Beruflich Dokumente
Kultur Dokumente
ENVIRONMENT
PERGAMON Atmospheric Environment 35 Supplement No. 1 (2001) $99-S105
www.elsevier.com/locate/atmosenv
Abstract
This work examines the deposition of particulate matter from elevated sources in a neutrally stratified atmosphere.
Two current models have been compared with the available experimental data. The particle trajectory model (PTM)
shows very good agreement with index of agreement of 0.99, while the fugitive dust model (FDM) tends to over-predict
the value of the concentration peak and under-predict the concentration profiles at large downwind distances. The paper
also reports on the influence of particle shape on the dispersion and deposition of heavy particles. Simulation runs have
been performed for 75 and 100 gm particles of different sphericity by means of the PTM. The results indicate that the
shape of the particle is a very important parameter in determining the deposition curves, but that its importance
decreases with the size of the particle. © 2001 Elsevier Science Ltd. All rights reserved.
1352-2310/01/$-see fi'ont matter © 2001 Elsevier Science Ltd. All rights reserved.
PII: S1 3 5 2 - 2 3 1 0 ( 0 0 ) 0 0 5 3 4 - 3
$100 V. Vesovic et al. / Atmospheric Environment 35 Supplement No. 1 (2001) $99-S105
correlation functions with that of the surrounding fluid, ground-level concentration is then given as
nor is one fluid time scale sufficient to describe the
relative particle motion.
C - Q exp - exp ~/expl-
A number of models have been proposed to account rcayazU ~ [_uaZ. 2u a~] 1_
for the dispersion and deposition of the coarse particles
in the atmosphere. They can be generally divided into
x 1 - ~(Vd -- 0 . 5 v g ) - - exp [72] erfc(7) , (1)
two classes, namely, those that rely on suitable modifica- 0"zU
tions of the Gaussian plume model (Ermak, 1977;
Winges, 1992) and those that simulate a large number of ,/2(V d -- 0.5Vg)X h
particle trajectories (Ley, 1982; Wilson and Legg, 1983; = + --, (2)
Walkate, 1987; Hashem and Parkin, 1991; Sawford and
Guest, 1991). The work on modifications of the Gaussian where Q is the emission rate, u the wind speed, h the
plume model culminated in the development of the fugi- source height, x the downwind distance, and Vg the gravi-
tive dust model (FDM) (Winges, 1992) which is currently tational settling velocity. The standard deposition curves,
the EPA-recommended model for predicting the disper- as reported by Winges (1992), were used to evaluate
sion and the deposition of particulate matter. The work ay and az. This led to a mass conservation failure when
on particle trajectory models was more geared towards making use of Eq. (1). In order to correct for this incon-
gaining an understanding of particle fluid interaction and sistency, Winges (1992) modified Eq. (1) to predict the
no general-purpose model, analogous to FDM, exists. corrected ground-level concentration, C .....
It is important to stress that all of the currently avail-
able models assume that the released particles are spheri- c .... = Cf(x) (3)
cal. This is certainly not true of most of the emitted
particles and shape is believed to play an important role by introducing a distance-dependent correction factor, f
in particle transport away from the source. Recently, The correction factors were obtained by numerical integ-
some improvements have been proposed to the particle ration of the mass conservation equation at various wind
trajectory models which incorporate the shape of the speeds, atmospheric stability classes, release heights and
particle and avoid having to estimate the particle re- particle sizes. The data were fitted as a function of down-
sponse timescale. A result of these modifications was wind distance and incorporated into the computer
the development of a new two-dimensional Lagrangian model.
model for heavy, non-spherical particle dispersion
from elevated sources under neutral stability conditions 2.2. Particle trajectory model (PTM)
(Calviac and Yesovic, 1998; Dauriat and Vesovic,
1999). A two-dimensional mathematical model of dispersion
The present paper reports on the comparison of the of heavy particles in the atmosphere under neutral stabil-
predictions of the two models against available ex- ity conditions has been recently developed (Calviac and
perimental data and presents the simulation results Vesovic, 1998; Dauriat and Vesovic, 1999) based on the
that show the effect of particle shape on the deposition previous work (Ley, 1982; Wilson and Legg, 1983; Wal-
curves. kate, 1987; Hashem and Parkin, 1991; Sawford and
Guest, 1991) in the field. The extension to the three-
dimensional model is straightforward, but would involve
2. Mathematical models a substantial increase in computational time. It is be-
lieved that the crosswind dispersion of particulates is
2.1. Fugitive dust model (FDM) Gaussian and of relatively short range. Thus, no major
gain in physical insight is expected by extending the
The F D M (Winges, 1992) is based on the analytical model to 3-D, atleast not in this preliminary stage of
equations derived by Ermak (1977) for dispersion and investigation. Furthermore, due to the complexities of
deposition of airborne particulate matter from the point modelling the turbulent flows, only neutral stability con-
source. Ermak (1977) used as a starting point the K- ditions are considered. Although the extension to stable
theory advection-diffusion equation. He solved it under atmospheres should present no problem, the extension to
the steady-state conditions assuming that the turbulent cover the unstable stability class would require much
transport takes place in the crosswind direction and as more work. For brevity, only the main features of the
a function of height, and that advective transport is by PTM will be discussed, since a more detailed description
wind and gravitational sedimentation. Furthermore, has already been published (Calviac and Vesovic, 1998;
he assumed that the ground-level particle flux is pro- Dauriat and Vesovic, 1999).
