions and the incorporation of dust emission rates described in sections 4.1–4.3.
[23] Hourly emission rates by vehicle exhaust for the
greater Attica area, provided initially by the Ministry of
Environment for 2002, have been refined to account for the
new (2004) highway inside the Athens basin. In particular,
the existing exhaust emissions have been redistributed
onto the new road network, including emissions associated
with toll stations (38 sites), long tunnels (1) and junctions
(25 sites) of the new highway road (Figure 7a).
[24] Large discrepancies between ammonia (NH3) predictions and observations were found for a summer case
[Athanasopoulou et al., 2008], explained by the fact that
NH3 emissions from agricultural activities were provided
only for the winter period. Sensitivity tests showed that the
winter NH3 emission rates should be multiplied by a factor
of 5 in order to match the NH3 observations during the
warm period.
[25] The chemical speciation and size distribution of
aerosol emissions by vehicle exhaust [Athanasopoulou
4 of 21
D17301
D17301
ATHANASOPOULOU ET AL.: DUST INCORPORATION IN PM SIMULATIONS
Table 1. Chemical Source Profiles of Aerosol Emissionsa
PM40
Organic carbon
Elemental carbon
Sulfates
Nitrates
Chloride
Sodium
Calcium
Potassium
Magnesium
Other mass
Re‐suspension
Road Abrasion
Vehicle Exhaust
Tire Wear
Brake Wear
Soil Dust
PM2
PM2–10
PM2
PM2–10
60
26.5
6.3
0
0
0
0
0
0
7.2
67.5
28.7
0.5
0.3
0.1
0.1
0.2
0.1
0
2.5
16.2
4
5.1
0.2
0.2
1.2
1.2
0.1
6.8
65
0
0
10
0
0
2.2
43.9
4.5
3.1
36.3
60
28.7
5.3
0
0
0.03
3.4
0
0.3
2.3
60
9
13.6
0
0
1.7
0.3
0.7
0.4
14.3
60
18
10
0
0
0.06
6.7
0
0.6
4
60
0
17
0
0
3.3
6.4
1
0.7
11
a
Mass fractions (%) for vehicle exhaust, soil dust, re‐suspension and road abrasion are based on Karanasiou et al. [2009]. The rest fractions are taken
from EEA [2007].
et al., 2008], is modified using recently available information. The inorganic chemical source profile for vehicle
exhaust is taken from Karanasiou et al. [2009] (Table 1).
The size distribution of the three mass fractions (PM1, PM2.5
and PM2.5–10) to ten size bins (Figure 2) is reached by using
measurements performed in this study. This led to emissions
with a primary accumulation mode peak around 0.5 mm.
The dust emission sectors described in the following paragraphs are also subjected to this size information.
4.1. Tire Wear, Brake Wear, and Road Abrasion
[26] In order to calculate aerosol emissions from tire wear,
brake wear and road abrasion processes, the detailed method
from the EMEP/CORINAIR emission inventory is applied
[EEA, 2007] to the Attica road network, using hourly,
gridded, exhaust emission rates. Data for the number of
vehicles in each defined class (two‐wheelers, passenger
cars, light duty trucks and heavy vehicles) within the Attica
network and the mileage driven by each vehicle in each
Figure 2. Size source profiles of aerosol emissions. Dots indicate the mean diameter of the each size
section treated by PMCAMx. The mass fractions (%) for vehicle exhaust and re‐suspension are based
on Karanasiou et al. [2009]. Wear processes and road abrasion fractions are taken from EEA [2007].
Soil dust sizes depend on the local soil classes provided by the MM5 database.
5 of 21
D17301
ATHANASOPOULOU ET AL.: DUST INCORPORATION IN PM SIMULATIONS
D17301
Figure 3. PM10 versus NOx concentrations at the (a) Athens city center (ARI) and (b) traffic urban area
(MAR) during 2005. The gray points are treated as data outliers: less than 15% and greater than 85% in
Figure 3a and less than 25% and greater than 75% of the data in Figure 3b. Least square equation and
correlation coefficient values (r2) of the central data (black points) are also displayed.
class are taken from the recent inventory provided by
Hellenic Ministry of Environment [2006].
[27] Dust production (per km basis) by the wear of vehicle
tires depends on vehicle class and speed, while in the case of
heavy vehicles, the load and the number of axles, also affect
dust emissions. Brake wear processes are affected by all the
factors above, except the number of axles of heavy‐duty
vehicles. Road abrasion is only affected by the class of the
vehicle.
[28] Typical size profiles (5 and 3 size classes for wear and
road abrasion particles, respectively) and chemical speciation of dust production processes (Table 1 and Figure 2) are
based on EEA [2007], except for the chemical profiles of
road abrasion emissions that are based on the road dust
source profiles for fine and coarse fractions, provided by
Karanasiou et al. [2009].
4.2. Vehicle‐Induced Re‐suspension
[29] Estimation of aerosol re‐suspension based on emission factors is highly uncertain as it depends on deposited
material from a wide range of sources and on precipitation.
An alternative approach is to use the mass closure method
[Ketzel et al., 2007; Patra et al., 2008; Thorpe et al., 2007].
Aerosol re‐suspension by vehicle movement is calculated as
the residual between the total aerosol mass from vehicle
operation and the exhaust, tire wear, brake wear and road
abrasion emissions.
[30] For the current study, aerosol re‐suspension rates by
vehicle movement are estimated using a combination of
aerosol vehicle emissions and measured aerosol ratios to
NOx [Omstedt et al., 2005]. A number of studies characterizing PM10 measurements in urban areas consider NOx
concentrations as a satisfactory indicator of vehicle exhaust
emissions when sampling is co‐located [Scheff and Valiozis,
1990; Grivas et al., 2004]. Therefore, regression analysis
between PM10 and NOx concentrations has been applied as a
method for the quantification of the total PM10 mass related
to vehicle movement [Fuller et al., 2002; Harrison et al.,
2003; Thorpe et al., 2007]. In the present study, NOx is
used as a tracer for vehicle emissions using two sampling
sites of the National Air Pollution Monitoring Network. One
is located along a busy inner‐city roadway (Aristotelous
Str.: ARI), and the other is in the suburban area Marousi
(MAR), located 300m from the heavily used Kifissias avenue. Neither site is near major industrial sources [Grivas
et al., 2004].
[31] Data analysis involves measurements of PM10 and
NOx during 2005 (Figure 3). This period was chosen not
only due to the available road transport data but also because
it is the only year after the new highway road construction
for which hourly PM10 measurements are available. Data
was treated to exclude background pollution and measurement errors. Windy hours (U > 2 m s−1) are omitted so as to
minimize pollutant transport, Saharan dust events (OMI
data) and high moisture data (RH > 60%) are removed in
order to avoid both particle washout and moisture problems
on particle counters. Then, the hours with high traffic loads
(07:00–10:00 and 20:00–23:00 LST) are selected, in order
to capture periods dominated by high vehicle source
strengths.
[32] The NOx−PM10 relationship is still somewhat scattered. Thorpe et al. [2007] suggested that points where the
PM10 to NOx ratios are larger than 0.1 should be excluded as
outliers. In the present study, the interquartile range (IQR),
being equal to the difference between the third and first
quartiles, is applied to identify outliers. The IQR is regarded
as a robust statistic and is often preferred to the total data set.
In this study, outliers with unreasonably high or low PM10
to NOx ratios may be related to events with additional
aerosol sources (other than vehicles) or problematic measurements, respectively.
[33] The IQR data for MAR have a satisfactory correlation
(r2 = 0.77), but a higher correlation (r2 = 0.87) is obtained
for the ARI data set, even for a wider data range (70%). The
representative ratios (PM10/NOx) found by this methodology are 0.26 and 0.51, for the ARI and MAR locations,
respectively, similar to these calculated by Grivas et al.
[2004]. The value of 0.26, being the most representative
and conservative is taken as the indicator of the total aerosol
related to road transport. Subsequently, the vehicle‐induced
re‐suspension emission rate (PM10res, g hr−1 m−2) is estimated as,
PM10 res ¼ 0:26 NOx exh PM10 exh PM10 tbr
ð1Þ
where, NOxexh and PM10exh are the emission rates of
nitrogen oxides and aerosol, respectively, by vehicle exhaust
6 of 21
D17301
ATHANASOPOULOU ET AL.: DUST INCORPORATION IN PM SIMULATIONS
Table 2. Fractions b(i, j) of the Soil Texture Classes in the Four
Soil Categories in Attica, as Indicated by the MM5 Databasea
Soil Texture Class (i)
Soil Category (j)
1: Loam
2: Sandy clay loam
3. Clay loam
4: Clay
g (i)
1: Clay (PM2) 2: Silt (PM2–50) 3: Sand (PM50–100)
0.20
0.34
0.39
0.45
0.08
0.40
0.10
0.29
0.27
1.00
0.40
0.56
0.32
0.28
0.12
a
Values are obtained from the U.S. soil texture triangle [Noorallah,
1999]. The mass fraction of each soil texture class available for uplift (g)
is also given [Nickovic et al., 2001].
and PM10tbr is the dust emission rate from tire wear, brake
wear and road abrasion processes.
[34] Mass and size fractions for road dust and vehicle
exhaust [Karanasiou et al., 2009] are averaged for the
estimation of speciated and size resolved re‐suspended
aerosol mass by vehicles (Table 1 and Figure 2, respectively). The deposited aerosol on the road network (especially at a city center location) comes mainly from road
transport, with vehicle exhaust contributing about equally
with dust production (tire and brake wear, road abrasion) (as
shown previously). This assumption will be assessed in a
sensitivity test concerning particle sizes. The alternative size
distribution (dotted line in Figure 2) is confined to the coarse
section. The latter is considered to be the result of collisions
between smaller and larger dust particles as they are set in
motion (saltation) before re‐suspension.
