Download as pdf or txt
Download as pdf or txt
You are on page 1of 16

JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 104, NO.

DI4, PAGES 16,827-16,842, JULY 27, 1999

A new model for dust emission by saltation


bombardment

Hua Lu, Yaping Shao


School of Mathematics, University of New South Wales, Sydney, Australia

Abstract. In this paper, we presentan analysison the mechanismof dust emission


and describe a new model for the prediction of dust emissionrate caused by
saltation bombardment,basedon the volume removalby impacting sand grains as
they plough into the soil surface. The model predicts a relationship between the
removedvolume V and the particle impacting velocity U of the form V cr U n, with
n _ 2 -• 3. The dust emissionrate is found to be proportionalto u,n+l , whereu, is
friction velocity. These model predictions are consistentwith wind tunnel and field
observations.The model offersa new interpretation of observeddata and a simple
scheme for the calculation of dust emission.

1. Introduction Wind tunnel experimentsallowingcontrolof soiltex-


ture and aerodynamicconditionshave beencarriedout
Dust produced by wind erosionis a major sourceof
atmosphericaerosolsthat may have a considerableim-
by $haoet al. [1993].It hasbeenfoundthat saltation
bombardmentappearsto be mainly responsiblefor dust
pact on atmosphericcirculationpatterns [Joussaume,
1990; Tegenand Fung, 1994]. For studiesof atmo- emission. On the basis of wind tunnel observations,
sphericradiation and global circulation patterns, it is $hao et al. [1993, 1996] proposedthat vertical dust
required to adequately estimate dust emissionrate so flux is proportional to horizontal sand drift intensity.
that dust concentrationin the atmospherecan be esti- They also proposeda theory based on the concept of
mated. However, a satisfactorydust emissionmodel so particle binding energy and energy balance of a single
far does not exist, albeit the considerableeffort of sev- saltator during particle and surfacecollision. The the-
eral researchgroups[e.g. MarticorenaandBergametti, ory of $haoet al. [1993]suggests that F is proportional
to u.,3 in contrastto the u,4 relationship suggested by
1995;Shaoet al., 1996].
Dust emissionrate has been conventionallyexpressed Gillette and Passi[1988].
in terms of aerodynamic variables. For instance, field The model of Shao et al. [1993]has two intrinsic
measurements problems. First, it is difficult to accurately estimate
have been carried out to establish a rela-
tionship between vertical dust flux F and surfacefric- the dust-particle binding energy • on either theoreti-
cal or experimentalgrounds. It can be shownthat the
tion velocityu. [Gillette and Walker,1977; Gillette,
1977;Nicklingand Gillies,1989, 1993].Field data sets uncertainty involved in the theoretical estimation of •
often have large scatter becausewind erosion is influ- is severalorders of magnitude. Second,during parti-
encedby a rangeofsoilproperties(e.g.,soiltexture,sur- cle/surfacecollisionthe kinetic energyof saltatingpar-
facecrust,soilcompactness, and soilmoisture),surface ticlesis not conservative,as a proportionof the particle
(e.g., roughness,vegetationcover,and kineticenergyconvertedto heat. Alfaro et al. [1997]
characteristics
proposeda conceptuallysimilar dust emissionmodel as
topography),and atmosphericconditions(e.g., wind
speedand precipitation).Consequently,
it is often dif- that of Shaoet al. [1993]. However,the two models
ficult to use the empirical relationshipsderived from are significantlydifferentfrom eachother in a quantita-
field experimentsto explain the physicsinvolvedin the tive sense,preciselybecauseof the differentassumptions
erosionprocess.Although there is a generalagreement made to accountfor the proportion of the kinetic energy
that dust emissionrate is proportional to u.• with n that is dissipatedduring particle-surfaceimpact.
varying between 2 and 5, there is no clear understand- Using high-speedphotographytechniques,Rice et al.
ing for what effectsn. [1996a,b] carriedout experiments
to examinethe possi-
ble mechanismof dust emissionby saltation. Their ob-
servationsrevealedthat sandgrainssaltatingovera sur-
face of loosefine particles excavatevoid-shapedcraters
Copyright 1999 by the American GeophysicalUnion.
in the bed. This phenomenonhas been known sincethe
earlierstudiesof Bagnold[1941,p. 57], who described
Paper number 1999JD900169. an eroded surfaceas being "pitted with tiny bombard-
0148-0227/99/1999JD900169509.00 ment craters a few grain diameters in size". Evidence
16,827
16,828 LU AND SHAO: MODEL FOR DUST EMISSION

