See

discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/226990494

A mathematical model of suspension with

saltation

Article in Acta Mechanica January 1995

DOI: 10.1007/BF01177324

CITATIONS READS

17 7

3 authors, including:

Bill Scott
Murdoch University
111 PUBLICATIONS 614 CITATIONS

SEE PROFILE

All content following this page was uploaded by Bill Scott on 27 March 2016.

The user has requested enhancement of the downloaded file. All in-text references underlined in blue are added to the original document
and are linked to publications on ResearchGate, letting you access and read them immediately.
 ACTA MECHANICA
Acta Mechanica 108, 1 - 2 2 (1995)
9 Springer-Verlag 1995

Invited Review Article

A mathematical model of suspension with saltation
W. D. Scott, Murdoch, J. M. Hopwood, Nedlands, and K. J. Summers, Applecross,
Western Australia

(Received December 15, 1992)

Summary.A theoretical approach to the treatment of wind erosion data, particularly from a wind tunnel, is
presented. Considerations are given to the utilisation of a real data set in validation of the model, data that
will be presented in a forthcoming paper, Following this, the physics of particle suspension, saltation and the
turbulent boundary layer are examined. Two different mathematical models evolve: one considers only
suspension, another evokes Bagnold's observation that eroding material merely shifts the velocity profile and
the effect of the airborne material on the effective density of the air parcel. These produce a final, relatively
simple expression that credibly fits the data of Gerety and Slingerland. A critique of the approach reveals it to
be an adequate expression of the known mechanisms of suspension and saltation. Derived algebraic forms for
integrated collectors show several of the same "logarithmic power" dependences. Importantly, the results
show little influence of saltation itself on the profile. It appears that the saltation process is responsible for
a feedback such that the eddy diffusion process for particle movement is effectively enhanced. The
combination of an appropriate correction of the pitot data (following Scott and Carter) and a complete mass
balance has removed the "kink" from the velocity profile and also the need to consider the saltation process
itself in the particle mass balance.

1 Introduction

The classic work of Bagnold [1] carries through to this day as the most definitive work on wind
erosion. His perceptive insights have formed the basis for "the wind erosion equation" [2], [3].
This is an empirical equation based on site-specific statistical data; soil loss is a function of the
non-erodible fraction, the surface roughness, the local climate, the upwind fetch and the
vegetative cover. Bagnold's initial studies of saltation and the lift-off of single particles has
evolved into some complex analytical and computer simulations [ 4 ] - [7]. In the field his detailed
analyses of both wind and source have not really been repeated; wind tunnel studies have been
able to corroborate several of his findings [8]-[11]. The reader is referred to Bagnold's b o o k or
the b o o k of Pye [12] for a recent view.
Also a number of workshops on wind erosion have been conducted. Noteworthy are the
International Workshop on the Physics of Blown Sand held at Aarhus University [13], the
Erosion Wind Tunnel Workshop held at M u r d o c h University [14] and the International
Workshop on Aeolian Grain Transport [15]. All have summaries; the thrusts of the last two seem
quite disparate. The Erosion Wind Tunnel Workshop has put forth a list of criteria for
appropriate measurement and an effort was made to produce a plan of action to improve our
instrumentation, particularly trap design and the measurement of turbulent shear stress in this
harsh environment [16]. The later international workshop was dominated by studies of particle
2 W.D. Scott et al.

trajectories in uniform beds of saltating grains. In the introductory summary article Anderson
and Willetts [17] state that the "considerable progress in mathematical models of aeolian sand
transport has not yet led to more reliable predictions of transport rates. It is, moreover,
questionable whether this approach will ever yield universal transport rate formulae".
Considering the limitations of identical trajectory, single particle analyses, it is not surprising
that Bagnold's work continues to stand out. The statistical requirements are not met with
two-dimensional beds that are 30 or so particles wide with particles cycling in and out of the bed.
Most of the studies rely on a source limitation with a 'splash function' that specifies a distribution
of ejected particles, their velocities and angles of exit from a given collision with the surface; the
calculations are completed only with simplest forms for the functions. If multiple sizes and highly
variable trajectories were included, the calculation would not be possible even on recent, fast
computers. The results of these simplest analyses create a maximum in the concentration, mass
flux and kinetic energy fluxes at the top of the particle path [7], [18], conditions that are not
observed. They cannot allow for the winnowing of particles within the fluid mixture, formation of
dunes or surface armouring. Still, the single particle models can simulate the timewise evolution
of the erosion process and show the difference between fluid threshold and impact threshold.
It is clear that there is a Froude 1 number limitation of about 10 on wind tunnel work [11].
Butterfield [19] and Carminati et al. [20] have found that it is not really possible to obtain
a shear stress from the steady velocity profile; it is suggested that extreme caution be used if
transport rates obtained from wind tunnel data are to be used in the field. Some of the
theoretical papers refute Bagnold's observation of a focal point in the logarithmic profile [7],
[21]; they do, however, show the same logarithmic slope with and without erosion. The
experimental work of White and Mounla [11] shows a focal point; they emphasise that it is
only obtained with equilibrium flows. The experimental data of Rasmussen and Mikkelsen
[22] exhibits the focus but it is not recognised in favour of the Owen formula. Most of the
theoretical and experimental studies acknowledge that it takes 1 - 2 seconds for the saltating
process to reach an equilibrium state. Anderson and Haft [7] suggest that there is no real
definition of layers but a continuous merging of an upper 'suspension layer' with a lower
'saltation layer'. The articles seem to have ignored the gross density increase that occurs close
to the bed; it appears as a perceived deviation in a nearly logarithmic wind profile that is
recognised as a 'saltation layer' [23].
Both Bagnold and Owen died in 1990 and the last international workshop was dedicated to
their memories. The present paper builds on their work. Importantly, there are two of Bagnold's
basic perceptions that we consider in this paper: these are his observation of logarithmic profile
both with and without erosion and the observation of a 'kink' in the profile that relates to
saltation.