4.3. Aeolian Erosion
[35] Natural production of soil dust aerosols depends on
wind energy and is controlled by soil surface properties and
atmospheric characteristics. The parameterization of dust
production from Aeolian erosion of the soil surfaces, AE
(g cm−2 s−1) is given as a function of friction velocity (u*),
based on Shaw et al. [2008]:
AE ¼ fa C u*
4
fw u*t
1
u*
ð2Þ
where a is the vegetation mask, fab is the product of the
clay mass fraction of the soil (b) multiplied by its area
fraction fa in a grid cell, g is the mass fraction of the soil
available for uplift, C (g s3 cm−6) is a coefficient of
proportionality, fw is the soil wetness factor and u*t (cm
s−1) is a threshold friction velocity below which dust
emission does not occur. Values of u* in the current study
are from MM5 output for each computational cell on an
hourly basis. Values for g are shown in Table 2 [Nickovic
et al., 2001], while fa is from the MM5 database.
[36] The above equation estimates the total mass of dust
generated by the action of wind on the soil surface. The
chemical profile for AE (Table 1) is based on the soil source
profiles provided by Karanasiou et al. [2009]. Soil dust
emissions are largely calcium, as calcium carbonate is a
major component of loose soil and dust in Greece [Scheff
and Valiozis, 1990].
[37] Applications by Shaw et al. [2008] assumed dry soil
( fw = 1), semi‐desert and desert areas (a = 0.5 or 1,
respectively), u*t = 20 cm s−1 and C = 1 × 10−14 g s3 cm−6.
D17301
In the present work, the available soil and atmospheric data
by MM5 have been incorporated into functional approaches
for the parameters needed, replacing the default values.
Thus, local soil categorization and vegetation data are used,
as well as functional forms of fw, u*t and C. Furthermore,
quarries and earthwork sites are treated as explicit natural
dust sources.
4.3.1. Soil Characteristics
[38] Soil particles distribute among clay, silt and sand
texture classes and correspond to dust particles with diameters up to 2, 50 and 100 mm, respectively [Shaw et al.,
2008]. The mass fraction of each soil texture class depends on the category of the soil. In the current study, the
soil categorization is based on 14 soil categories and their
area fraction, obtained by MM5 database. In the Attica
region, there are four soil categories, each having a specific
mass fraction b of each soil texture class (Table 2) obtained
from the U.S. soil texture triangle [Noorallah, 1999]. Thus,
equation (2) gives up to 12 AE values for each computational cell, instead of a simple calculation based on average
soil characteristics of each cell.
[39] A first approximation of AE for Attica is based on the
fixed fw, u*t and C aforementioned values. AE ranges
between 10−10 to 10−6 g cm−2 s−1 (Figure 4) similar to what
was found in previous modeling and experimental studies
[Nickovic et al., 2001; Pulugurtha and James, 2006]. The
variation in AE for the same friction velocity reaches up to 2
orders of magnitude and depends on the soil characteristics
( fab values) of the dust productive areas in Attica.
[40] Local calculations of the dust size distribution in
Attica showed that PM2–50 accounts for 80% of the total
flux, followed by PM50–100 and PM2 (11 and 9%, respectively). This suggests that silt is the most dust productive
soil class, and is due to the high mass fraction available for
uplift (g). The sand mass exclusion from aerosol modeling
does not significantly underestimate the total Aeolian dust
production.
4.3.2. Vegetation
[41] Vegetation coverage of Attica is obtained from the
monthly MM5 database for each computational cell. The
vegetation mask, a, is the inverse of vegetation coverage,
ranging from zero (green field) to one (bare soil). For the
Attica area, a ranges from 0.72 to 0.87. The current calculation for dust emissions is a conservative dust estimate of
fluxes for a typical year because the simulation period is in
April, which is the most vegetated month of the year
(Figure 4). However, the variation of AE through the year,
in respect to vegetation, is small.
4.3.3. Soil Moisture
[42] Soil moisture increases the erosion threshold friction
velocity (u*t) due to increased cohesion. When the capacity
of the soil to absorb water is exceeded, soil dust does not
entrain into the atmosphere. This effect is included by using
the soil wetness factor ( fw) schemes [Nickovic et al., 2001;
Zender et al., 2003; Vautard et al., 2005; Cheng et al., 2007;
Choi and Fernando, 2008; Shaw et al., 2008]. This factor
depends on the soil category and is empirically determined
as a function of the clay soil content [Fécan et al., 1999].
Based on historical soil moisture values less than 16%,
Fécan et al. calculated an fw range of 1–2.7.
[43] In the present study, volumetric soil moisture data are
obtained by MM5 output for each computational cell on an
7 of 21
D17301
ATHANASOPOULOU ET AL.: DUST INCORPORATION IN PM SIMULATIONS
D17301
Figure 4. Total dust flux calculations (AE) as a function of friction velocity u*, from the active dust
emission areas of Attica after Shaw et al. [2008] assumptions. The bar plot shows total AE values
during the same wind conditions (u* > 1 m s−1), in respect to the yearly variation of vegetation in Attica,
as obtained by the MM5 database.
hourly basis. In order to be used in the fw calculation, they
are converted to gravimetric water content, wg [Zender et
al., 2003]. This conversion requires soil density data,
which depend on soil category, thus each cell bears up to
four wg values.
[44] MM5‐calculated wg are much higher than the soil
water capacity for all local soil categories (Figure 5). In
particular, wg always exceeds 11% and reaches up to 30%
over active dust emission regions, resulting in fw values
from 1.9 to 2.8, much higher than those derived by Fécan et
al. [1999]. Such high fw values can lead to unrealistically
low dust flux values. Therefore, the local soil water capacity
was multiplied by 1.5, similar to Zender et al. [2003],
leading to fw values in the range calculated by Fécan et al.
For the same wg values occurring in the Attica area, the soil
moisture effect is stronger as the soil category changes from
clay to clay loam, sandy clay loam and loam. The resulting
average threshold friction velocity value is 0.35 m s−1,
ranging from 0.2 (dry clay loam or clay soil) to 0.7 m s−1
(soil moisture highly above water capacity of a sandy clay
loam or clay loam soil), which is realistic [Marticorena et
al., 1997; Nickovic et al., 2001; Shaw et al., 2008]. This
soil moisture effect reduces the number of active dust
emission cells by 10% but does not affect the order of
magnitude of AE values calculated for a dry soil.
4.3.4. Coefficient of Proportionality
[45] Previous studies on Aeolian dust production have
expressed the coefficient of proportionality (equation (2))
involved in dust flux equations as a function of the acceleration gravity g (cm s−2) and of the air density rair (g cm−3)
[Marticorena et al., 1997; Zender et al., 2003; Vautard et
al., 2005]:
C ¼ fbc c air
g
ð3Þ
The parameter c was introduced by Zender et al. [2003] and
is a constant of proportionality with a value of 2.61 determined and confirmed by experiments [White, 1979; Greeley
and Iversen, 1985]. The factor fbc is related to soil crustation
and coverage [Vautard et al., 2005].
[46] The coefficient C given in g s2 cm−4 (equation (3)), is
divided by the kinematic viscosity of air (n = m / rair) (cm2
s−1), to provide C in g s3 cm−6, as necessary for equation (2).
C ¼ fbc c air 2
g
ð4Þ
m is the dynamic viscosity of air (g cm−1 s−1), calculated by
the Sutherland’s formula [Crane Company, 1988].
[47] In this study, the implementation of fbc as defined by
Vautard et al. [2005] gave large overestimations in calcium
concentrations when compared to Attica observations
[Scheff and Valiozis, 1990; Colbeck et al., 2002; Pateraki
et al., 2008]. This was partly attributed to the 2 order of
magnitude higher C values (driven by fbc) than those calculated by Shaw et al. [2008]. Therefore, in order to obtain
the order of magnitude of C values similar to those calculated by Shaw et al. [2008] and subsequently of the AE
values, and calcium concentration predictions, fbc is set to
8 of 21
D17301
ATHANASOPOULOU ET AL.: DUST INCORPORATION IN PM SIMULATIONS
D17301
Figure 5. Calculated soil wetness factor, fw [after Fécan et al., 1999] as a function of the gravimetric soil
moisture wg (MM5 output) for each soil category of Attica. Markers for fw = 1 reflect the water capacity
(w’) of each soil category. Empty symbols are based on fw calculations directly by Fécan et al. [1999],
while solid symbols are the tuned fw calculations.
4 × 10−5. This approximation results in C values ranging
from 6 to 9 × 10−14 g s3 cm−6.
4.3.5. Erosion Threshold Friction Velocity
[48] In order to avoid the large fractional errors in
equation (2) caused by the use of a constant threshold
friction velocity (u*t) value [Shaw et al., 2008], the Bagnold
[1941] formulation is applied, similar to previous studies
[Marticorena and Bergametti, 1995; Nickovic et al., 2001].
This formulation provides u*t as a function of soil class, thus
of particle size. Subsequently, erosion thresholds are multiplied by 0.9, to account for dust production during lower
wind speeds due to inertia [Nickovic et al., 2001].
[49] The effect of soil class on dust productivity is realized
when the erosion threshold friction velocity based on the
assumption of Shaw et al. [2008] is compared with the one
from Bagnold, after both are multiplied by the soil moisture
effect, fw. The latter results in u*t values from 0.2 to 1.4 m
s−1, for the Attica area, while previous u*t calculations did
not exceed 0.7 m s−1. Therefore, when incorporating the soil
class (particle size) effect [Bagnold, 1941] instead of the
first u*t approximation in this study [Shaw et al., 2008], dust
production from clay (PM2) is not significantly affected, but
silt and sand soil classes do not release particles (PM2–50 and
PM50–100, respectively), unless higher wind speeds (with u*
above 0.5 m s−1) prevail.