portional to the local concentration, where the constant The particle trajectory of a single, non-spherical,
of proportionality is the deposition velocity, yd. The heavy particle in air is modelled by means of Newton's
V. Vesovic et al. / Atmospheric Environment 35 SupplementNo. 1 (2001) $99-S105 SlO1
dx~ At z) 1/2
- - = vi, (5)
dt
where v~ and u~ are the components of the particle and where r/ is the Gaussian random number and the esti-
wind velocity, respectively, and subscript i indicates mates of the Lagrangian time scale, TL, and the velocity
either x or z-component. Both the particle and the fluid standard deviations, cry and az, are given for neutral
motion are viewed relative to the stationary frame of stability conditions by (Hanna, 1982)
reference fixed in space in such a way that the negative 0.4z
z direction indicates the gravity direction and z = 0 TLx = TLz u* (9)
indicates the ground level. The parameter z is the
particle aerodynamic response time, the factor f is the ax = 2.4u*, a~ = 1.25u*. (10)
drag force correction factor given by a semi-empirical
expression As described previously (Calviac and Vesovic, 1998;
Dauriat and Vesovic, 1999), a set of four coupled first-
1 + O.1118[ReKxK2] °'6s67 order stiff differential equations, Eqs. (4) and (5), has been
f- solved by a forward difference method based on variable
K1
time step and originally proposed by Gear (1971). A com-
prehensive series of tests have been carried out by de-
+O.O1794ReK2 1 + ReK1K2 ' (6)
creasing the time step, to ensure that the solutions are
stable and well behaved. By decreasing the size of turbu-
where parameters K1 and K2 are Stokes' and Newton's
lent fluctuations it was demonstrated that the numerical
shape factors, respectively, which are related to the
solution obtained corresponds, in the limit, to the ana-
sphericity, qS, and projected area diameter, da, for a given
lytical solution for the wind profile given by Eq. (7).
non-spherical particle (Ganser, 1993). It measures devi-
Finally, an automatic adjustment of the time step
ations of the drag force experienced by the particle mov-
throughout the integration was implemented, thus lead-
ing relative to the fluid from the Stokes's law.
ing to a large reduction in computation time. The Gaus-
In order to solve a set of four coupled differential
sian random numbers, r/, were generated by means of the
equations, Eqs. (4) and (5), and thus compute the particle
minimal standard algorithms of Park and Miller (Press
trajectory from its release to its deposition, it is assumed
et al., 1992).
that, in principle, the wind components ux and uz are
No attempt has been made to model the particle tra-
known and thus calculable at any position and at any
jectories inside the roughness layer. For the average
instant in time. In this respect, the present model differs
release heights, 10-15m, and the average roughness
from most of the other previous particle trajectory mod-
height, 0.005-0.1 m, considered in this work, the amount
els (Ley, 1982; Wilson and Legg, 1983; Walkate, 1987;
of time that the particle spends inside the roughness layer
Hashem and Parkin, 1991; Sawford and Guest, 1991), in
is less than 1% of the total travel time. It is thus reason-
that it does not require an estimate of the particle time
able to assume, as a first approximation, that only grav-
scale. Thus, it uses a type of Markov chain process to
ity acts on the particle and that the particle does not
model the wind velocity, rather than particle velocity.
undergo any horizontal movement once inside the
This results in an increase in computational time, but
roughness layer. Taking into account that the velocity
avoids the complexities of having to estimate the particle
components inside the layer are on average small, the
time scale.
numerical error introduced in the overall landing posi-
As is customary (Zannetti, 1990) the instantaneous
tion of the particles is believed to be negligible in the
wind velocity u has been decomposed into two compo-
context of this work.
nents, namely the ensemble-average wind velocity,
fi(fix,0), given by
U* z -I- z o 3. Results
fix = - - In , (7)
1£ Zo
3.1. Comparison with field data
where u* is the friction velocity, Zo is the surface
roughness length, ~ is the yon Karman's constant Reliable experimental data on particle deposition, that
equal to 0.4, and the turbulent fluctuation u' (u', u/) can be used to validate and compare different models, are
given by (Hanna, 1982; Walkate, 1987; Hashem and rather scarce. The majority of the field data that exist on
S102 K Vesovic et al. / Atmospheric Environment 35 Supplement No. 1 (2001) $99-S105
0.010 Table l
Statistical measures of model performance
MA£ IA
0.008
PTM 0.00045 0.99
FDM 0.00138 0.92
i
¥ 0.006
0.016
¥
[a.