2 Backbone

Dust movement originates with seed grains that plummet to earth and dislodge other grains (see
Fig. 1). This results in a cascade of temporarily suspended material in 'saltation'. Saltating grains
vibrate the surface grains and cause 'surface creep' when surface adhesion is sufficiently small;
this effect commonly moves about 25 % of the material. Importantly, the saltating grains knock
smaller grains off the surface; these smaller particles have relatively small settling velocities and

The Froude number is Ux2/gHwhere U 1 is the mean velocity upstream of the bed, g is the acceleration
of gravity, and H is the tunnel height.
A mathematical model of suspension with saltation 3

become dust clouds that can penetrate to great heights, forming an aerosol or near aerosol of
particles in 'suspension'.
Of course, all these movements derive their energy from the wind or, rather, turbulent eddies
that wiggle and occasionally dislodge saltable grains ( ~ 100 gin) which emerge into the
airstream, acquire the momentum of the airstream and later plummet back to release their energy
to other, perhaps smaller, particles and may cause larger grains (or clods) to fragment into finer
particles [24]. As shown, the saltating particles are effectively shot vertically into the airstream
and fall back to collide with the surface at an angle of 10~ to 16~,

Suspension

Saltation
Surface creep Fig. 1. Schematic view of particle trajectories

The wind structure is affected by the overriding wind, surface roughness and air stability. The
dramatic mechanical mixing in wind-driven dust is expected to overcome any non-neutral
tendency of the air so the primary parameters that determine the dust movement are surface
roughness zo and drag velocity u,, defined by the logarithmic equation

u,
u(z) = ~- in , (1)

where k is von Karman's constant (k = 0.40). This equation is of general validity and fits the
profile of mean horizontal wind speed near the ground in various conditions [25] but most
accurately describes the profile in conditions of stationary 'frozen' turbulence, in flat conditions
with sufficient fetch and neutral stability. Importantly, it describes the wind profile in both
eroding and non-eroding conditions.
In eroding conditions, there has been some contention regarding the values and the number
of parameters necessary to describe the profile. It has been shown conclusively that, when the
density of the driving sand is accounted for, the logarithmic profile Eq. (1) extends down into the
saltating layer or rather to a distance of one surface roughness length from the real surface [23].
The saltation 'kink' in the profile is absent.
Bagnold [1] has related the parameters in an eroding condition to those in a non-eroding
condition. In eroding conditions Eq. (1) becomes modified so that u(z) becomes the velocity
in excess of the impact threshold velocity u~; u, is little affected by t h e presence of
erosion and is determined by the shear stress of the overriding wind but zo is dramatically
increased to a value Zo,.
In our data set we propose that wind profile data are available for a large range of imposed
wind speed conditions. The procedure for fitting is as follows:

(i) Fit a linear-log regression line to the data set. Get u, and Zo.
(ii) Test the values of u, and Zo for differences statistically. Divide into eroding and non-
eroding fractions. Scott [26] outlines a simple technique for sorting the type of erosion based on
the way u, and Zo' relate. Note that, since the eroding surfaces are evolving with time, it is possible
that several different, non-sequential eroding conditions may occur.
(iii) Obtain zo and the u, value required for particle lift-off (u,t).
4 W.D. Scott et al.

3 Saltation

Saltation is initiated at the impact threshold velocity u,t, which is physically related to particle
size by the expression [1]

Prediction of this threshold velocity follows from the size distribution of the saltating
particles or the surface particle size distribution [27], [28]. In the case of a distribution of grain
sizes, the size that is expected to describe the onset of wind erosion is

\ nt /

where n~ is the fractional number of particles of size di. Note that this expression is a natural
consequence of the mathematical separation that occurs and is formalised in eq. [25], see also [29].
The overall picture is therefore:
(i) u* and Co' describe the forcing influence, the wind.
(ii) d (through u.t) determines the susceptibility of the particles for lift off.
(iii) In the eroding condition saltating surface particles use the kinetic energy of the
horizontal airstream to inject particles into the whole profile.
(iv) The concentration of particles of a given size at a given height is determined by the
competing influences of turbulent transport enhanced by saltation and particle fallout.
In what follows only the case of wind-driven particles will be considered. A model of the
suspension-saltation process will evolve.

3.1 The flux balance equation

In the unsteady state, the mass densityfof particles of effective diameter D is given by the partial
differential equation

~f ~[K~J] ? ~(vtf) OS
& - ~z ~z - ~z ( v f ) + ~z Oz' (4)

wheref6D is the mass of particles contained in effective size interval D to D + 5D per m 3 of air.
The terms on the right are, respectively, the divergence of the eddy diffusion flux, the divergence
of the horizontal advective flux, the divergence of the sedimentation flux and a saltation term.
The velocity v is the longitudinal (x direction) velocity which may be a function ofx and height z.
Hindered settling is ignored so that the terminal velocity vt is a function of particle size only and
K is the turbulent eddy diffusivity which may be a function of x and z. The equation assumes
homogeneity in the lateral (y) direction and that the eddy flux along the flow direction (x) is
negligible compared to the advective flux. Also implicit in this formulation is that diffusion and
settling are independent and additive mechanisms.
In the initial derivations, the eddy diffusivity is considered to be independent of the particle
size. Also, it is considered that the particles are approximately moving with the horizontal
velocity of the air and the saltating layer is presumed to exist within a constant stress layer. The
effect of these rather gross assumptions will be assessed later (see Section 4).
A mathematical model of suspension with saltation 5

Further, we will assume that the boundary layer has developed as a steady state with
homogeneity along the longitudinal flow direction, assumptions which improve tractability and
are justified in flows with sufficient fetch. In the present the saltation term will also be ignored but
will be considered later; as a result the equation is expected to have limited applicability. With
these assumptions Eq. (4) becomes

dz & - v, & (5)

or, simply the flux balance

df
K dz = - vtf (6)

with the gradient and the concentrations vanishing at large z.

Integrating with K = ku,z,
vt

f = fo \~o',] = fo ~ = fo , (7)

Z i)t
where X = ~ and 7 = - - are dimensionless ratios that characterise the height and the relative
Zo ku,
importance of the fall velocity. The parameterfo is the mass density of particles of effective size
D at zo', near the surface; it is not necessarily the same as the surface size distribution; it is
dependent on particle size. This equation was derived by Budd [30] to treat the problem of
blowing snow. It is noteworthy that Budd summed this equation to care for the whole size
distribution and its change with height. The results he obtained were a credible, realistic fit to his
real data despite the fact that rounded snow particles of millimetre size were present. Allan [31]
also presents the same equation; it has been used variously to describe suspension of particulates
and snow [32]. It is a simplest expression that is valid for suspended particles, the "Rouse"
equation of sedimentology.
The relative decrease of airborne material with height as predicted by Eq. (7) is presented in
Figs. 2a and 2b. Note that the vertical scale is equivalent to 50 cm of height with a roughness
length of 1 ram. In this case, when the particles are of small size or the overriding wind is large
(y = .01) there is little change in concentration with height. If, however, 7 = 2 (e.g., the friction
velocity is 50 cm/sec and the fall velocity is 40cm/sec), only 0.1% of the ground level
concentration is found at X = 30 (e.g., at 3 cm above the ground when zo' = 1 ram).
In the experimental arrangement, however, the mass of particles in a given aerodynamic size
interval is collected over height interval zl and z2. We use the standard presumption here, namely
that the particles are moving with the air, to derive the total amount of material collected (see
Section 4). The mass collected per second in small size interval D to D + 5D is
z2 X2

R= f f uwdz - f~176 f X - ' ln XdX - k ~, (1, 2), (a)

zl X1
where [331
X2

I,(1,2)= f X-'lnXdX= [ XI-' XI-' ~ X2

L 1 - 7 in X (1 Z ))2ix 1"
X1
6 W.D. Scott et al.