[50] This modification results in a decrease of the number
of active dust emission cells by 30%, causing a reduction in
the average dust flux of about 1 order of magnitude to the
aforementioned dust flux versions (Figure 6).
[51] Including the effect of roughness elements on the
threshold friction velocity of adjacent erodible surfaces, as
described by Marticorena and Bergametti [1995], was attempted, but rejected from the parameterization. Considering the effect of roughness elements (drag partition), u*t was
increased by 3–4 times, resulting in little or no dust pro-
duction in Attica, which is unrealistic given the local
observations of dust aerosol tracers [Colbeck et al., 2002;
Pateraki et al., 2008; Karanasiou et al., 2009].
4.3.6. Quarries and Earthwork Sites
[52] Anthropogenic disturbances by quarry and construction activities cause additional dust emissions, due to vehicle
circulation and intense surface dust loading in the greater
area of these sites. In the current parameterization, the wind
erosion mechanism is simulated, as vehicle load data are not
available. A simple source function [Gillette and Passi,
1988] is applied in each soil texture class (i) at the quarry
and construction areas of the nested domain. The simplifications for the specific dust emission calculations are that
the soil is dry and sandy and the simulated soil dust is
equally distributed in the 5 coarser size sections, as crustal
material. The coordinates and dimensions of the quarry and
earthwork sites are taken from recent Google earth satellite
images (Figure 7d).
5. Results
5.1. Dust Emissions
[53] The updated vehicle exhaust emission rate (PM10exh)
on a typical weekday in Attica range from 3 to 4.4 × 103 kg
day−1 (Figure 7a). This variation depicts the uncertainty
(15–25%) due to statistical and systematic errors in the
applied emission factors, driving pattern, traffic volume,
spatial distribution of fleet composition and temporal resolution of emissions [Kühlwein and Friedrich, 2000]. These
emissions reflect the 20% of the total conventional aerosol
emissions, while the other 80% of the mass reflects maritime, off‐road and agricultural activities, railways and air
traffic aerosol emissions.
[54] The above emission projection is evaluated against a
rough emission calculation based on EEA [2007] exhaust
9 of 21
D17301
ATHANASOPOULOU ET AL.: DUST INCORPORATION IN PM SIMULATIONS
D17301
Figure 6. As in Figure 4 but after all modifications from the current study.
emission factors and road transport data for 2005 [Hellenic
Ministry of Environment, 2006]. The result, PM10exh = 3.6 ×
103 kg day−1, falls into the applied mass emission range.
[55] Emission rates (PM10tbr) from dust production processes (tire wear, brake wear and road abrasion) on a typical
weekday in Attica range from 1.6 to 3.5 × 103 kg day−1
(Figure 7b). This variation stems from applying the
parameterization over the ranges of proposed emission
factors [EEA, 2007], vehicle speed (30–100 m s−1), load
factor of heavy vehicles (0–1) and the number of axles of
heavy vehicles (2–6). The dust source strength is slightly
less than that of vehicle exhaust, though for particles with
diameters up to 40 mm, dust production ranges from 2.6 to
6.5 × 103 kg day−1 due to the large coarse fraction produced
by tire wear and road abrasion (Figure 2). The relative
contribution from brakes, tires and road abrasion, to dust
production is 40, 30 and 30% of PM10 mass, respectively.
These values change to 27, 32 and 41% of PM40 mass,
respectively.
[56] Vehicle‐induced re‐suspension (Figure 7c) dominates
vehicle‐induced production of aerosol (exhaust, wear processes and road abrasion). In particular, the calculated
PM10res on a typical weekday in Attica ranges from 16 to
18 × 103 kg day−1, 65% of which is PM2.5 (Figure 2). This
suggests that traffic‐induced re‐suspension is a major
aerosol source for Attica, as also found for the greater Paris
area [Jaecker‐Voirol and Pelt, 2000]. Therefore, road
aerosol emissions are likely underestimated by a factor of
4–5, at the Attica area if only car exhaust is considered, while
this factor reduces to 2–3 if re‐suspended mass is compared
to all production processes.
[57] Aeolian erosion emission rates are calculated for the
10 simulated days in Attica (Figure 8). In the evening and
early morning hours, AE is minimal due to calm winds near
the land surface and increased humidity that prevail during
this time of the days examined. The dust production period
is limited to the driest daylight hours, with its peak around
1100 and 1600. Peak comparisons among all days give an
indication of the average wind regime over the whole simulated domain. The highest hourly dust production (12/04/’06,
13:00) corresponds to a spatial mean friction velocity of
0.9 m s−1, while the lowest peak (11/04/’06, 15:00) is
produced by a mean value around 0.2 m s−1.
[58] The AE production for these 10 days ranges from 0.8
to 255 × 103 kg day−1. In order to examine more closely the
atmospheric and soil effects on dust production by Aeolian
erosion processes, April 10, 2006 is selected for further
analysis (Figure 7d). For this particular day, under a wide
range of surface wind conditions (u* ranges from 0.2 to
1.1 m s−1) the calculated AE value is 14 × 103 kg day−1,
emitted from areas away from the urban core. An analysis of
AE data for the different soil categories existing in Attica
(clay, clay loam, loam and sandy clay loam) revealed that a
dry ( fw = 1) clay soil releases 3 orders of magnitude more
dust than a moist (fw = 1.7) sandy clay loam soil under the
same wind conditions (u* = 0.6 m s−1).
5.2. Simulated Aerosol Concentrations
5.2.1. PM10 and PM2.5
[59] Daily average comparisons of PM10 concentrations at
the measuring sites find a low bias in the model simulated
values (Table 3). PM10 and PM2.5 comparisons show an
average model underestimation, but average PM2.5/PM10
ratios are predicted higher than observed (73 and 61%,
respectively). This indicates an average trend for coarse
aerosol underestimation. Model performance is greatly
10 of 21
D17301
ATHANASOPOULOU ET AL.: DUST INCORPORATION IN PM SIMULATIONS
D17301
Figure 7. Daily average aerosol emission rates (g hr−1 km−2) in Attica, by (a) vehicle exhaust, (b) vehicle dust production (tire wear, brake wear and road abrasion), (c) vehicle re‐suspension, and (d) Aeolian
erosion of soil (10/04/2006). Black symbols show the sites of tolls, tunnels and junctions of the highway
in Figure 7a and quarries and earthwork sites in Figure 7d. The daily average aerosol emission rates of
these sites are below 20 g hr−1 km−1.
improved when excluding from comparisons not only the
Saharan dust intrusion day, but all the events with winds
above 4 m s−1. Increased atmospheric cleansing effectiveness is rejected as a reason for model discrepancies, since
wind speeds are slightly underestimated by MM5. It is
argued that local wind‐dependent aerosol sources are not
completely captured in the current inventory.
[60] Results compare well at the ARI, PIR, MAR, AGP
and DEM sites, indicating that anthropogenic aerosol sources including road dust are better captured inside or near the
Figure 8. Total hourly emission rates of dust (103 kg hr−1) generated by Aeolian erosion processes over
the whole Attica domain, during 10 simulated days. Black vertical lines define the 24‐h period
corresponding to the average daily values referred in this article.
11 of 21
12 of 21
26/05
30, 46
32
34
42
64
12/04
16, 9
4.4 (5.3f), 5.1 (0.9f)
W, W
49,f 56f
17, 17
1.7 (1.1f), 2.9 (0.9f)
SSW, SSW
56,f 60f
14, 14
1.0, 2.7
SW, NW
68,g 63g
20, 21
54
29
63
94
23, 20
20,
23,
23,
28,
27, 53
36, 82
28, 117
10, 137
2006
1.2 (2.7f), 1.9 (0.3f)
SSW, SSW
48,f 61f
17, 16
37, 65
45, 90
43, 56
24,
31,
27,
37,
11/04
62, 67
45, 102
23, 56
34, 42
55, 54
34, 55
16, 34
10/04
18, 35
1.8, 5.2
NE, NE
75,g 71g
19, 21
16, 63
14, 36
24, 22
19, 36
14, 33
25, 47
44,
24,
28,
20,
36,
52,
15, 35
17, 45
43, 56
17, 38
4, 20
38, 42
17, 36
11, 27
13, 65
2005
38
24
58
47
35
57
63, 48
52, 36
17, 16
25/05
1.9 (3.5f), 2.7 (0.5f)
SSW, SSW
53,f 56f
14, 14
28, 39
45, 57
22, 26
36, 28
50, 43
45, 58
13, 29
13/04
3.4, 5.7
NE, SW
45, 41
16, 18
8, 18
15, 21
1.2, 3.2
SW, SW
50, 49
15, 18
7, 46
16, 27
14, 39
13, 24
14, 26
27, 45
12, 49
10, 42
2.1, 3.1
SW, NE
46, 50
23, 23
21, 36
19, 34
14, 22
32, 49
58
72
21
48
54
38
35
17/05
38,
29,
13,
45,
30,
21,
21,
2007
26/04
22, 35
33, 40
31, 50
18, 30
21, 32
62, 48
29, 55
12, 27
25/04
39
41
12
31
29
25
20
3.24.5
S, NNE
57, 64
18, 20
17, 19
8, 18
7, 11
15, 25
23,
23,
10,
23,
19,
12,
12,
18/05
Errorc
0.33
0.71
0.80
0.15
0.46
0.53
0.58
0.38
0.51
0.60
0.41
0.90
0.84
0.35
0.54
0.65
0.69
0.50
0.45
0.57
−0.19
−0.63
−0.79
−0.15
−0.38
−0.52
−0.58
−0.38
−0.50
−0.60
−0.41
−0.90
−0.84
−0.26
−0.51
−0.65
−0.69
−0.50
0.02
−0.45
−0.45
−0.05
−0.45
−0.53
−0.18
−0.22
−0.44
−0.37
−0.09
−0.32
−0.38
−0.37
Biasd
0.31
0.24
0.55
0.54
0.18
0.33
0.45
0.37
0.09
0.34
0.38
0.37
Errore
Mean Fractional Values
Biasb
N
i¼1
2
jCeCoj , where Ce are the modeled and Co are the measured PM10 values for the i out of N sites.