of craterswere alsofound in severalmore recent exper- saltatingparticlesby a superpositionof individual im-


imentalstudies[Willerrsand Rice,1985;Linsey,1989; pacting events.
McEwanet al., 1992].A similarphenomenon hasbeen A similar approachused by Finnie and McFadden
observedin metal erosionwhichoccursat largerparticle [1978]for ductilemetal erosionis adoptedhere. The
impactingvelocity. It has been found that an impact- ploughingprocessof a singleparticle is illustrated in
ing particle may penetrate into the surfaceto a depth Figure 1. The impact particle is assumedto be angular
I 3
of a fraction of its own diameter, deforming the sur- but with an equivalent
diameterd, a massm = g•rppd
face plastically to form a concavityhaving a radius of (whereppis particledensity),and a momentof inertia
curvaturesimilar to that of the particle. On the basis I = md•'
12 ' Theoriginof thecoordinate
system
is lo-
of the observationsof crater formation, severalmodels cated at the centre of gravity of the impactingparticle
have been proposed for metal or ductile material re- when the particle starts to contact the surface.It is as-
movalby the impact of abrasiveparticles[Finnie and sumedthat the particle doesnot break during impact.
McFadden,1978;Hutchings,1977],whichare generally It impingesupon the soil surfaceat a velocityU (im-
in good agreementwith experimental data. The start- mediatelybeforeimpact) and an angleof attacka. The
ing point of these models is to considerthe motion of particle then ploughesinto the surfaceand pushessoil
impactingparticlesduring particle/bedcollision.The particles ahead of it. The target soil flows plastically
similarity between soil erosion and metal erosion mo- during the ploughing without fracture. The protrud-
tivates us to apply some of the analytical techniques ing tip of the incident particle ploughesa trajectory
usedin the industry wear researchto study dust emis- (XT, YT) into the target soil and formsa crater. The
sion during wind erosion. total volume of the grains ejected from the crater into
This paper presentsan analysison the mechanismof the air is assumedto be equal to that of the crater.
dustemissionand describes a newmodelfor the predic- The volume of the crater can be estimated by solving
tion of dust emissionrate causedby saltation bombard- the equationof particle motion and by determiningthe
ment. The model estimates dust emission rate on the
path of the particle as it movesthroughthe soil (see
basisof the volume removedby impacting sand grains Figure 1).
as they plough into the soil surface. In section 2, a To derivethe equationsof motionfor the ploughing
theoretical prediction for the volume removal, its rela- saltator, the forcesexerted by the target soil upon the
tionshipwith particle impact velocityand impact angle, saltator during the ploughing processneed to be de-
and that for dust emissionrate are derived. A compari- termined. For an analytical model, this can only be
sonof the theoretical predictionswith wind tunnel and doneunder severalsimplificationsthat make the analy-
field data is given in section 3. Discussionson the role sis mathematically tractable. These simplificationsare
of soil erodibility and the possiblemeasureof soil erodi- listed below.
bility are given in section4. Conclusionsof the paper 1. On average,impact particleshave no initial rota-
appear in section 5. tion and small rotation during the ploughingprocess.
It follows
thatforpolyhedral
particles,
XT "• X + •b
2. Dust Emission Model and YT • Y, with •bbeing the rotation angle, as shown
in Figure 1.
in our dust emissionmodel, we considersaltation 2. The ratio of the vertical force to the horizontal
bombardment as the main mechanismresponsiblefor force on the particle during ploughingis a constant K.
dust emission. The dust emission model consists of two
A reasonable
valuefor abrasivegrainsis K - 2 [Finnie
components. The first component is to predict vari-
andMcFadden,1978].
ables that characterizethe saltation process,such as
the number of saltating particles, particle impact ve-
locity, and impact angle. The secondcomponentis to
predict the amount of dust particlesejectedby saltation
bombardment. The saltation componentof the model
hasbeenextensively
studiedby manyresearchers
[Bag-
nold,1941;Owen,1964;Andersonand Haft, 1991]and
thereforeis not describedin detail in this paper. The
dust componentof the modelis describedbelow.