0.004 0.012
¢-
g o.oo8
"o
I19
0.002 *6
~5 0.004
13-
0.000
0.000 0.000 0.004 0.008 0.012 0.016
0 200 400 600 800
Experimental frequency, m -1
Downwind distance, m
Fig. 2. Deposition curve for the conditions corresponding to Fig. 4. Measured vs. predicted frequency based on the data
those of the original Trial 8: (average wind speed at the release generated in Trials 5, 8 and 9: (U] FDM; A PTM).
height, u~ s m = 9.1 m s 1, Zo = 0.005 m): (11 experimental data
points (Hage, 1961); . . . . FDM; - - PTM).
reported (Calviac and Vesovic, 1998) primarily due to the
use of a more accurate expression for the drag force,
0.016 Eq. (6), and due to improvements made in a numerical
technique used to solve the set of differential equations,
I' which in turn allows for a better resolution of the peak
0.014 [i concentration.
I I
iq
in order to examine the overall agreement between the
o.o, fl predicted and experimental data the use is made of two
statistical indices, namely, the mean absolute error
(NAg)
0.010
¥ 1 N
M A E = N~-~I IPi - OiL (11)
~, 0.008
g
u-
and the index of agreement (IA),
0.006
ZL~ (Pi o,F
-
Table 2
Statistical analysis of the deposition curves for different particle shapesa
afpea k -- normalised number frequency corresponding to the maximum in the deposition curve; Xp~ak - the landing position
corresponding to the maximum in the deposition curve;xm,ai,n - the landing position at which 50% of the released particles have landed;
x < 3 Xm~ai,n -- the percentage of particles that have landed by the distance corresponding to 3 Xm~alan X70O/o -- the landing position at
which 70% of the released particles have landed.
3.2. Influence oJ))article shape the peak concentration is very sensitive to a number of
parameters and thus has the highest uncertainty of all the
A series of parametric studies have been carried out to quantities tabulated in Table 2. Therefore, a number of
ascertain the influence Of particle shape on the deposition additional statistical measures were used to examine the
curves. All the simulations have been performed by run- influence of the shape of the particles.
ning PTM, since the FDM does not allow for the explicit The median of the deposition curves, as illustrated in
inclusion of the shape factor. No reliable experimental Table 2, increases for decreasing sphericity. The change is
data exist on release of non-spherical particles, so the more pronounced for the larger 100 gm particles where
validity of P T M for non-spherical particles rests on the the median for the disk deposition is twice the median
good agreement achieved with the data of Hage. For for the deposition of spherical particles. For smaller
the purposes of this paper the results are illustrated by particles the analogous change in the median is 60%.
simulations carried out on particles of 75 gm equivalent Similar conclusions can be drawn by examining the per-
volume diameter and nominal density of 2700 kg m-3. centage of the particles that have landed by a distance of
The use of the equivalent volume diameter for different three times the median. Again the influence of the particle
shape particles ensures that all the particles are of t h e shape is more prominent for the larger particles. Only at
same volume and therefore weight. In the example very large downwind distances from the source, does the
presented, all the particles were released from a height of shape of the 75 gm particles appreciably influence the
10m into average wind of 5 m s-1 over the flat terrain deposition rate. Thus, by 173 m, 70% of the spheres have
with the roughness length of Zo = 0.1m, corresponding landed, while the disks travel approximately twice as
to grassland. Simulations were performed for three differ- long and 70% of the disks would land by 357m.