X
O O O O
O ~ O r

2 2 ~'-- 10i ' . . . . . . . . . i ......... , ......... , .........

-- ~/= .o5 ~ _ ~

-?=.1

o~
" .__.__.-- "/= 2 ~
o

><
(:) O O (:D
CD
Lr r r

. ...... ' J ' .......... ' ...... '71 I

I Fig. 2. The relative decrease in particle mass

density as a function of relative height. Small
particles correspond to small values of y, the
relative fall velocity, a y between .01 and 1,
b 7 between .4 and 2

The same expression was first derived by Gillette and Goodwin [34]. It is the mass collected with
effective, aerodynamic particle size D with a collector of width w. An elutriator [35] m a y be used
to aerodynamically sort the collected material though it undoubtedly tends to break up the
grains [36]. Note that the integral, defined as the function I~(1, 2), contains the limits X1 and X2,
which are the dimensionless counterparts of zl and z2, the collector height interval. The
integrated form becomes infinite when ~ = 1; in fact it has a limit. This awkwardness is removed if
the integral is evaluated numerically.
If a collector extends from the surface (X1 = 1) to a chosen height (X2 = X), then the total
amount of material collected is given by
X

R(X) - fowu,zo'k f X-~ In XdX - fowu,zo'k I~(X). (9)

t

This integral function I~(X) is presented in Fig. 3. Note that if7 < 1 an infinite-sized collector
is necessary to collect all the airborne material, whereas ify > 1, nearly everything is collected in
a collector of dimensionless height 500 (e.g.i if Zo' = 1 m m then a collector of 50 cm height is
sufficient).
If, indeed, it is intended to collect material over the entire profile, an appropriate limiting form
obtains with total horizontal flow of material given by

fowu, zo' (10)

RT -- k (1 -- 7) 2,
A mathematical model of suspension with saltation 7

Iy(X)
0 50 lOO 500

/
3O0

~.. ~ 200 X

,-~"

.~~. z I00
Fig. 3. The suspended mass integral Iy(X), for
the total amount collected between relative
height 1 and relative height X. The relative fall
velocity ?, varies between .i and 1.5

which is valid only if ? > 1. If 7 < 1, particles are present to great heights and the upper limit of
the integral becomes infinite. In the extreme of very large particles when 7 ~> 1,

fowku,3zo
R r (large 7) - vt2 (11)

This form m a y dominate in windblown sand and produce the well-known empirical equation of
Bagnold [1] which relates the mass flow of saltating material to the cube of the friction velocity u,.
This expression (11) may show the dependence of Bagnold's constant on roughness length and
fall velocity.
Large 7, however, means large particles and it is unlikely that a simple flux balance which
ignores the effect of saltation is valid in this case; the next section attempts to quantify this effect
on the profile.

3.2 The sahation term

All the above equations presume that the mass per unit volume of air containing the driving
grains is constant in the vertical. This means that either there must be such a small number of
suspended particles in the air that the density is the same as the air density or that the equations
are only applicable over a limited range in the vertical. This must restrict the validity of the results
since it is well known that saltation causes gross density changes and a dramatic uplift of the
velocity profile with a considerable increase in the roughness length. The friction velocity, u,,
however, is not appreciably changed as it is dependent upon the wind patterns far removed from
the surface. A reasonable way the friction velocity can remain constant during the onset of wind
erosion is for the density increase in an air parcel (with associated m o m e n t u m increase) to be just
absorbed by the m o m e n t u m extraction due to injected, saltating particles. If ~o is the overall
density of an air parcel

o = ~. + c , (12)
8 W.D. Scott et al.

where 0a is the density of clean air and C is the mass of particulates per unit volume. With
saltation drag, the total shear stress is given by a turbulent Reynold's stress

'
"r,=QUW ', (13)

where u' and w' are fluctuating components of the horizontal and vertical wind speeds,
respectively. Similarly, in the case without wind erosion,
"CO = Oa U r N ' ~ O a U , 2 (14)

and, generally,

e u'w' = O,,u'w' + C u'w' (15)

or

= ro + ~ , (16)

so that

C
r~ = - - Zo (17)

or shear stress changes are directly proportional to density and one expects an increased shear
stress due to the extra particles in the air.
This does not happen because of the negative feedback stress due to injected particles with
zero horizontal momentum. If

G~ = mass of particles added per sec per unit area due to saltation injection

then

c~Gs
- gain of saltating particles per sec per unit volume
0z
and the m o m e n t u m lost by the air per sec is:

3G~
--U --
Oz
The force per unit volume due to saltation is the divergence of %; therefore, from Newton's
law

&~ - u 0G~ (18)

& & 9

This means that, from (17)

OG~&=(u_~)OC~z. (19)

This is an overall particle balance, including the mass of particles of all sizes as a single mass.
In fact, Eq. (4) is for particles of a given size and,
Dmax
C,s = sdo
Drain (20)
Dmax
C= ~ fdD
Dmin
A mathematical model of suspension with saltation 9

so, if we integrate (4) between the limits Dm~nand D .... we can use the integrated expressions (19
and 20). In steady-state with no divergence of the horizontal flux, the integrated form is

Dmax Dmax Dmax

8z 8z dV + ~z (v,f) dV - a~. dD = O. (21)

Drain Drain Drain

Note that a general indefinite integration would produce either a constant or a function of
z on the right side. Since the limits are given and finite, the right side must be zero. Allow that the
order of integration/differentiation be changed; this is possible for well-behaved functions. Then,
from Eqs. (20) and Eq. (19)

Drnax

~z dD = 0--7- = ~z (22)
Dmin

U 2
but ~* does not depend on D and C can, itself, be written as an integral form (20). This integral
u
form can, again, be interchanged with the partial derivative with respect to z. This means that

Dmax Dmax

V dD = - -

u ~z
dD (23)
Drain Drain

and the full Eq. (21) becomes

Dmax /)max Dmax

fO[g(~f
~z 8z I dD + f ~87z (vtf) dD - fu*2~fdO=
~ - 8z 0 (24)
Drain Drain Drain

Looking at Eqs. (4), (21) and (23), it is clear that the differential form of Eq. (4) and the finite,
definite integrals require that the differential form carry through to each size class.'- That is,
general continuity requires that

3z ~z + ~z (vtf) - u*z = O. (25)

u 3z

Note that this is a general form, without restrictions, other than the steady-state requirement.
z
Writing this equation in terms of X = - - and the Rouse number 7,
Z0 r

O/xT~f~ ~ 1 Of 0. (26)
+ g 2 (Tf) ln X a x

Substitute

= -71nX

2 The intent here is clear: the limits cannot be too rigid and still have a steady condition. Note that
Eq. (25) obtains if Eq. (17) is applied to infinitesimals.
10 W.D. Scott et al.