CoþCe
2
d
Same as footnote b but excluding events with wind speeds above 4 m s−1.
e
Same as footnote c but excluding events with wind speeds above 4 m s−1.
f
Refers to values from DEM.
g
Rain event.
MFE =
c
N
P
1
a
Bold indicates modeled results, and regular font indicates measured results, in mg m–3. Observations for aerosol are from the National Air Pollution Monitoring Network (except from DEM data of this study).
Meteorological data are from and the National Observatory of Athens.
N
P
b
ðCeCoÞ .
MFB = N1
i¼1
CoþCe
ARI
LYK
THR
PIR
MAR
ZOG
AGP
DEM
COL
IRA
GOU
HER
PAT
VOL
Average
PM2.5
LYK
PIR
AGP
DEM
Average
Wind speed (m s−1)
Wind direction
RH (%)
Temperature (C)
PM10
Table 3. Daily Average PM10, PM2.5, Meteorological Values and Evaluation Statistics, for Modeling Periods, at Several Greek Stationsa
D17301
ATHANASOPOULOU ET AL.: DUST INCORPORATION IN PM SIMULATIONS
D17301
D17301
ATHANASOPOULOU ET AL.: DUST INCORPORATION IN PM SIMULATIONS
D17301
Figure 9. Diurnal (April 10 and 11, 2006) variation of observed (dots) and predicted (black lines) PM10
values (mg m−3), at the highway sites of (a) IRA and (b) COL. Red lines show the diurnal variation of the
conventional anthropogenic emissions applied in the model (daily average value around 2 kg hr−1 km−2).
urban core. Increased model efficiency during weak wind
conditions shows the importance of coarse dust aerosol
production when wind acts on urban surfaces. The underestimation at the highway stations (COL and IRA) is readily
explained since the model considers the intense local
emissions homogeneously and instantaneously mixed over a
grid cell, while the measured morning peaks reflect the local
emission regime (Figure 9). These discrepancies become
more obvious during the afternoon hours when the mixing
height predictions become higher. This is also supported by
the low predicted aerosol, NO2 and CO values, despite the
emission rates that follow the daily variation of measurements remaining high during the afternoon (Figure S1 in the
auxiliary material).1 The morning and night traffic peaks are
well captured by the model. Increased discrepancies at THR
and LYK (Table 3) suggest greater emission uncertainties in
these rapidly developing, suburban areas, due to construction activities and other wind‐related processes. Model
discrepancies at HER and PAT (Table 3) are mainly
attributed to the point (measurements) being less reflective
of area predictions, as the model resolution at the periphery
of Greece is lower (6 km).
5.2.2. Sensitivity Study
[61] A sensitivity analysis is conducted in order to assess
the response to inclusion of road and soil dust production
mechanisms in aerosol modeling. The 10th of April is
selected to show the current dust incorporation effects to
predictions (Figure 10). Comparison of simulations with and
without dust emissions shows that the dust component
corresponds to about 15–40% (5–20 mg m−3) of the total
PM10 predictions. Larger dust mass quantities at the traffic‐
oriented stations (ARI, MAR, COL, IRA) indicate that
vehicle‐induced dust is the major PM10 dust component
near the road network. This is tested with a simulation
without the natural component of dust emissions (Aeolian
erosion). Subtracting results from the total PM10 predictions
1
Auxiliary materials are available in the HTML. doi:10.1029/
2009JD013207.
shows insignificant Aeolian soil dust fraction at the traffic
areas. Results in the urban core (ARI) indicate that road dust
emissions are well estimated by the model. The emission
rates of dust re‐suspension from the road network greatly
dominate other primary aerosol production mechanisms
(section 5.1) and this process typically is responsible for a
large majority of the dust component of aerosol in the
greater Athens area. The exception to the above occurs
when high winds are observed above rather dry soil surfaces
with a large fraction of clay. In such conditions, Aeolian
erosion can produce much greater quantities of dust than the
road re‐suspension process (section 5.1). Nevertheless, in
the greater Athens area road dust still dominates because
such land use characteristics occur far from the Athens
basin.
[62] In order to minimize arbitrary approximations, in this
study the constant input parameters in the emission parameterizations schemes were replaced with functional forms,
when this was feasible. Nevertheless, the coefficient fbc
(in equation (4)), that forms directly the order of magnitude
of dust emission rates, remains a constant. Thus, its value
is settled so that concentration predictions are reasonable
according to local measurements (section 4.3).
5.2.3. Wind Fields
[63] MM5 predicted calmer conditions during the 11th of
April and consequently increased simulated pollution
accumulation from CAMx (“non‐dust PM10” in Figure 10),
but less than observed. Diurnal variation of aerosol levels
(Figure 9) confirms that although the applied emission
database is similar for each weekday, the model captures the
different morning peaks. Nevertheless, a possible increase in
real emissions during the 11th of April, might also contribute to the existing discrepancies. High observed PM10 at
LYK is similar to the next day’s measurement that is
diagnosed to be a Saharan dust event, although all other
stations seem unaffected.
[64] Comparison of simulated inorganic aerosol mass to
measurements varies by day. Simulations of the events in
10–11 and 12–13 of April, 2006 show better performance
13 of 21
D17301
ATHANASOPOULOU ET AL.: DUST INCORPORATION IN PM SIMULATIONS
D17301
Figure 10. Measured and simulated daily average PM10 concentrations (mg m−3) in the Attica area
(April 10 and 11, 2006). Location of all sites is shown in Figure 1. Two measuring sites (linked with
an error bar) exist in the location named as LYK.
(Figures 11 and 12). DEM wind field data for the two events
are similar (SW winds around 1 m s−1), and reflect the
Saronic Gulf breeze effect frequently observed and simulated in Athens [Melas et al., 1998; Dandou et al., 2009].
During a short afternoon period of the 10th of April, the
wind direction changes to SE and intensifies up to 3 m s−1 at
the DEM station. This is already characterized as the effect
of the Evoicos Gulf breeze that cannot be captured by
modeling at this site [Bossioli et al., 2007]. Local aerosol is
found to contain sodium, chloride and magnesium in the
fine particles that are possibly attributed to the SE air mass
entrainment. The fine potassium peak around the diameter
of 0.3mm may be related to biomass burning [Karanasiou et
al., 2009] from sources located southeasterly of DEM. The
applied emission inventory does not currently involve biomass burning activities, thus, even if the model could capture the SE air intrusion, it couldn’t have shown increases in
potassium.
5.2.4. Sodium, Chloride, Magnesium, and Potassium
[65] Simulated sodium compares very well to measurements, while chloride observations show very low values,
possibly due to depletion during transport as observed in
earlier studies [Eleftheriadis et al., 2006b]. Predictions of
both species follow the regular size profile reported for
seawater [Lewis and Schwartz, 2005] and are predicted to be
sea‐salt dominated (95%). The sea‐salt origin of sodium and
chloride is supported by aerosol observations during 12–13/
04/2006, due to constant onshore WSW winds.
[66] Magnesium shows the best performance of all species. Sodium to magnesium PM10 mass ratio is calculated
(around 6) similar to the value (of 8) reported for seawater
[Lewis and Schwartz, 2005], while the measured and modeled size distribution of the coarser fraction follows that of
sodium chloride during 12–13/04/2006. The highest marine
contribution to magnesium mass (75%) is reached during
this period, as expected. Observations at a southwestern
Mediterranean coast also showed the almost exclusive
marine origin of magnesium [Nicolás et al., 2009]. Predictions for potassium are limited to its coarse part, which
follows measurements, especially for 10–11/04/2006. Sea
salt and dust contribute equally to its predicted value.
5.2.5. Calcium
[67] Wind field data (Table 3) for these two successive
events give a clear indication that the synoptic flow during
12–13/04/2006 (W and SSW winds with hourly values up to
11 m s−1) decreases PM2.5/PM10 values and total PM10
concentrations at DEM. This is depicted in both measurements and predictions (Figures 11 and 12) and shows the
importance of fine aerosol during low winds, which are
effectively removed by strong winds. The simultaneous
production of local dust can be identified by calcium, which
is commonly used as a tracer for dust aerosol.
[68] Aeolian dust production from soil surfaces results in
reasonable calcium levels according to this and other studies
[Scheff and Valiozis, 1990; Colbeck et al., 2002; Pateraki
et al., 2008]. Calcium predictions are around 2–4 mg m−3
(Figure 13a) and the size distribution predicted in areas
located near high Aeolian dust emissions follows measurements during this study (bold line in Figure 11). Simulated
calcium levels at DEM are below observations, largely in
the coarse fraction. This supports previous concerns of a
missing wind‐driven but local dust source, possibly
corresponding to re‐suspension from urban surfaces, such as
the road network and its adjacent open land surfaces or local
construction areas.
[69] The model underestimation of total aerosol and calcium in the PM10–40 fraction intensifies during 12–13/04/
2006 because of the Saharan dust intrusion. Measured winds
at DEM for this event are similar to the previous (Table 3)
supporting that increases in PM10–40 measurements is dust
transported from the Sahara desert by the higher synoptic
winds. The primary aerosol peak around 10 mm sums up to a
difference of 9.4 mg m−3 between the measured PM40 concentrations of the two events. Focusing on PM10 values, the
14 of 21
D17301
ATHANASOPOULOU ET AL.: DUST INCORPORATION IN PM SIMULATIONS
Figure 11. Size distribution of measured (dots) and modeled (lines) values of aerosol species (10–11 of
April, 2006, DEM site). The bottom line shows simulated non‐dust aerosol values. The top line shows
total simulated species concentrations. The striped area is the dust component of aerosol. The bold line
in calcium plot shows Aeolian erosion calcium at a high erosion emissions cell. The bold line in PM plot
shows the output of the alternative size profile for re‐suspension. Total PM10 values (PM10m: modeled /
PM10o: measured, in mg m−3) are shown at the top right corner of each species plot.