ent shapes, namely sphere, tetrahedron (q5 = 0.67) and
disk (~b = 0.23; d,/dv = 0.4), thus covering a wide range of
sphericity. The results of the simulations, presented here, 4. Conclusions
supplement the investigations reported previously under
the same conditions, but for 100gm particles (Dauriat The ground-level deposition curves of the coarse par-
and Vesovic, 1999). ticles released from the elevated source have been
Table 2 summarises the main statistical characteristics simulated by the recently developed model of heavy
of the deposition curves for 75 and 100 ~tm particles, for particle dispersion (PTM) and FDM. The results have
the three different sphericities. The results indicate that been compared to the available experimental data. The
for 100 ~tm particles the peak concentration reduces dras- overall agreement of both models is very good as demon-
tically with decreasing sphericity resulting in approxim- strated by the high values of the index of agreement (IA)
ately a 65 % reduction between deposition of spheres and of 0.99 for PTM and 0.92 for FDM. The detailed analysis
disks. The position of the peak also changes, shifting indicates that the agreement between the PTM predic-
downwind from 52m for spheres to 70m for disks. For tions and the experimental data is very good everywhere
75 gm particles the effect is less pronounced. The compari- except near the concentration peak, while the F D M
son between spheres and the disks indicates that the peak tends to over-predict the value of the concentration peak
concentration drops off by about 40%, while the position and under-predict the concentration profiles at large
of the peak increases by 13 m. As discussed previously, downwind distances.
V. Vesovic et al. / Atmospheric Environment 35 Supplement No. 1 (2001) $99-S105 S105
Influence of the particle shape on the deposition of Gear, C.W., 1971. Numerical Initial Value Problems in Ordi-
particles was also examined by means of the P T M simu- nary Differential Equations, Prentice-Hall, Englewood
lation runs performed for 75 and 100gm particles of Cliffs, NJ.
different sphericity. The results indicate that the shape of Hage, K.D., 1961. On the dispersion of large particles from
the particle is a very important parameter in determining a 15-m source in the atmosphere. Journal of Meteorology 18,
534-539.
the deposition curves. Although, as the particle size de-
Hanna, S.R., 1982. Applications in air pollution modelling in
creases the differences in shape become less important in
atmospheric turbulence. In: Nieuwstadt, F.T., van Dop, H.
influencing the particle deposition. (Eds.), Air Pollution Modelling. D. Reidel, Dordrecht,
pp. 275-310.
Hashem, A., Parkin, C.S., 1991. A simplified heavy particle
References random-walk model for the prediction of drift from
agricultural sprays. Atmospheric Environment 25A (8),
Calviac, G., Vesovic, V., 1998. Modelling dispersion of partic- 1609-1614.
ulate matter in the mining environment. In: Pasameh- Ley, A., 1982. A random walk simulation of two-dimensional
metoglu, A.G., Ozgenoglu, A. (Eds.), Proceedings of the Fifth turbulent diffusion in the neutral surface layer. Atmospheric
International Symposium on Environmental Issues and Environment 16 (12), 2799-2808.
Waste Management in Energy and Mineral Production Press, W.H., Teukolsky, S.A., Vetterling, W.T., Flannery, B.P.,
- SWEMP '98. Balkema Publishers, Brookfield, pp. 143-148. 1992. Numerical Recipes; the Art of Scientific Computing.
Dauriat, A., Vesovic, V., 1999. Modelling dispersion of heavy Cambridge University Press, Cambridge, pp. 266-280.
particulate matter. In: Brebbia, C.A., Jacobson, M., Power, Sawford, B.L., Guest, F.M., 1991. Lagrangian statistical simula-
H. (Eds.), Air Pollution VIII. WIT Press, Southampton, tion of the turbulent motion of heavy particles. Boundary-
pp. 843-852. Layer Meteorology 54, 147-166.
Doran, J.C., Abbey, O.B., Buck, J.W., Glover, D.W., Horst, T.W, Walkate, P.J., 1987. A random-walk model for dispersion of
Lee, R.N., Lloyd, F.D., 1984. Field Validation of Exposure heavy particles in turbulent air flow. Boundary-Layer
Assessment Models, Vol. 1: Data. EPA/600/3-84/0924, Meteorology 39, 175-190 and references therein.
NTIS, Springfield. Wilson, J.D., Legg, B.J., 1983. Calculation of particle trajectories
Doran, J.C., Horst, T.W., 1985. An evaluation of Gaussian in the presence of a gradient in turbulent-velocity variance.
plume depletion models with dual-tracer field measure- Boundary-Layer Meteorology 27, 163-169 and references
ments. Atmospheric Environment 19, 939-951. therein.
Ermak, D., 1977. An analytical model for air pollutant transport Winges, K.D., 1992. User's Guide for the Fugitive Dust Model.
and deposition from a point source. Atmospheric Environ- Environmental Protection Agency Report, EPA-910/
ment 11,231-237. 9-88-202R.
Ganser, G.H., 1993. A rational approach to drag prediction of Zannetti, P., 1990. Air Pollution Modelling; Theories, Computa-
spherical and non-spherical particles. Powder Technology tional Modelling and Available Software. Computational
77, 143-152. Mechanics Publications, Southampton.