Eq. (26) becomes

042 ~ = 0, (27)

which has the solution

of
a~ - fo~d,

f = fo[er -- {e~] + go. (28)

Back substituting;

f =foX-~[7 In X + 1] + go, (29)

wherefo and go are functions of the diameter D. When X is large,f should go to zero if the particles
are not too small (suspended). In any case, if there is no vertical flux, the derivative should be zero.
This means that go ~ 0, so

f =foX-~[ylnX + 1]. (30)

Expression (30) rivals Eq. (7) for simplicity yet it includes saltation. Note that the second term
on the right side produces the equivalent of Eq. (7). The first term is from the saltation effect; its
functional form is exhibited in Figs. 4 a and 4 b. Note the tendency for little change with height
when y is small, trends exhibited by Eq. (7). At small heights, however, the function has a very
sharp 'knee' when ~ is about 1. This allows a definition of saltation and suspension while one
process melds into the other. We would define saltation particles as particles with effective Rouse
numbers greater than 1. This leaves the majority of saltation close on the ground. But the form of
Eq. (30) suggests little real distinction between suspension and saltation. Larger particles actually
71nX
penetrate far into the vertical. The ratio ~ i s very large at most heights. I f X = 10, and 7 = 2,
the ratio is 4.6.
The form of Eq. (29) was fitted to the data of Gerety and Slingerland [10]. Following
Hopwood and Scott [29] the data for all three minerals were organised in a distribution of fall
velocity (see Fig. 5). This is the fall velocity distribution for the source used in all four of their
experiments. The present form Eq. (29) has the characteristics of a gamma function and also
the fall velocity distribution has the characteristics of a gamma distribution. The best fit is
shown by the dotted line on Fig. 5 (a = 5.48, b = 0.0202), the technique is presented in the
Appendix.
Proceeding, we assume the distribution of the source, in the air, to also be a gamma func-
tion. Gamma functions were used effectively by Budd [30] to characterise the distribution of
rounded grains of blowing snow in saltation; Sorensen [37] also assumed a gamma distribution
in his theory of sand grains in saltation. The functional form Eq. (29) preserves the gamma distri-
bution

Aba (31)
f0 = r ~ ~~
A mathematical model of suspension with saltation 11

X-'Y I n X
0 . 1 .2 .3 .4 .5 .6 .7 .8 .9 1.0 1.2 1.4 1.6 1.8
500
-,,--r-,,,,,1T " v - ' r ' ~ r ,'- 1 - ~ -'v-"l"m"T"~r- ' ~ ' v W , ' , , ~ t

400

300

200

100

X-~ln X
.02 .04 .06 o 0B . 1O . 12: . 14 . ~6 . 16
500

qO0

Lllll
\\.... 300

200
Fig. 4. Salation function relation in
terms of the relative mass density of
IOO all particles at different relative
heights X and overall relative fall
velocities ~: a between .1 and 1,
b b between .6 and 5

o L Composited Distribution
~ | fron Geret9 & Slingerland
~I (c. 9. s. uni is)

E_~I n

ClLC~ -
0~ C~ a

3 ~. .. .,
~ o 9 .
Fig. 5, The distribution of source materials used by Gerety
I and Slingerland. Data are composited from the Quartz,
~oo 2oo 300 4oo 500 soo 700 Olivine and Garnet Distributions and presented with the
Fall Velocit9 argument of fall velocity
12 W.D. Scott et al.

where, for convenience, ~ is used as a surrogate for the fall velocity, in a given experiment with
known u.. Then, at any height

A In Xbayae-(b+t'x)v Ab"7"-le-(b+l"x)v
f = + +go.
r(a) F(a)
Define
~
b
g(a, b) = F ~ 7"-le-bv,

then

f=A --g(a+l,B)+g(a,B) +go, (32)

where B = b + In X and go is an arbitrary function of particle size; effectively, a distribution

function independent of height.
Integrate with respect to y from 0 ~ oe; the two g functions are normalised and
become equal to 1:

C=A
I J~ aln
b+lnX b+ln~ +1
1 +Co (33)

and, now, C is the integral o f f over all values of 7 and Co is a constant independent of size and
height.
The fitting procedure uses a trial and error guess of a and b and a least squares, analytical
solution for A. The fit is presented in Table 1 and Fig. 6. It is surprisingly good despite the fact
that Eq. (33) shows little of the normal power law or direct log-log dependence. There is a mild
bend in the curves that appears in the data. The ratio a/b is a measure of the mean size, in ~ units.
All the values are well above 1, suggesting an importance of saltation. The measure of the
standard deviation is ]/a/b (not shown in the table); it is low ( ~ .1) at the lower wind speeds
(EXP 1 - EXP 3) but high ( ~ .9) in EXP 4. Compared with the source distribution (Fig. 5, note
that the units are cm/sec), most of the material in motion is of smaller size than the modal value of
the source destribution.
Further, the two terms in Eq.(33) care for saltation and suspension, respectively;
Table 2 shows their contribution to the relative error 3. There is little gain in including the

Table 1. Fitting of Eq. (32), Gerety and Slingerland data (see Scott and Carter [23])
Zo Ustar A Co Vertical flux* Relative
(ram) (cm/sec) (kg/m 3) a a/b (kg/m 3) (kg/m 2) error

EXP 1 0.05 28. 379. 470. 2.16 0.00083 -.0002 3.5%

EXP 2 0.30 46. 73. 110. 2.61 0.0095 -.0046 ,16.0%
EXP 3 1.40 76. 17. 331. 2.76 0.0060 -.0050 6.2%
EXP 4 0.26 113. 5556. 21. 3.98 0.0031 --.0059 11.7%

* This value is presented for comparison. It is derived from the unjustified assumption that the distribution
of particles is equal to the distribution at ground level (z = zo). Then fro + J/t~ = Co(a + 1) and the vertical

flux is -ku. Co(ab+ 1), a mean sedimentation flux

3 With terms missing, the parameters are not so realistic.

A mathematical model of suspension with saltation 13

saltation term; if it has an influence, it is at the lower velocities with little spread in the
distribution. It appears saltation is more important with monodisperse particles, close to the
surface.
As before, we also consider the effect of an integrating collector. Eq. (32) is not directly
integrable but the distribution function Eq. (30) gives:

W~2-
owzou.[
k
x2
7 ~ X-r(lnX)ZdX+ ~ X-~lnXdX
l -owzo. [7K,(1,2)+I,(1,2)].
X1 X1 ]s
(34)

Where the integral Kr(1, 2) has a tractable solution [33]

X2

K,(1,2)= X - ' ( l n X ) 2dX [1-7 (lnX) 2 1-7 (1-7) 2 x, (35)

X1

and, as before, the integral Kr(X) can be defined where X1 = 1, X2 = X. The result of this
integration is depicted in Fig. 7. These data should be directly comparable to the integral data of
Fig. 3, derived for the simple flux-balance solution. Overall, the level of airborne mass is higher
for a given value of 7, though the trends are similar.