15 of 21
D17301
D17301
ATHANASOPOULOU ET AL.: DUST INCORPORATION IN PM SIMULATIONS
D17301
Figure 12. As in Figure 11 but for 12–13 of April, 2006.
Saharan dust contribution corresponds to 21% of PM10,
which is close to 25%, found in an Athens suburb by
Mitsakou et al. [2008]. Calcium exhibits a primary coarse
peak around the same diameter that corresponds to the half
mass of Saharan dust contribution (4.7 mg m−3).
5.2.6. Carbon
[70] Elemental and organic carbon were simulated but not
measured for this study. Their calculated dust fraction
corresponds to 55% and 75% of their PM10 mass (3.5 and
4.5 mg m−3, respectively). Available PM10 measurements
16 of 21
D17301
ATHANASOPOULOU ET AL.: DUST INCORPORATION IN PM SIMULATIONS
D17301
Figure 13. Spatial distribution of average daily (12–13/04/2006) concentrations (mg m−3) over Attica:
−
(a) PM 10 calcium (Ca 2+ ), (b) PM 10 sulfate (SO 2−
4 ), (c) PM 10 nitrate (NO 3 ), (d) PM 10 ammonium
+
(NH4 ), (e) total PM2.5, and (f) total PM10. The black dot indicates the DEM measuring station.
of elemental and organic carbon showed average values of
1.8 ± 0.6 and 5.7 ± 2.8 mg m−3, respectively [Prosmitis et al.,
2003]. Total carbon predictions are satisfactory, while the
organic fraction is largely improved after dust incorporation
due to its large fraction that is re‐suspended from road surfaces. Thus, including wind‐driven aerosol re‐suspension
from the road network is believed to further improve the
organic carbon model performance. Simultaneous carbon
measurements would help to evaluate tire wear, road abrasion and re‐suspension algorithms (Table 1) and assess
wind‐driven re‐suspension effects.
5.2.7. Secondary Inorganic Species
[71] Simulated sulfate concentrations are moderately
lower than the observed values during calm conditions
(Figure 11). Results from a process analysis of measurement
data conducted by the National Air Pollution Monitoring
17 of 21
D17301
ATHANASOPOULOU ET AL.: DUST INCORPORATION IN PM SIMULATIONS
Network revealed that SO2 values have increased since 2004
at Athens traffic sites, possibly reflecting a higher sulfur
content in vehicular fuel not being reported in the emission
database. This could lead to the poor correlation between
fine PM measurements and simulations during this event.
Another reason for the differences could be the SE afternoon
winds that the meteorological model failed to capture. The
windier event shows a good model performance (Figure 12).
Sulfate‐rich air masses are predicted to come from the
industrial area west of the city center (Figure 13b). A sensitivity simulation performed without the conventional area
sources confirms the industrial origin of sulfate. Measured
values during 12–13/04/2006 drop due to the higher winds
that disperse the fine aerosol emitted by the sulfate‐rich
vehicular fuel combustion or other area sources.
[72] Nitrate and ammonium concentrations (Figures 13c
and 13d) depend on the relative molar quantities of sulfate, ammonium, crustal and sodium species [Fountoukis
and Nenes, 2007]. Observations and predictions of aerosol
over the greater Athens area show a regime poor in sulfate
and crustal species but rich in sodium. Comparison of
simulations with and without dust emissions shows that the
dust component of PM10 sulfate corresponds to about 10–
15% (Figures 11 and 12). Reactions of sulfuric acid on sea
salt and crustal particles leave little sulfate for ammonia to
neutralize.
[73] Free ammonia reacts with nitric acid to form
ammonium nitrate (NH4NO3). Total nitrate compares very
well to measurements, while total ammonium is found
moderately overestimated. Their predicted amounts are
slightly affected by dust production (Figures 11 and 12).
This is also deduced from coarse nitrate measurements,
which follow the sea‐salt profile of sodium (12–13/04/
2006). Its measured coarse peak value is similar between
events regardless the large calcium differences due to Saharan dust. It is suggested that adjustments in the ammonia
emission values are the major reason for the good total
nitrate and ammonium model performance, though may
have been slightly overestimated.
[74] Differences between measured and predicted size
distributions of secondary species are found mainly in the
fine fraction. The observed sulfate peak around 0.6mm is
followed but underestimated (Figure 11). In the measurements, nitrate and ammonium concentration peak at the
same diameter as the sulfates, while in the model the peak of
the fine mode is predicted at around 0.2mm. Measurements
reflecting 0.6mm mass fraction, show air masses rich in
sulfate, thus sulfate captures ammonia toward ammonium
sulfate [(NH4)2SO4] formation and then nitrate reacts with
available ammonia toward NH4NO3. A difference in RH
between measurements and predictions (Table 3) may have
caused differences in the deliquescence and subsequent
growth in ionic mixtures. Higher RH measurement values
can lead to increased deliquescence and growth of NH4NO3
particles, while lower RH predictions limit the hydroscopic
growth of particles in the model.
5.2.8. Total Aerosol Mass
[75] The size distribution of total aerosol mass predictions follows measurements (Figure 11), though a shift of
mass from finer to coarser bins could improve performance. The application of the alternative size distribution
of re‐suspended material (Figure 2) caused a decrease in
D17301
simulated PM10 due to the increased dry deposition flux of
coarse particles. The effect of the size shift on fine aerosol
shows that vehicle‐induced re‐suspension is the basic component of dust aerosol at DEM. This finding follows results
in dust emissions (paragraph 5.1), expected for traffic‐
affected areas. Coarse aerosol is still underestimated, showing that size modifications in the applied source profiles
cannot compensate model discrepancies. Total aerosol mass
closure is mainly affected by deficiencies in particle diameters greater than 10mm. This chemically unresolved sand
material is regarded of minor health impact, thus of less
importance for air quality management and modeling.
[76] During the moderate rainy event of 25–26/05/2005
(Figure S2 in the auxiliary material), simulated levels of all
species are low, while fine particles dominate aerosol
measurements. The wet deposition effect is predicted very
efficient in all sizes, while fine aerosol measurements seem
unaffected by deposition processes possibly due to a positive correlation of humidity with aerosol of such sizes
[Nicolás et al., 2009].
6. Conclusions
[77] An emission inventory for road dust emissions
developed from detailed parameterizations finds that vehicle‐induced re‐suspension is the major contributor to aerosol emissions from the Athens’ road network, followed by
brakes, tires and road abrasion. The simultaneous parameterization of soil dust production can give higher emissions,
though located in less urbanized but more extended areas at
the periphery of the city. Not only does the use of the
developed method extend model capabilities to better capture dust levels, but also these results show that inclusion of
a spatially temporally more accurate approach to estimating
dust emissions does alter the concentrations of related pollutant species, and can improve how accurately air quality
models simulate the impact of emission sources on air
quality.
[78] Comparisons of PM10 concentrations at traffic‐
affected stations show that incorporating dust of different
types improves model accuracy. Toward the periphery of
the urban core, dust appears to be underestimated, most
probably because its wind component is misrepresented by
the model in areas where soil surfaces co‐exist with
urbanized land. Incorporation of an emission scheme representing wind‐induced particle re‐suspension from urban
surfaces is believed to further improve model’s efficiency.
[79] Carbon, calcium, potassium, magnesium and sulfates
are the influenced species by incorporating dust in the
aerosol modeling. Calcium, which is regarded as a soil dust
tracer, reaches the measured values outside the city center
where high Aeolian dust emissions are simulated. Saharan
dust is currently omitted from the applied dust emissions,
thus model results concerning crustal species were expected
to be poor. However, predictions concerning sodium,
chloride and magnesium species can be compared to observations during such episodes, given the strong southern
winds that form well‐established sea‐salt profiles.
[80] A meteorological parameter that appears critical to
aerosol predictions is the diffusion efficiency of the atmosphere. During the afternoon and evening hours, model
discrepancies increase although the emission trend follows
18 of 21
D17301
ATHANASOPOULOU ET AL.: DUST INCORPORATION IN PM SIMULATIONS
the observations. This finding is attributed to an overestimation in the mixing height from meteorological simulations and is clearly illustrated at sites adjacent to emission
sources. Humidity is also found to affect model performance, particularly impacting the size distribution of secondary aerosol formation. The atmosphere is predicted to be
drier than measured, thus fine ammonium nitrate predictions
peak at smaller sizes due to limited deliquescence and
growth.
[81] Total secondary inorganic aerosol depends on sulfate
production and formation. The spatial distribution of sulfates shows a strong industrial origin and low concentrations
are predicted inside the urban core during lower winds. This
allows for greater ammonium nitrate production, which was
well captured by the model. During higher wind conditions
observed sulfate concentrations are lower but not in the simulations, indicating that their local origin is not fully
captured in the emission database. Higher sulfate predictions
lead to nitrate underestimations due to increased ammonia
being tied up with sulfate.
[82] The parameterization developed in this work can be
used in other urban, regional and global modeling applications, which can be combined so as the boundary conditions
for the chemical species, included dust, to be provided to
regional applications by the global models. Having a more
accurate approach to simulating emissions across scales can
provide consistent dust concentration fields for assessing
impacts on not only air quality but radiative transfer as well.