2.1. Model Concept and Assumptions


The proposedmodel is basedon the observationthat Figure 1. A schematicillustrationof the saltation
sand grains saltating over a surfaceof loosefine par- bombardment process. A saltating particle ploughes
through the soil, createsa small crater, and ejects par-
ticles excavate void-shapedcraters in the bed. The
ticles into the air. The horizontal and vertical compo-
model first calculates the crater volume and dust re-
nents of the force exerted on the particle by the target
leasecreated by individual saltating particles,and then soil are FH --pYb and Fv - KpYb, respectively. See
estimates the dust emissiondue to a large number of text for more details.
LU AND SHAO: MODEL FOR DUST EMISSION 16,829

3. A constant plastic flow pressure exists during ploughingparticle. It is assumedthat YT -- Y and


d
ploughing,and its horizontal componentis p.
4. The depth over which surfacecontactsthe particle Equations(1) and (2) can be solvedwith the ini-
is the same as that of the crater YT, as illustrated in tial conditions
(X It=0,Y It=0)- (0,0) and
dY
Figure 1. , at It=0)- (U cosa, U sinc•)'
5. The removed volume is the product of the area
swept out by the particle tip and the width b of the
ploughingface, X(t)--
UfiK
sinc• (Ucos•-Using)
sinfit+
U
K t (4)
V - b YTdXT
- b fO
tc
YTdXT
dt dt Y(t)-• sin
• sin
fit (5)
where tc is the time at which ploughingceases.
6. The vertical and horizontal forceson the particle
where
fi - •.
With the initial conditions
½ It=0= 0 andd•/dt It=0=
are located at the centre of the surface soil material in 0, the solution for angular rotation ½ is
contact with the particle. The symmetrical picture of
two-dimensionalploughing shown in Figure 1 can be
understood as the averagesituation for grains that are
½(t)
- 3'75U2
sin2
• [2(fit)
2+cos
tilted in either direction as they strike the surface. To 3U sin •
be consistentwith assumption2, the projected contact +d• (sin
fit- fit) (6)
area in the horizontal plane is twice that in the vertical
plane.
2.3. Volume Removal
Assumption 2 is sufficiently accurate if the particle
rotation during the ploughingis small, and hencea ge- It followsfrom assumption5 and equations(4), (5),
ometrically similar configurationis maintained through- and (6) that the volumeremovedby a singleimpactcan
out the process. Although there is no direct observa- be calculatedusing the followingexpression:
tional evidenceof the geometry,it is plausibleto assume
that K is a constantas the particle/bedcontacttime is V U 2 sin2 a
small(< 10-2 s, [Riceet al., 1996a]).Experiments
on (2 - cota) cos/•tc+ cota - 1.5
forcemeasurementsof dry surfacegrinding[Marshall
and Shaw,1952]showan overallvalueof K closeto 2
for angular abrasive grains. It seemsreasonableto use
this value in our analysisbecausethe processesinvolved
+U sin
-0.5cos2/•tc/• a 3.75U
sin
2a
7.5U2sin2 a
in ploughing and grinding are similar. The magnitude
of K may vary with the shape of the impact particle,
1
sin•t•+ • sin3•t•+ ) d

and it is possiblethat K increasesas the particles be- (7)


comelessangular and more spherical. Experiments of sin
fit•- •t•cos
rite
Marshalland Shaw[1952]on dry surfacegrindingalso
showsthat the value of K is usually larger than I and where t• is the impact duration.
lessthan 5. This suggeststhat the uncertainty brought Two ploughing cases should be distinguished. In
by K may be limited. As it will becomeclear later, the case 1, the impact particle ploughes into the target
precisevalue of K does not significantlyinfluencethe soil and, subsequently,leavesit when YT becomeszero.
conclusionsof our study. In case 2, the particle stops dur!•ngits scoopingac-
tion at some depth as its kinetic energy is exhausted,
dXTt=tc--O. Forcase1, wehaveY(t•) -0,
that is, W
2.2. Equations of Motion for Ploughing sin•t• = 0, and t• = •r/•, as can be seenfrom equation
Particles (5). It followsthat equation(7) canbe simplifiedto
The equations of particle motion in the X and Y V U2 7.5•rU 3 sin3 a
directionsand the equation of angular rotation are '•-= /•2(sin
2a--4sin
2a)+ /•3d (8)
d2X
m•-•-+ pYb- 0 (1) Forcase2, theinitialcondition
a•
dt = ax -[-•a•a4-- 0
d--•-
leadsto a rather complexexpressionof to, which can be
d2y derivedfrom equations(4) and (6)
m• + KpYb- 0 (2)
3.75U 2 sin2 a
d2•+ d_y)_2(KpYb)Y
- 0 (3) U cosa - 2U sina - sin(2/•t•)