Table 2. The relative fits with suspension or saltation

No suspension No saltation Complete
solution

EXP 1 3.6% 3.4% 3.5%

EXP 2 16% 18% 16%
EXP 3 5.8% 7% 6.2%
EXP 4 11.8% 11.8% 11.7%

Data o f Geret9 & 511ngerland

ExperiMent I O
Experinent 2 A
ExperiMent 3 +
x
6

~4
x

-7 -6 -5 -4 -3 -2 -1 0 1 2 3 Fig. 6. Fitting of the data of Gerety and Slingerland

InC (see Scott and Carter, 1984)
14 W.D. Scott et al.

G(X)
0 ...............r , l lIir'SI IQI'"III
..... lOO

j
z' / - zoo

Fig. 7. Collected material integral Kv(X) for the

total amount collected to relative height X (see Eq.
(35)) Integral cares for both saltation and the
effectiveair density. The overall relative fall veloci-
ty 7 varies between 0.5 and 2

Noteworthy are the observations of d e Ploey [38] that the cumulative weight of trapped
sediment is related to the logarithm of the height by a predominantly parabolic function which is
of the form a(ln X) z + b In X, where a and b are arbitrary constants. These data were from
collections in the "panne" of a crescent dune in Belgium. Nearly all collections of windblown
sand show logarithmic trends, including the data of Bagnold [1], Zingg [8] and Williams [9].
Nickling [39] has fitted his data with a power function similar to Eq. (7). In this respect, consider
Eq. (35) when 7 = 1:

(ln X ) 3
KdX) - 3

when X~ = 1 and X2 = X. Logically, 7 = 1 specifies that grain properties near the surface are
such that the grains are just starting to suspend. A surface with a broad grain size distribution in
steady eroding conditions should lose its smaller fraction to suspension with time leaving the
saltating fractions with 7 ~ 1. Then, the amount of material of a given size collected by a collector
of height X should be

Wx = fowzo'u, (ln X) 3 Awzo' (ln X) 2 (36)

3k + 2k

3.3. Vertical flux

In wind tunnel studies the full depth of the suspension/saltation layer may not evolve. This means
that the vertical flux may not be so small at large heights; this adds constant particle
concentration through the vertical (see Eq. (33)). Of course, in the real situation, there should be
a vertical flux that depletes the surface and the horizontal flux must increase with x.
The constant term, Co, in Eq. (33) allows that there be a net, steady-state vertical flux, in the
absence of streamwise divergence, i.e., Qf/bx = 0. This could result from a uniform vertical
emission from the source or sedimentation down through the whole profile, i.e., material from
outside or upstream uniformly distributed in the vertical. Alternatively, this flux could simply be
a result of omission of the horizontal divergence term in the mass balance Eq. (4). Here we explore
A mathematical model of suspension with saltation 15

the fluxes that contribute to the total and calculate a value for Go, the steady vertical flux.
The total vertical flux is the sum of the eddy flux, the sedimentation flux and the saltation
flux.

Eddy flux

K ~ d? = ku*Aba(a + 1) a In X
3 OZ B a+ 2
all7

Sedimentation flux

ku,yfd7 - B. B2 -k ~ - k u , (ff0
ally

Where go is the first moment of the distribution function go.

Saltation flux

ku,Aab ~
G~= f ~_ O~-
3C~ dz + const -
Ba+ i
+ ku.N;

Here the term const is distribution dependent but is likely to be zero because saltation has
a limited reach in the vertical. In the wind tunnel the vertical reach is limited, however, so the term
remains.
The steady, vertical flux is the sum of the eddy, sedimentation and saltation fluxes:

Go = Geddy q- Gsed -}- Gsal~,

Go = - k u . % + ku,Jgo,

which is a consistent result. As this is a linear problem, the whole problem can be considered
a sum of three separate solutions. One, totally a steady-state with no vertical flux, has both terms
making up Go equal to zero. This is the normal steady solution without outside influence where
the suspended material goes to zero at z = infinity. A second solution is for the case of no wind.
Here, we note that No and Jfo are both first moments of distribution functions in 7 and the u,
dependence does not appear. This should apply to particles (small) that are within the air and
simply fall out. The third solution consists of particles injected through the whole profile by
saltation. These particles should be large in order that they have sufficient momentum and be
carried to heights larger than the inner, constant stress layer.
Presently we have no information on either c~o or 24~ and it might be expected that they
would be zero. Within the latter two separated solutions mentioned above, they are associated
with distribution functions that simply amend the gamma functional form. In the case of source
limitation, it is expected that only go could be non-zero; in the case of transport limitation,
however, a 'fluffy' source might simply be injected through the whole profile.
A net vertical flux would usually be associated with streamwise divergence. If, indeed, this
16 w.D. Scott et al.

divergence were constant, a term containing ~ would appear in the solution to Eq. (28). This is,
perhaps, a rough way of determining the contribution of streamwise divergence to the vertical
flux.

4 Foundations
There is some question as to how large saltating particles may be and the simple expression for
K (above Eq. (7)) still remain valid. Gillette and Goodwin [34] used rather conservative criteria to
calculate the height below which their flux equation (similar to Eq. (8)) was applicable. Here we
take an optimistic approach and look for the maximum particle size for which we can expect this
approximation to have some physical credibility.
The eddy diffusion coefficient K derives from a first order approximation to the Reynold's
shear stresses using a gradient driving force for transport of horizontal momentum in the vertical.
If the particles closely follow the air movements one expects that the turbulent exchange of
particulates should equal the momentum exchange and the expression be valid. It is still possible,
though, that the eddy exchange coefficient concept can be used with some degree of success even
if the particles do not follow the air movements. This can be expected as long as the effect of
turbulence on the exchange is adequately described by an expression of the form

g = kcu,z, (37)

where for small particles kc = k, Von Karman's constant. In this paper it is assumed that kc
characterises the eddy exchange process for particle transfer; it may be a function of particle size,
density and shape, and when there is a distribution of sizes present, the overall turbulence in the
air produced by all the sizes present. The factor may also be a function of u,, Zo' and z. Here we
only presume that the linear form Eq. (37) obtains. We see that any other influence simply alters
the 'effective' value of 7 in Eqs. (4) or (25). This means that inferring 7 from erosion rates would
produce a mean value which might not be directly related to particle size.
Exploring the limits of validity of this optimistic approach, the classic data of Bagnold and
nearly all collections of saltating material show the logarithmic trends expected from the
solutions of Eqs. (4) and (5), measurements of particulates and snow. This suggests that the forms
are generally valid and that the profile can be fitted with an average 7. In addition, the data of Lin
and Agarwal, as presented in Hidy [40], suggest that the dimensionless particle deposition
velocity on a vertical surface without sedimentation is approximately constant at 0.1 for large
particles. That is, we expect that

6C
--K ~zz ~ 0.1 u , C .