[ 83 ] Acknowledgments. This work was supported by PENED,
which is co‐financed by E.U.‐European Social Fund (75%) and the Greek
Ministry of Development‐GSRT (25%). We want to thank V. A. Karydis,
A. P. Tsimpidi, and S. N. Pandis for use of the new code routines employed
in the model. We also acknowledge the provision of the emission data employed in the study by the Greek Ministry of the Environment and the
HPC‐PASECO company.
References
Abu‐Allaban, M., S. Hamasha, and A. W. Gertler (2006), Road dust resuspension in the vicinity of Limestone Quarries in Jordan, J. Air Waste
Manage. Assoc., 56, 1440–1444.
Alfaro, S., and L. Gomes (2001), Modeling mineral aerosol production by
wind erosion: Emission intensities and aerosol size distributions in source
areas, J. Geophys. Res., 106(D16), 18,075–18,084, doi:10.1029/
2000JD900339.
Alfaro, S., A. Gaudichet, L. Gomes, and M. Maillé (1998), Mineral aerosol
production by wind erosion: Aerosol particle sizes and binding energies,
Geophys. Res. Lett., 25(7), 991–994, doi:10.1029/98GL00502.
Alfaro, S. C., S. Lafon, J. L. Rajot, P. Formenti, A. Gaudichet, and
M. Maillé (2004), Iron oxides and light absorption by pure desert dust:
An experimental study, J. Geophys. Res., 109, D08208, doi:10.1029/
2003JD004374.
Allwine, K. J., F. C. Rutz, W. J. Shaw, J. P. Rishel, B. G. Fritz, E. G.
Chapman, B. L. Hoopes, and T. E. Seiple (2006), DUSTRAN 1.0 user’s
guide: A GIS‐based atmospheric dust dispersion modeling system, Tech.
Rep., PNNL‐16055, Pac. Northwest Natl. Lab., Richland, Wash.
Alpert, P., P. Kishcha, A. Shtivelman, S. O. Krichak, and J. H. Joseph
(2004), Vertical distribution of Saharan dust based on 2.5‐year model
predictions, Atmos. Res., 70(2), 109–130, doi:10.1016/j.atmosres.2003.
11.001.
Astitha, M., and G. Kallos (2008), Gas‐phase and aerosol chemistry interactions in south Europe and the Mediterranean region, Environ. Fluid
Mech., 9, 3–22, doi:10.1007/s10652-008-9110-7.
Athanasopoulou, E., E. Bossioli, and M. Tombrou (2005), Modelling of
aerosol in the greater Athens area, Greece, Int. J. Environ. Pollut.,
24(1–4), 230–246, doi:10.1504/IJEP.2005.007395.
Athanasopoulou, E., M. Tombrou, S. Pandis, and A. Russell (2008), The
role of sea‐salt emissions and heterogeneous chemistry in the air quality
D17301
of polluted coastal areas, Atmos. Chem. Phys. Discuss., 8(1), 3807–3841,
doi:10.5194/acpd-8-3807-2008.
Bagnold, R. (1941), The Physics of Blown Sand and Desert Dunes,
Methuen, New York.
Bauer, S. E., Y. Balkanski, M. Schulz, D. A. Hauglustaine, and F. Dentener
(2004), Global modelling of heterogeneous chemistry on mineral aerosol
surfaces: The influence on tropospheric ozone chemistry and comparison
to observations, J. Geophys. Res., 109, D02304, doi:10.1029/
2003JD003868.
Bossioli, E., M. Tombrou, A. Dandou, and N. Soulakellis (2007), Simulation of the effects of critical factors on ozone formation and accumulation
in the greater Athens area, J. Geophys. Res., 112, D02309, doi:10.1029/
2006JD007185.
Capaldo, K. P., C. Pilinis, and S. P. Pandis (2000), A computationally efficient hybrid approach for dynamic gas/aerosol transfer in air quality
models, Atmos. Environ., 34, 3617–3627, doi:10.1016/S1352-2310(00)
00092-3.
Chang, C. T., and C. J. Tsai (2003), A model for the relative humidity
effect on the readings of the PM beta‐gauge monitor, J. Aerosol Sci.,
34(12), 1685–1697, doi:10.1016/S0021-8502(03)00356-2.
Chang, J. S., R. A. Brost, I. S. A. Isaksen, S. Madronich, P. Middleton,
W. R. Stockwell and C. J. Walcek (1987), A three‐dimensional Eulerian
acid deposition model: Physical concepts and formulation, J. Geophys.
Res., 92(D12), 14,681–14,700, doi:10.1029/JD092iD12p14681.
Cheng, T., Y. Peng, J. Feichter, and I. Tegen (2007), An improvement on the
dust emission scheme in the global aerosol‐climate model ECHAM5‐
HAM, Atmos. Chem. Phys. Discuss., 713,959–13,987, doi:10.5194/
acpd-7-13959-2007.
Chin, M., P. Ginoux, S. Kinne, O. Torres, B. N. Holben, B. N. Duncan,
R. V. Martin, J. A. Logan, A. Higurashi, and T. Nakajima (2002),
Tropospheric aerosol optical thickness from the GOCART model
and comparisons with satellite and sunphotometer measurements,
J. Atmos. Sci., 59, 461–483, doi:10.1175/1520-0469(2002)059<0461:
TAOTFT>2.0.CO;2.
Chin, M., D. A. Chu, R. Levy, L. Remer, Y. Kaufman, B. N. Holben,
T. Eck, and P. Ginoux (2004), Aerosol composition and distributions in
the Northern Hemisphere during ACE‐Asia: Results from global
model, satellite observations, and surface sunphotometer measurements,
J. Geophys. Res., 109, D23S90, doi:10.1029/2004JD004829.
Chin, M., T. Diehl, P. Ginoux, and W. Malm (2007), Intercontinental transport of pollution and dust aerosols: Implications for regional air quality,
Atmos. Chem. Phys., 7(21), 5501–5517, doi:10.5194/acp-7-5501-2007.
Choi, Y., and H. Fernando (2008), Implementation of a windblown dust
parameterization into MODELS‐3/CMAQ: Application to episodic PM
events in the US/Mexico border, Atmos. Environ., 42(24), 6039–6046,
doi:10.1016/j.atmosenv.2008.03.038.
Colbeck, I., M. C. Chung, and K. Eleftheriadis (2002), Formation and
transport of atmospheric aerosol over Athens, Greece, Water Air Soil
Pollut. Focus, 2(5–6), 223–235, doi:10.1023/A:1021335401558.
Crane Company (1988), Flow of fluids through valves, fittings, and pipe,
Tech. Pap., 410, Stamford, Conn.
Dall’Osto, M., and R. M. Harrison (2006), Chemical characterisation of
single airborne particles in Athens (Greece) by ATOFMS, Atmos. Environ., 40(39), 7614–7631, doi:10.1016/j.atmosenv.2006.06.053.
Dandou, A., M. Tombrou, E. Akylas, N. Soulakellis, and E. Bossioli
(2005), Development and evaluation of an urban parameterization
scheme in the Penn State/NCAR Mesoscale Model (MM5), J. Geophys.
Res., 110, D10102, doi:10.1029/2004JD005192.
Dandou, A., M. Tombrou, K. Schäfer, S. Emeis, A. Protonotariou,
E. Bossioli, N. Soulakellis, and P. Suppan (2009), A comparison between
modelled and measured mixing‐layer height over Munich, Boundary
Layer Meteorol., 131(3), 425–440, doi:10.1007/s10546-009-9373-7.
Eleftheriadis, K., D. Balis, I. Ziomas, I. Colbeck, and N. Manalis (1998),
Atmospheric aerosol and gaseous species in Athens, Greece, Atmos.
Environ., 32(12), 2183–2191, doi:10.1016/S1352-2310(97)00412-3.
Eleftheriadis, K., K. Ochsenkuehn, T. Lyberopoulou, P. Razos, and
M. Ochsenkuehn‐Petropoulou (2006a), Size distribution of inorganic
water soluble species in the Mediterranean suburban background atmospheric aerosol, paper presented at Atmospheric Chemistry at the Interfaces, Comm. on Atmos. Chem. and Global Pollut., Cape Town, South
Africa, 17–22 Sept.
Eleftheriadis, K., I. Colbeck, C. Housiadas, M. Lazaridis, N. Mihalopoulos,
C. Mitsakou, J. Smolik, and V. Zdimal (2006b), Size distribution, composition and origin of the submicron aerosol in the marine boundary layer
during the eastern Mediterranean “SUB‐AERO” experiment, Atmos.
Environ., 40, 6245–6260, doi:10.1016/j.atmosenv.2006.03.059.
ENVIRON (2003), Development of an Advanced Photochemical Model for
Particulate matter: PMCAMx, Novato, Calif. (Available at http://www.
camx.com)
19 of 21
D17301
ATHANASOPOULOU ET AL.: DUST INCORPORATION IN PM SIMULATIONS
ENVIRON (2004), User’s guide to the comprehensive air quality model
with extensions (CAMx), version 4.10s, report, Novato, Calif. (Available
at http://www.camx.com)
European Environment Agency (EEA) (2007), EMEP/CORINAIR Emission
Inventory Guidebook—2007, Copenhagen. (Available at http://www.eea.
europa.eu/publications/EMEPCORINAIR5)
Fécan, F., B. Marticorena, and G. Bergametti (1999), Parametrization of the
increase of the aeolian erosion threshold wind friction velocity due to soil
moisture for arid and semi‐arid areas, Ann. Geophys., 17(1), 149–157.
Fountoukis, C., and A. Nenes (2007), ISORROPIA II: A computationally
efficient thermodynamic equilibrium model for K+–Ca2+–Mg2+–NH+4 –
−
−
Na+–SO2−
4 –NO3 –Cl –H2O aerosols, Atmos. Chem. Phys., 7(17), 4639–
4659, doi:10.5194/acp-7-4639-2007.
Fuller, G. W., D. C. Carslaw, and H. W. Lodge (2002), An empirical
approach for the prediction of daily mean PM10 concentrations, Atmos.