where pYb and KpYb are the horizontal and verti- 7.5U2 sin2 c•
+2Usina cos•t• + /•t• = 0 (9)
cal componentsof the resistanceforce acting upon the
16,830 LU AND SHAO: MODEL FOR DUST EMISSION

The impact duration tc can be determinednumerically u. = 0.4 m/s. By setting Uref----3 m/s, K = 2, b = d,
from the above equation for a given c• and then used in andpp-- 2650kg/ma. Table1 summarizes
thevalues
equation(?) to calculatethe volumeremoval. of /•ref, /• (corresponding
to U), and n estimatedfor
We now examine the relationshipbetweenthe volume differentof p, a, and U/Uref. It revealsthe following
removal and the particle impact velocity on the basisof important facts:
equation(8). We note that the maximumdepth of the 1. For fixed impact saltator conditions(givena and
crater is U/Uref), n increasesfrom 2 to 3 with decreasingp. This
U
impliesthat V tendsto be proportional to U3 for soft,
¾max
= • sin
c• (10) loosesoils(corresponding to small p and hencelarge
and the ratio of Ymaxto the particleradiusd/2 is A), whileit tendsto be proportionalto U2 for more
compactsoils(corresponding to largep and hencesmall
,X- 2Usin
a / •rpp
f-• ),).
d3 : 2Usinav6-•
V• 2. For givenp, n is rather insensitiveto U/Uref for
fixed a. This implies that a plot of V against U on a
log-log scale is essentially a straight line for given soil
Both ¾maxand X are functionsof particle impact ve- surface conditions.
locity and angle;X is independentof particle diameter 3. For givenp, n increaseswith a for fixed U/Uref.
d if b = d is assumed.In terms of X, equation(8) can 4. Equation (12) showsthat the saltatingparticle
be rewritten as
size d has no explicit influenceon the value of n and A
mU 2 1 for givenU and a (as b = d is assumed).However,both
V = •-(sin2c• - 4sin2c•)+ 0.947rd2b/k
3 (11) U and a vary with d becauseof the particle/air inter-
2 p
actionon the saltator trajectories. Rice et al. [1995]
where we assumed K - 2. showed that U decreases and a increases with increas-
By fixingthe properties(U, a, d, and pp) of impact ing d. Since n is more sensitiveto the changeof a than
saltator, the effects of soil surface hardnesson the vol- to the changeof U, the overall effect of d on n is likely
ume removal and the relation between V and U can be that n increases with d.
shownby equation(11). The larger horizontalplastic For case2, in which the particle is trapped in the soil
flowpressurep (hardersurface),the smallerV (lessero- and doesnot rebound,equation(9) canbe simplifiedto
sionby saltationbombardment)is obtained.A precise
analysisof the effect p on the relationship between V 7.5Ufitc sin a
2 cosfit• - 2 + cot a +
and U can be done by consideringthe ratio of the vol-
ume removalsby the same impacting particle with two 3.75Usina
- sin 2fit• = 0 (14)
different velocities Uref and U
or in terms of A
v(v)
V(Uref) 2 cosfit• - 2 + cot a + 3.75Afit•
sin 2a - 4 sin2 a +
7.5•rUsinaa/fid -1.875Asin2fit• - 0
sin2a - 4sin2c•+ 7.57rUref
sinaa/fid
where Urefis an arbitrary referencevelocity that can be For A -• 0, equation(15) has no solutionfor t• for
set, for instance,as the averageparticle impact velocity a <_tan-l(1/4). It is thereforeunlikelyfor particles
during a wind erosion event. If the volume removal is with an impacting angle between 0ø and 14ø to be
proportional to the nth power of U, we obtain trapped in soil. Under this circumstance,particlesmust
leave the surface while still in the ploughing motion.
The experimentalresultsof Rice et al. [1996a]sup-
port this conclusionand their high-speedphotography
sin 2a - 4 sin2 a -1-3.757r(U/Uref),kref
sin2 showedthat saltating sand grains with an impact angle
(13) between 10ø and 15ø usually rebound after excavating
sin2a - 4 sin2a + 3.75W,kref
sin2a
a crater on the soil surface. However, solution exists
where /•ref is the crater depth to particle radius ratio for t• for larger impacting angles, which implies that
correspondingto Uref. particlesmay be trapped in soil in those circumstances.
From equation(12), effectiveexponentn can be cal- For instance, for a - 30ø and A - 0.1, the solution of
culated
foranarbitrary
given
value
ofa, /•ref,
andv-•fß rite from equation(15) is fit• = 1.8289.It followsfrom
It is widely acceptedthat a is typically between10ø and equation(7) that n - 3. A large impact angledoes
15ø [e.g. Andersonand Haft, 1991;Rice et al., 1995]. happen over an irregularly shapedsoil surface. As ob-
The wind tunnelstudyof Rice et al. [1995]showsthat servedin the wind tunnelstudyby Alfaroet al. [1997],
U falls mostly between3 and 4 m/s for saltator di- a considerableproportion of sand grains saltating over
ameter d - 150 •0 600 /•m and friction wind velocity a loosely packed clay particle surface do not rebound.
LU AND SHAO: MODEL FOR DUST EMISSION 16,831