But, to a rough approximation in this near-surface condition,

0C C
az z

if the surface is sticky. This means that

K ~ 0.1 u , z

for large particles. This rather rough form suggests that kc ~ k for large particles.
A mathematical model of suspension with saltation t7

Dyer and Soulsby [41] summarise the data on the ratio 4

K k
fi = -- ~ -- (38)
Kc kc
and suggest that fl < 1 for airflow. The inability of grains to follow the motion suggests that
fi > 1. They quote a value of 5 [41, p. 308] from observations of suspended sand clouds in water
[42]. The presence of the sand lowered the turbulence intensity from 20% to 11%. The
concentration fluctuation was of the same order as the concentration.
Opposing this, it is expected that centrifugal effects in eddies allow that fl < 1. Hinze [43] and
Graf [44] present data from both points of view. Ffirber [45] measured particles in a flame and
obtained 0.1 < fl < 0.5; Van Rijn [4t, p. 318] suggests that fl increases from less than one to
greater than one with increasing U,/vertical velocity of sediment. Lees [46] obtained 0.1 < fi < 1
from profiles of velocity and concentration in the sea. Generally, in sedimentology, the behaviour
of fi has been well studied but remains poorly understood; fi = 1 is probably a safe value.
Noteworthy is that fl is considered only a function of grain size, not z [41, p. 311].
Measurements on blowing snow by Budd [30] and Schmidt [47] also produce logarithmic
relationships of the form suggested. Generally these are rounded particles in saltation up to
about .3 mm in size. Schmidt [47] presents relative values of kc for different particle sizes that
suggest that kc ~ 1.6 for small particles but that kc ~ .4 for large sizes. The data presented in
Greeley and Iversen [32] from Budd [30] or Mellor [32] also suggest kc values above 0.4. In this
experiment it appears that there is a distribution of particle sizes with a 'more or less' constant
mean fall velocity. These data also suggest that kc is decreased with increasing wind speed. One
explanation for this is that significant turbulence is created by the saltation process itself.
To explore this matter in the present set of data, we allow a different k for particle movement,
k~ in Eq. (27). Substituting

= lnX

Eq. (27) becomes

af--5 + 7- ~--0, (39)

where fl is as defined above, Note that the only assumption is that kc is not a function of z,
following the general approach in sedimentology. The solution to this equation is

f = for(/~ + 1, ~r + 1), (40)

where

F(a, x) = ~ e-tto-ldt
x

is a form of the incomplete gamma function and F(a) is the normal gamma function (where x = 0
in the integral). As before, Eq. (39) is linear in the two independent variables ( and 7, so the direct
effects of the velocity of fall can be exactly separated. This means that integration with respect to
the velocity of fall is always possible; even though we may not know the velocity of fall. The
vertical dependence is separate from this and the equation can be integrated with respect to

4 This is the inverse of their ratio to conform with Eq. (39).

18 W.D. Scott et al.

velocity of fall 5 for all values between 0 and infinity. This probably means that the "velocity of
fall" can have no deterministic dependence on height.
Allowing that/? m a y be size dependent, Eq. (40) was integrated using Simpson's Rule. The
fitting of this alorithm to the data of Gerety and Slingerland [10] made it possible to estimate/L
Table 3 shows the results, assuming/~ = 1 + eT, where e is a constant. This suggests that, indeed,
is a constant, a r o u n d 0.01 to 0.02 sec/cm. This result is questionable because of the small
number of experiments and data. Still, the fit is at least as good as the fit w i t h / / = 1 (see Table 1).
Minimum and m a x i m u m values are calculated on the basis that the airborne particles cannot
be larger or smaller than the sizes of particles measured at the source (see Fig. 5). The average e is
0.014.
It is somewhat to be expected that/? should increase with fall speed (particle size) and kc be less
than k. The worst case (see Table 3), however, with vt ~ 680 cm/sec or a quartz particle of about
0.3 ram, is with kc = 0.1. This is a factor of 4 but, surprisingly, has little effect on the fit. This is
probably because these largest particles are present in relatively small quantities. Overall, it is not
clear how valid it is to use Eq. (37). Still, the evidence does not invalidate the general form and
even the use of kc = 0.4 has pragmatic value.
Another assumption is that the particles move with the velocity of the horizontal wind. This
influences the calculation of horizontal fluxes and affects the analysis of Scott and Carter [23].
There is evidence that, on average, even the largest particles m a y move with the horizontal wind.
Schmidt [47] made measurement of the velocity of individual (rounded) crystals of blowing snow
with sizes between .06 and 0.4 ram. He found no statistically significant systematic deviation of
the mean particle velocities and the mean wind speed between heights of 0.05 and 1.0 m. Greeley
et al. [48] have measured individual particle velocities of quartz particles of 92 ~tm and 350 gm in
conditions appropriate for Earth, Venus and Mars. In the dense Venus atmospheres particles
almost always attain the velocity of the airstream; in the Earth case, however, this is not the case
with individual particles. As expected, the particles exhibited horizontal velocities less than that
of their surroundings if they arc ascending and velocities greater than their surroundings when
descending. On average, however, these particles have velocities not very different than their
surroundings. This suggests that it is reasonable to presume that average particle velocities
approximate the free stream velocity up to sizes of about 0.3 ram.
Another important assumption is that the layer with erosion is a 'constant stress' layer. This
without doubt exists within the atmospheric boundary layer, particularly when it is near to
equilibrium [49]. However, if a constant stress layer exists in the wind tunnel, it is of a relatively
shallow depth [50], [51]. This does not mean that a logarithmic profile of wind does not exist but
simply that an alternate form of the equation may be necessary when, say, a linear shear profile
exists.