Environ., 36(9), 1431–1441, doi:10.1016/S1352-2310(01)00580-5.
Gaydos, T. M., B. Koo, S. N. Pandis, and D. P. Chock (2003), Development and application of an efficient moving sectional approach for the
solution of the atmospheric aerosol condensation/evaporation equations,
Atmos. Environ., 37(23), 3303–3316, doi:10.1016/S1352-2310(03)
00267-X.
Gery, M. W., G. Z. Whitten, J. P. Killus, and M. C. Dodge (1989), A photochemical kinetics mechanism for urban and regional scale computer
modeling, J. Geophys. Res., 94, 12,925–12,956.
Gidhagen, L., C. Johansson, G. Omstedt, J. Langner, and G. Olivares
(2004), Model simulations of NOx and ultrafine particles close to a Swedish highway, Environ. Sci. Technol., 38(24), 6730–6740, doi:10.1021/
es0498134.
Gillette, D. A., and R. Passi (1988), Modeling dust emission caused by
wind erosion, J. Geophys. Res., 93(D11), 14,233–14,242, doi:10.1029/
JD093iD11p14233.
Gillies, J., V. Etyemezian, H. Kuhns, D. Nikolic, and D. Gillette (2005),
Effect of vehicle characteristics on unpaved road dust emissions, Atmos.
Environ., 39(13), 2341–2347, doi:10.1016/j.atmosenv.2004.05.064.
Ginoux, P., M. Chin, I. Tegen, J. M. Prospero, B. Holben, O. Dubovik, and
S. Lin (2001), Sources and distributions of dust aerosols simulated with
the GOCART model, J. Geophys. Res., 106(D17), 20,255–20,273,
doi:10.1029/2000JD000053.
Ginoux, P., J. M. Prospero, O. Torres, and M. Chin (2004), Long‐term simulation of global dust distribution with the GOCART model: Correlation
with North Atlantic Oscillation, Environ. Model. Softw., 19(2), 113–128,
doi:10.1016/S1364-8152(03)00114-2.
Greeley, R., and J. D. Iversen (1985), Wind as a Geological Process, Cambridge Planet. Sci. Ser., vol. 4, Cambridge Univ. Press, New York.
Grell, G. A., J. Dudhia, and D. Stauffer (1994), A description of the fifth‐
generation Penn state/NCAR mesoscale model (MM5), Tech. Note,
NCAR/TN‐398+STR, 138 pp., Natl. Cent. for Atmos. Sci., Boulder, Colo.
Grivas, G., A. Chaloulakou, C. Samara, and N. Spyrellis (2004), Spatial
and temporal variation of PM10 mass concentrations within the greater
area of Athens, Greece, Water Air Soil Pollut., 158(1), 357–371,
doi:10.1023/B:WATE.0000044859.84066.09.
Harrison, R. M., A. M. Jones, and R. G. Lawrence (2003), A pragmatic
mass closure model for airborne particulate matter at urban background
and roadside sites, Atmos. Environ., 37, 4927–4933, doi:10.1016/j.
atmosenv.2003.08.025.
Hellenic Ministry of Environment (2006), Report on the control of air pollution sources for the year 2005 (in Greek), Athens, Greece.
Hien, P., V. Bac, and N. Thinh (2005), Investigation of sulfate and nitrate
formation on mineral dust particles by receptor modeling, Atmos. Environ., 39(38), 7231–7239, doi:10.1016/j.atmosenv.2005.09.003.
Hodzic, A., B. Bessagnet, and R. Vautard (2006), A model evaluation of
coarse‐mode nitrate heterogeneous formation on dust particles, Atmos.
Environ., 40(22), 4158–4171, doi:10.1016/j.atmosenv.2006.02.015.
Jaecker‐Voirol, A., and P. Pelt (2000), PM10 emission inventory in Ile de
France for transport and industrial sources: PM10 re‐suspension, a key
factor for air quality, Environ. Model. Softw., 15(6–7), 575–581,
doi:10.1016/S1364-8152(00)00044-X.
Jalava, P. I., R. O. Salonen, A. S. Pennanen, M. S. Happo, P. Penttinen,
A. I. Hälinen, M. Sillanpää, R. Hillamo, and M. Hirvonen (2008), Effects
of solubility of urban air fine and coarse particles on cytotoxic and inflammatory responses in RAW 264.7 macrophage cell line, Toxicol. Appl.
Pharmacol., 229(2), 146–160, doi:10.1016/j.taap.2008.01.006.
Kallos, G., A. Papadopoulos, P. Katsafados, and S. Nickovic (2006), Transatlantic Saharan dust transport: Model simulation and results, J. Geophys.
Res., 111, D09204, doi:10.1029/2005JD006207.
Kallos, G., M. Astitha, P. Katsafados, and C. Spyrou (2007), Long‐range
transport of anthropogenically and naturally produced particulate matter
in the Mediterranean and North Atlantic: Current state of knowledge,
J. Appl. Meteorol. Climatol., 46, 1230–1251, doi:10.1175/JAM2530.1.
D17301
Karanasiou, A., P. Siskos, and K. Eleftheriadis (2009), Assessment of
source apportionment by Positive Matrix Factorization analysis on fine
and coarse urban aerosol size fractions, Atmos. Environ., 43(21),
3385–3395, doi:10.1016/j.atmosenv.2009.03.051.
Ketzel, M., G. Omstedt, C. Johansson, I. Düring, M. Pohjola, D. Oettl,
L. Gidhagen, P. Wåhlin, A. Lohmeyer, and R. Berkowicz (2007), Estimation and validation of PM2.5/PM10 exhaust and non‐exhaust emission factors for practical street pollution modelling, Atmos. Environ.,
41, 9370–9385.
Kondratyev, K. Y., L. S. Ivlev, V. F. Krapivin, and C. A. Varostos (2005),
Atmospheric Aerosol Properties: Formation, Processes and Impacts,
1st ed., 572 pp., Springer, Berlin.
Koo, B., T. M. Gaydos, and S. N. Pandis (2003), Evaluation of the equilibrium, hybrid, and dynamic aerosol modeling approaches, Aerosol Sci.
Technol., 37, 53–64, doi:10.1080/02786820300893.
Krueger, B., V. Grassian, J. Cowin, and A. Laskin (2004), Heterogeneous
chemistry of individual mineral dust particles from different dust source
regions: The importance of particle mineralogy, Atmos. Environ., 38(36),
6253–6261, doi:10.1016/j.atmosenv.2004.07.010.
Kühlwein, J., and R. Friedrich (2000), Uncertainties of modelling emissions from road transport, Atmos. Environ., 34(27), 4603–4610,
doi:10.1016/S1352-2310(00)00302-2.
Lee, Y. H., K. Chen, and P. J. Adams (2009), Development of a global
model of mineral dust aerosol microphysics, Atmos. Chem. Phys., 9,
2441–2458, doi:10.5194/acp-9-2441-2009.
Lewis, E. R., and S. E. Schwartz (2005), Sea Salt Aerosol Production:
Mechanisms, Methods, Measurements, and Models—A Critical Review,
Geophys. Monogr. Ser., vol. 152, AGU, Washington, D. C.
Liu, M., and D. Westphal (2001), A study of the sensitivity of simulated mineral dust production to model resolution, J. Geophys. Res.,
106(D16), 18,099–18,112, doi:10.1029/2000JD900711.
Lohmeyer, A., U. Lambrecht, Z. Peranic, and R. Boesinger (2002), Validation of vehicle road PM10 emission models by the Karlsruhe validation
data set and the results of the regular German state monitoring stations,
paper presented at the Eighth International Conference on Harmonisation
within Atmospheric Dispersion Modelling for Regulatory Purposes, Natl.
Inst. of Meteorol. and Hydrol., Sofia, Bulgaria.
Mansell, G. E., M. Wolf, J. Gillies, W. Barnard, and M. Omary (2004),
Final report: Determining fugitive dust emissions from wind, ENVIRON,
Novato, Calif.
Marticorena, B., and G. Bergametti (1995), Modeling the atmospheric dust
cycle: 1. Design of a soil‐derived dust emission scheme, J. Geophys.
Res., 100(D8), 16,415–16,430, doi:10.1029/95JD00690.
Marticorena, B., G. Bergametti, B. Aumont, Y. Callot, C. N’Doume, and
M. Legrad (1997), Factors controlling threshold friction velocity in semiarid and arid areas of the United States, J. Geophys. Res., 102(D4),
4387–4404, doi:10.1029/96JD02964.
Matsuki, A., et al. (2004), Mineralogy of tropospheric dust aerosols and its
influence on the heterogeneous sulphate formation, in Nucleation and
Atmospheric Aerosols 2004: 16th International Conference, edited by
M. Kasahara and M. Kulmala, pp. 653–656, Kyoto Univ. Press, Kyoto,
Japan.
Melas, D., I. Ziomas, O. Klemm, and C. S. Zerefos (1998), Anatomy of the
sea‐breeze circulation in Athens area under weak large‐scale ambient
winds, Atmos. Environ., 32(12), 2223–2237, doi:10.1016/S1352-2310
(97)00420-2.
Mitsakou, C., G. Kallos, N. Papantoniou, C. Spyrou, S. Solomos, M. Astitha,
and C. Housiadas (2008), Saharan dust levels in Greece and received inhalation doses, Atmos. Chem. Phys., 8(23), 7181–7192, doi:10.5194/acp-87181-2008.
Moosmüller, H., J. Gillies, C. Rogers, D. DuBois, J. Chow, J. Watson, and
R. Langston (1998), Particulate emission rates for unpaved shoulders
along a paved road, J. Air Waste Manage. Assoc., 48(5), 398–407.