Table 1. Maximum Crater Depth Z•ref,z•, EffectiveExponent n for PloughingCase 1 for Different Plastic
Pressurep, Impact VelocityRatio U/Uref and Impact Angle a

p (N/m•.) v
Uref C•= 10ø a = 13ø a = 15ø
•ref • n •ref • n •ref • n

1.25 x 105 2 0.0776 0.1552 2.1515 0.1006 0.2011 2.2661 0.1157 0.2314 2.3588
1.5 0.1164 2.1329 0.1508 2.2380 0.1735 2.3257
1.25 0.0970 2.1224 0.1257 2.2214 0.1446 2.3057
1 x 105 2 0.0868 0.1736 2.1664 0.1124 0.2248 2.2883 0.1293 0.2587 2.3846
1.5 0.1302 2.1463 0.1686 2.2587 0.1940 2.3506
1.25 0.1085 2.1349 0.1405 2.2412 0.1617 2.3298
0.75 x 105 2 0.1002 0.2004 2.1871 0.1298 0.2596 2.3184 0.1494 0.2987 2.4188
1.5 0.1503 2.1652 0.1947 2.2872 0.2240 2.383
1.25 0.1253 2.1526 0.1623 2.2685 0.1867 2.3623

1.25 x 10a 2 0.2455 0.4909 2.3590 0.3180 0.6360 2.5316 0.3658 0.7317 2.6364
1.5 0.3682 2.3259 0.4770 2.4960 0.5488 2.6034
1.25 0.3068 2.3059 0.3975 2.4732 0.4573 2.5817
I x 10a 2 0.2744 0.5489 2.3848 0.3555 0.7110 2.5589 0.4090 0.8181 2.6615
1.5 0.4116 2.3508 0.5333 2.523 0.6135 2.6297
1.25 0.3430 2.3300 0.4444 2.5010 0.5113 2.6085
0.75 x 10a 2 0.3169 0.6338 2.4191 0.4105 0.8210 2.5937 0.4723 0.9446 2.6927
1.5 0.4753 2.3841 0.6158 2.5593 0.7085 2.6625
1.25 0.3961 2.3626 0.5131 2.5369 0.5904 2.6422

1.25 x 10s 2 0.7762 1.5524 2.6366 1.0055 2.0111 2.78O4 1.1569 2.3138 2.8459
1.5 1.1643 2.6037 1.5083 2.7561 1.7354 2.8275
1.25 0.9703 2.5819 1.2569 2 7394 1.4461 2.8146
1 x l0 s 2 0.8678 1.7356 2.6618 1.1242 2.2484 2 7988 1.2935 2.5869 2.8598
1.5 1.3017 2.6299 1.6863 2 7761 1.9402 2.8428
1.25 1.0848 2.6088 1.4053 2 7603 1.6168 2.8308
0.75 x l0 s 2 1.0021 2.0041 2.6929 1.2981 2.5963 2 8208 1.4936 2.9871 2.8762
1.5 1.5031 2.6627 1.9472 2 8000 2.2404 2.8609
1.25 1.2526 2.6424 1.6227 2.7855 1.8670 2.8501

Here Uref--3 m/s.