Table 3. An indication of the effect of variable/~ on the fitting process

assumes/? = + ey
Exp. e u. /~ Rel. standard
rain. max. deviation

1 .012 28 1.03 3.7 3%

2 .02 46 1.03 3.8 15%
3 .015 76 1.01 1.5 10%
4 .01 113 1.00 1.2 12%

5 Note that we specifically say "velocity of fall" as distinct from "fall velocity" because the particles,
especially the larger ones, will not have reached their distinct "fall velocity".
A mathematical model of suspension with saltation 19

5 Conclusion

This approach supplies an adequate simulation of the effects of saltation following the original
perceptions of Bagnold and more recent work. The surprising result is that the observation of
saltating or jumping grains plays but a small part in a practical simulation of the vertical profile
of particulates in the air. That is, the data presented are close to the ground (0.5 mm) and contain
details of the grains in the air. Correction of the velocity profile for overall density removes the
"artificial" boundary of the saltating layer as far as dynamics is concerned. The above work
firmly asserts that, with the use of the velocity of fall and the vertical height as separate,
independent variables, the mass balance of individual sized particles extends to include
a saltation term and any portion of the size distribution must be able to act independently.
An alternative is that ]3 simply increases with large particle values of kc less than k. Equa-
tion (25) and Eq. (39) have both been fitted to the data and show little improvement in the fit
over the simple suspension model (Eq. (7)). Either the 'logarithmic' term that represents saltation
is ineffectual or the identity of saltation is lost within be eddy exchange process. In either case,
there is no quantitative reason to consider saltation as directly contributing to the concentration
profile. It may require an extreme lack of a steady-state to see any practical effect of saltation
on the particulate profile. The saltation effect is most simply, conveniently, and logically
considered as equivalent to a component of turbulent, eddy exchange, with perhaps an exhanced
exchange coefficient. Consideration of saltating layers may be unnecessary, except that they
correspond to large values of zo' and increased surface concentrations. Occam's razor suggests
that we simply consider saltation is a variety of the eddy exchange process, with an increased
eddy K (see Table 2). Easier, yet, we simply ignore saltation as far as the concentration profile
is concerned.
These conclusions are moderated, however, by a 'forward' approach [18]. That is, the
'forward' problem of finding a concentration profile from a given profile of scalar source or sink
density is more tractable and well-posed than the 'inverse' problem of finding the source or sink
density for a given profile. Carminati et al. [20] show clearly that the problem of obtaining
a near-surface shear stress from a velocity profile is not properly posed. The opposite 'forward'
case of describing a velocity profile from shear stress information is insensitive to that
information. Hence, concentration profile data, no matter how detailed, may not resolve all the
details of the erosion process. More and different information may be needed. This could be one
reason that little progress has been made in obtaining universal formulae for transport rates [17].

6 Summary

Two expressions are presented to treat particle lift-offduring wind erosion: a simple flux balance
produces a simple expression for suspended particulate concentrations Eq. (4). This form is really
only applicable to material in suspension. The influence of saltation is considered by evoking
Bagnold's observation that the u* is unaffected by the presence of driving sand; it follows that the
effect of density caused by the airborne grains is exactly offset by momentum extraction of the
saltating grains. A simple tractable solution to the overall suspension/saltation problem is
produced Eq. (25) and Eq. (29). A fitting to real data, however, suggests that saltation does not
play a direct part in the development of the concentration profile.
20 W.D. Scott et al.

Appendix
The fitting of a g a m m a distribution to a size distribution function follows the notes of D r e w [52].
The general distribution is of the form

ba
fit) = ~ t a - l e -bt,

where a and b are the two parameters of the distribution. Using the principle of m a x i m u m
likelihood, we find the derivations of the measuredJ~ from the values calculated for given G when
there are i d a t a points in the distribution. The sum of the squares of these deviations is formed;
this sum is minimised with respect to a and b by taking derivatives. This gives

and

In a F'(a) _ In [ - In t,
r(a)
where the g a m m a function ratio is the logarithmic derivative of the g a m m a function, the psi
function. Algorithms for the psi functions are given in A b r a m o w i t z and Stegun [53]. Solution is
by trial a n d error.

Acknowledgements
The authors would like to thank Dr. D. J. Carter and Paul Findlater of the W. A. Department of Agriculture
for their helpful criticism and guidance. This work was supported by the National Soils Conservation pro-
gramme and the Land and Water Resources, Research and Development Corporation under grant UMU5.

References
[1] Bagnold, R. A.: The physics of blown sand and desert dunes. London: Chapman and Hall, 1941,
reprinted 1984.
[2] Chepil, W. S., Woodruff, N. R: The physics of wind erosion and its control. Adv. Agronomy 15, 211 -- 302
(1963).
[3] Skidmore, E. L.: Wind erosion calculator: revision of residue table. J. Soil Water Conservation 38,
110-112 (1983).
[4] Owen, P. R.: Saltation of uniform grains in air. J. Fluid Mech. 20, 225-242 (1964).
[5] Ungar, J. E., Haft, P. K.: Steady state saltation in air. Sedimentology 34, 289-299 (1988).
[6] Werner, B. T.: A steady-state model of wind blown sand transport. J. Geology 98, 1 - 1 7 (1990).
[7] Anderson, R. S., Haft, P. K.: Wind modification and bed response during saltation of sand in air. Acta
Mech. Suppl. 1, 21--51 (1991).
[8] Zingg, A. W.: Wind-tunnel studies of the movement of sedimentary material. Studies in Engineering:
Bulletin No. 34, Iowa University 1953.
[9] Williams, G. P.: Some aspects of the eolian saltation load. Sedimentology 3, 257-287 (1964)_
[10] Gerety, K. M., Slingerland, R.: Nature of the saltating population in wind tunnel experiments with
heterogeneous size-density sands. In: Developments in Sedimentology 38 Eolian Sediments and
Processes (Brookfield, M. E., Ahlbrandt, T. S., eds.), pp. 15--131. Amsterdam: Elsevier 1983.
[11] White, B. R., Mounla, H.: An experimental study of Froude number effect on wind-tunnel saltation.
Acta Mech. Suppl. 1, 145-157 (1991).
A mathematical model of suspension with saltation 21