Myhre, G., A. Grini, J. M. Haywood, F. Stordal, B. Chatenet, D. Tanré,
J. K. Sundet, and I. S. A. Isaksen (2003), Modeling the radiative impact
of mineral dust during the Saharan Dust Experiment (SHADE) campaign,
J. Geophys. Res., 108(D18), 8579, doi:10.1029/2002JD002566.
Nickovic, S., G. Kallos, A. Papadopoulos, and O. Kakaliagou (2001), A
model for prediction of desert dust cycle in the atmosphere, J. Geophys.
Res., 106(D16), 18,113–18,129, doi:10.1029/2000JD900794.
Nicolás, J., N. Galindo, E. Yubero, C. Pastor, R. Esclapez, and J. Crespo
(2009), Aerosol inorganic ions in a semiarid region on the southeastern
Spanish Mediterranean coast, Water Air Soil Pollut., 201(1–4), 149–159,
doi:10.1007/s11270-008-9934-2.
Noorallah, G. J. (1999), Pedosphere and Its Dynamics, Salman Prod.,
Sherwood Park, Alberta, Canada.
Omstedt, G., B. Bringfelt, and C. Johansson (2005), A model for vehicle‐
induced non‐tailpipe emissions of particles along Swedish roads, Atmos.
Environ., 39(33), 6088–6097, doi:10.1016/j.atmosenv.2005.06.037.
20 of 21
D17301
ATHANASOPOULOU ET AL.: DUST INCORPORATION IN PM SIMULATIONS
Ostro, B., L. Roth, B. Malig, and M. Marty (2009), The effects of fine particle components on respiratory hospital admissions in children, Environ.
Health Perspect., 117(3), 475–480.
Park, S., and H. In (2003), Parameterization of dust emission for the simulation of the yellow sand (Asian dust) event observed in March 2002 in
Korea, J. Geophys. Res., 108(D19), 4618, doi:10.1029/2003JD003484.
Pateraki, S., T. Maggos, J. Michopoulos, H. Flocas, D. Asimakopoulos,
and C. Vasilakos (2008), Ions species size distribution in particulate matter
associated with VOCs and meteorological conditions over an urban region,
Chemosphere, 72(3), 496–503, doi:10.1016/j.chemosphere.2008.02.061.
Patra, A., R. Colvile, S. Arnold, E. Bowen, D. Shallcross, D. Martin,
C. Price, J. Tate, H. ApSimon, and A. Robins (2008), On street observations of particulate matter movement and dispersion due to traffic
on an urban road, Atmos. Environ., 42(17), 3911–3926, doi:10.1016/
j.atmosenv.2006.10.070.
Perez, L., A. Tobias, X. Querol, N. Küzli, J. Pey, A. Alastuey, M. Viana,
N. Valero, M. González‐Cabré, and J. Sunyer (2008), Coarse particles
from Saharan dust and daily mortality, Epidemiology, 19(6), 800–807,
doi:10.1097/EDE.0b013e31818131cf.
Pérez, N., J. Pey, S. Castillo, M. Viana, A. Alastuey, and X. Querol (2008),
Interpretation of the variability of levels of regional background aerosols
in the Western Mediterranean, Sci. Total Environ., 407(1), 527–540,
doi:10.1016/j.scitotenv.2008.09.006.
Prosmitis, A. B., E. Diapouli, A. Chaloulakou, and N. Spyrellis (2003),
Organic and elemental carbon particulate continuous field measurement
in Athens urban area, paper presented at 8th International Conference
on Environmental Science and Technology, Global Network of Environ.
Sci. and Technol., Lemnos Island, Greece, 8–10 Sept.
Pulugurtha, S. S., and D. James (2006), Estimating windblown PM‐10
emissions from vacant urban land using GIS, J. Hazard. Mater., 132(1),
47–57, doi:10.1016/j.jhazmat.2005.11.089.
Rodríguez, S., X. Querol, A. Alastuey, M. Viana, M. Alarcón, E. Mantilla,
and C. R. Ruiz (2004), Comparative PM10–PM2.5 source contribution
study at rural, urban and industrial sites during PM episodes in eastern
Spain, Sci. Total Environ., 328(1–3), 95–113, doi:10.1016/S0048-9697
(03)00411-X.
Scheff, P. A., and C. Valiozis (1990), Characterization and source identification of respirable particulate matter in Athens, Greece, Atmos. Environ., 24A(1), 203–211.
Schwarze, P., J. Øvrevik, R. Hetland, R. Becher, F. Cassee, M. Låg,
M. Løvik, E. Dybing, and M. Refsnes (2007), Importance of size and
composition of particles for effects on cells in vitro, Inhal. Toxicol.,
19, suppl. 1, 17–22, doi:10.1080/08958370701490445.
Seinfeld, J. H., and S. N. Pandis (2006), Atmospheric Chemistry and Physics: From Air Pollution to Climate Change, 2nd ed., Wiley‐Interscience,
New York.
Shao, Y. (2001), A model for mineral dust emission, J. Geophys. Res.,
106(D17), 20,239–20,254, doi:10.1029/2001JD900171.
Shao, Y., G. H. McTainsh, J. F. Leys, and M. R. Raupach (1993), Efficiencies of sediment samplers for wind erosion measurement, Aust. J. Soil
Res., 31, 519–532, doi:10.1071/SR9930519.
Shaw, W. J., K. Jerry Allwine, B. G. Fritz, F. C. Rutz, J. P. Rishel, and
E. G. Chapman (2008), An evaluation of the wind erosion module in
DUSTRAN, Atmos. Environ., 42(8), 1907–1921, doi:10.1016/j.atmosenv.
2007.11.022.
Siskos, P. A., E. B. Bakeas, I. Lioli, V. N. Smirnioudi, and P. Koutrakis
(2001), Chemical characterization of PM2.5 aerosols in Athens‐Greece,
Environ. Technol., 22, 687–695, doi:10.1080/09593332208618241.
D17301
Strader, R., C. Gurciullo, S. N. Pandis, N. Kumar, and F. W. Lurmann
(1998), Development of gas‐phase chemistry, secondary organic aerosol,
and aqueous‐phase chemistry modules for PM modeling, Rep. STI‐
97510–1822‐FR, Sonoma Technol., Petaluma, Calif.
Takahashi, K., H. Minoura, and K. Sakamoto (2008), Examination of discrepancies between beta‐attenuation and gravimetric methods for the
monitoring of particulate matter, Atmos. Environ., 42(21), 5232–5240,
doi:10.1016/j.atmosenv.2008.02.057.
Tanré, D., J. Haywood, J. Pelon, J. F. Leon, B. Chatenet, P. Formenti,
P. Francis, P. Goloub, E. J. Highwood, and G. Myhre (2003), Measurement and modeling of the Saharan dust radiative impact: Overview of
the Saharan Dust Experiment (SHADE), J. Geophys. Res., 108(D18),
8574, doi:10.1029/2002JD003273.
Tegen, I., and I. Fung (1994), Modeling of mineral dust in the atmosphere:
Sources, transport, and optical thickness, J. Geophys. Res., 99(D11),
22,897–22,914, doi:10.1029/94JD01928.
Thorpe, A., and R. M. Harrison (2008), Sources and properties of non‐
exhaust particulate matter from road traffic: A review, Sci. Total Environ., 400(1–3), 270–282, doi:10.1016/j.scitotenv.2008.06.007.
Thorpe, A. J., R. M. Harrison, P. G. Boulter, and I. S. McCrae (2007),
Estimation of particle resuspension source strength on a major London
Road, Atmos. Environ., 41(37), 8007–8020, doi:10.1016/j.atmosenv.
2007.07.006.
U.S. Environmental Protection Agency (1995), User’s Guide for the Industrial Source Complex (ISC3) Dispersion Models, vol. II, Description of
Model Algorithms, Emissions, Monit., and Anal. Div., Off. of Air Qual.
Plann. and Stand., Research Triangle Park, N. C.
U.S. Environmental Protection Agency (2003), Miscellaneous sources, in
AP‐42, vol. 1, 5th ed., chap. 13, Research Triangle Park, N. C. (Available
at http://www.epa.gov/ttn/chief/ap42/ch13/final/c13s0202.pdf)
Vautard, R., B. Bessagnet, M. Chin, and L. Menut (2005), On the contribution of natural Aeolian sources to particulate matter concentrations
in Europe: Testing hypotheses with a modelling approach, Atmos. Environ., 39(18), 3291–3303, doi:10.1016/j.atmosenv.2005.01.051.
Watson, J. G., and J. C. Chow (2000), Reconciling urban fugitive dust
emissions inventory and ambient source contribution estimates: Summary of current knowledge and needed research, DRI Doc., 6110.4F,
Desert Res. Inst., Reno, Nev.
Westphal, D., O. Toon, and T. Carlson (1987), A two‐dimensional numerical investigation of the dynamics and microphysics of Saharan dust
storms, J. Geophys. Res., 92(D3), 3027–3049, doi:10.1029/
JD092iD03p03027.
White, B. R. (1979), Soil transport by winds on Mars, J. Geophys. Res., 84,
4643–4651, doi:10.1029/JB084iB09p04643.
Zender, C. S., H. Bian, and D. Newman (2003), Mineral Dust Entrainment
and Deposition (DEAD) model: Description and 1990s dust climatology,
J. Geophys. Res., 108(D14), 4416, doi:10.1029/2002JD002775.
E. Athanasopoulou, A. Dandou, and M. Tombrou, Department of
Applied Physics, National and Kapodistrian University of Athens,
GR‐15784 Athens, Greece. ([email protected])
K. Eleftheriadis and A. Karanasiou, Institute of Nuclear Technology and
Radiation Protection, National Center of Scientific Research “Demokritos,”
GR‐15310 Athens, Greece.
A. G. Russell, School of Civil and Environmental Engineering, Georgia
Institute of Technology, Atlanta, GA 30332, USA.
21 of 21