[12] Pye, K.: Aeolian dust and dust deposits. London: Academic Press 1987.
[13] Barndorff-Nielsen, O. E., Jensen, J. L., Neilsen, H. L., Rasmussen, K. R., Sorensen, M.: Proceedings of
the international workshop on the physics of blown sand. Memoirs No. 8. Denmark: Dept. Theor.
Statist., Aarhus University 1985.
[14] Brancatisano, T., Carter, D., Hopwood, J., Raupach, M., Scott, W. D., Smith, S., Tubb, J.: Erosion wind
tunnel workshop. Perth: Div. of Env. Sciences, Murdoch University 1991.
I151 Barndorff-Nielsen, O. E, Willetts, B. B.: International workshop on aeolean grain transport. Acta
Mech. Suppl. 1 and Suppl. 2 (1991).
[16] Scott,W. D., Findlater, R, Hopwood, J. M., Tubb, J.: Precis of the erosion wind tunnel workshop.
Contributions fi'om the wind erosion study group, 1994. See also B.A.M.S. 73, 821 (1992).
[17] Anderson, R. S., Willetts, B. B.: A review of recent progress in our understanding of aeolian sediment
transport. Acta Mech. Suppl. 1, 1 - 1 9 (1991).
[18] Raupach, M. R.: Saltation layers, vegetation canopies and roughness lengths. Acta Mech. Suppl. 1,
8 3 - 9 6 (199t).
[19] Butterfield, G. R.: Grain transport rates in steady and unsteady wind flows. Acta Mech. Suppl. 1,
97-122 (t991).
[20] Carminati, J., Devitt, J. S., Scott, w. D.: On the non-uniqueness of solutions in the modelling of steady
wind flows. Computers and Fluids (accepted for publication).
[21] McEwan, I. K., Willetts, B. B.: Numerical model of the saltation cloud. Acta Mech. Suppl. 1, 5 3 - 6 6
(1991).
[22] Rasmussen, K. R., Mikkelsen, H. E.: Wind tunnel observations of aeolian transport rates. Acta Mech.
Suppl. 1, 135-144 (1991).
[23] Scott, W. D., Carter, D. J.: The logarithmic profile in wind erosion: an algebraic solution. Boundary
Layer Meteorology 34, 303-310 (1986).
[24] Gillette, D. A., Walker, T. R.: Characteristics of airborne particles produced by wind erosion of sandy
soil, high plains of West Texas. Soil Science 123, 97-110 (1977).
[25] Smith, J. D., McLean, S. R.: Spatially averaged flow over a wavy surface. J. Geophysical Res. 82,
1735 - 1746 (1977).
[26] Scott, W. D.: Wind erosion of residue waste, part I. Using the wind profile to characterize wind erosion.
Catena 21, 291-303 (1994).
[27] Gillette, D. A.: Fine particle emissions due to wind erosion. Trans. ASAE 20, 890-897 (1977).
[28] Gillette, D. A.: Environmental factors affecting dust emission by wind erosion. In: Sahara dust
mobilization, transport, deposition (Morales, C. J., ed.), pp. 3 - 9 1 , New York: Wiley 1979.
[29] Hopwood, J., Scott, W.: The sizing of particles in terms of terminal velocity: A note on the interpretation
of data of Gerety and Slingerland with particles sized in terms of terminal velocity. Catena 17, 327 - 332
(1990).
[30] Budd, W. E: The drifting of non-uniform snow particles. In: Studies in Antarctic Meteorology, Antartic
research Series 9, (Rubin, M., ed.), pp. 59-70. American Geophysical Union 1966.
[31] Allen, J. R. L.: Physical processes of sedimentation. London: George Allen and Unwin 1970.
[32] Greeley, R., Iversen, L D.: Wind as a geological process. Cambridge: Cambridge University Press 1985.
[33] Weast, R. C., Astle, M. J. (eds.): CRC Handbook of chemistry and physics, 60th ed. Boca Raton: CRC
Press 1981.
[34] Gillette, D. A., Goodwin, P. A.: Microscale transport of sand-sized soil aggregates eroded by wind. J.
Geophys. Res. 79, 4080-4084 (1974).
[35] Tommerup, I. C., Carter, D. J.: Dry separation of microorganisms from soil. Soil Biology Biochem. 14,
6 9 - 7 1 (1982).
[36] Bell, E. P.: Statistical and micrometeorological modelling ofwindborne dust. Honours Thesis, Murdoch
University, Western Australia 1984.
[37] Sorensen, M.: An analytical model of wind-blown sand transport. Acta Mech. Suppl. 1, 6 7 - 8l (1991).
[38] de Ploey, J.: Some field measurements ans experimental data on wind-blown sands. In: Assessment of
Erosion (M. de Boot, D. Gabriels, eds.), pp. 541 - 552, Workshop on Assessement of Erosion in USA and
Europe, State University Ghent. New York: Wiley 1978.
[39] Nickling, W. G.: Eolian sediment transport during dust storms: Slims River Valley, Yukon Territory.
Can. J. Earth Sci. 15, 1069--1084 (1978).
[40] Hidy, G. M.: Aerosols, an industrial and environmental science. Orlando: Academic Press 1984.
 22 W.D. Scott et al.: A mathematical model of suspension with saltation

[41] Dyer, K. R., Soulsby, R. L.: Sand transport on the continental shelf. Annu. Rev. Fluid Mech. 20,
297- 324 (1988).
[42] Soulsby, R. L., Salkield, A. E, Haine, R. A., Wainwright, B.: Observations of the turbulent fluxes of
suspended sand near the sea-bed. In: Transport of suspended solids in open channels (Bechteler, W., ed.),
pp. 183-186. Rotterdam: Balkenner 1986.
[43] Hinze, J. O.: Turbulence. New York: McGraw-Hill 1975.
[44] Graf, W. H.: Hydraulics of sediment transport. Littleton: Water Resources Publ. 1984.
[451 F/irber, K.: Investigations of particle motions in turbulent flow. In: Transport of suspended solids in
open channels (Bechteler,W., ed.), pp. 33-36. Rotterdam: Balkenner 1986:
[46] Lees, B. J. : Relationship between eddy viscosity of seawater and eddy diffusivity of suspended particles.
Geo. Mar. Lett. 1, 149-254 (1981).
[47] Schmidt, R. A.: Vertical profiles of wind speed, snow concentration, and humidity in blowing snow.
Boundary Layer Meteorology 223, 223-246 (1982).
[48] Greeley, R., Williams, S. H., Marshall, J. R.: Velocities of windblown particles in saltation: Preliminary
laboratory and field measurements. In: Developments in Sedimentology 38 Eolian Sediments and
Processes (Brookfield, M. E., Ahlbrandt, T. S., eds.), pp. 133-148. Amsterdam: Elsevier 1983.
[49] McDonald, H., Camarata, E J.: An extended mixing length approach for computing the turbulent
boundary layer development. In: Computations of turbulent boundary layers (Kline, S. J., Morkovin,
M. V., Sovran, G., Cockrell, D. J., eds.), Vol. 1, Thermosciences Division, Dep. of Mech. Eng., Stanford,
page xxiii (1969).
[50] Raupach, M. R., Leys, J. E: Aerodynamics of a portable wind erosion tunnel for measuring soil
erodability by wind. Aus. J. Soil Res. 28, 177-191 (1990).
[51] Scott,W. D., Tubb, J.: An assessment of agricultural wind tunnels. In: Preprint of Conf. on Agricultural
Eng. (M. McKay et al., eds.), pp. 138-146, Toowoomba, Queensland 1990.
[52] Drew, D. R.: Introduction to operations research. Lecture Notes #2. Bangkok: Asian institute of
Technology 1973.
[53] Abramowitz, M., Stegun, I. A.: Handbook of mathematical function with formulas, graphs and
mathematical tables. New York: Dover 1972.

Authors' addresses: W D. Scott, Division of Environmental Sciences, School of Biological and Environmen-
tal Sciences, Murdoch University, Murdoch, Western Australia, 6150; J. M. Hopwood, Department of
Mathematics, University of Western Australia, Nedlands, Western Australia, 6009; K. J. Summers,
Engineering Department, ALCOA of Australia, GPO Box 252, Applecross, Western Australia, 6153,
Australia

View publication stats