There exists a critical angle ac that separatesthe two pu.2


(1
2 2
ploughing casesand C•cis positively related to A. For = m(Ux
-u't/u*)
cos
ax- Ucos
a)
(17)
instance, for A = 0.1, C•cis about 20ø compared with
about 14 ø for A • 0.
where u,t is the threshold friction velocity for sand par-
2.4. Vertical Dust Flux
ticles, U1 is the ejection velocity of sand particles, and
c• is the ejectionangle[Raupach,1991].
We first consider dust emission from a soil that con- Substituting(17) and (8)into (16), we obtain
tains multisized dust particles and uniform sand parti-
cles(with diameterd). If ns is the particlenumberflux
density(numberof impactingparticlesper unit areaper cNfpbU2 pu,•(l_ 2 2
F(d) =
unit time) and f is the Ëactionof dustcontainedin V, 2p U1cosal - U cosa
the vertical dust flux causedby saltation bombardment
can be calculated by (sin2a- 4sin
2a+7'5•rU
sin3
•d a)
F(d) = CNnspb
f V (16)
According
to Owen[1964],the horizontalsaltationflux
where Pb is the bulk density of soil and cs is a con- for a dry soil with uniform particlesis
stant of proportionality lessthan 1, as a proportion of
dust particles may stick on aggregatescontained in V.
Following the well-developedline of sand saltation mo-
mentum transfer near the surface,we have
Q(d)
= cpu*3
(1- u,t
2 /u,•) (19)
16,832 LU AND SHAO: MODEL FOR DUST EMISSION

where c is the Owen coefficient and g is gravitational face),kl -• 1, and for smallerp (softsurface),kl • 1.5;
acceleration.Equation (18) can be expressed
in terms
of Q(d) as (3)forsufficiently
large
psothat
C•u,v•<<0.24,
F/Q wouldbe independentof u,, while for sufficiently
F(d)= C•gf
2ppbQ(d) small
psothat
C•u,
V• >>0.24,
F/Qincreases
lin-
early with u,.
The value of cs, the fraction of dust particles that
(sin
2a
- 4sin•
a-[-
C•u,
••-•) (20) becomesuspendedfrom the ejected soil volume,is not
amenable to direct measurement. However, the exper-
where
imentsof Rice et al. [1996a]showthe grainsfromthe
crater are ejectedinto the air as a densecloud and grad-
Ca -• ½N
½ U1cos
oz
1- Ucos
oz ually disperses. Dust emissionsin natural conditions
are usually observed as soon as saltation is initiated
Cs = 7.5•sin
sa • • . [Gilletteand Walker,1977].It suggests
that cs should
not be muchsmaller(severalorderssmaller)than 1 for
loosely picked particle surfacesthat are prone to wind
As U/(U1 cosal - U cosa) is of the orderof 1, U/u.
is of the orderof 10 [Shaoet al., 1993]and the Owen erosion. More detailed experimental studiesare needed
coefficientis typically0.8 • 1 [Shaoet al., 1996];C• is to fully describethis parameter. In reality, particle im-
approximately10CN. By assuminga = 13ø and b = d, pact velocity U and angle a are also functionsof par-
C• isapproximately
0.137•,which
isoftheorder
of1. tide sized [Riceet al., 1995, 1996b]. If we substitute
Equation(20) then can be simplifiedto equation(20) into equation(23), the behaviorof F/Q
is much more complex. The analysis presentedin this
sectioncan be usedto explain someof the large scatter
= 2p (0.4
+ encountered in field data.
Using equation(22) to calculatethe total vertical
To estimate dust emission rate from a soil with adust flux F, two types of information about soil texture
particlesizedistributionp(d), we separatesoilparticles are required. One is a minimally dispersedparticle size
into the categoriesof dust and sand. An integration distribution(MDPSD) whichreflectsthe in-situparti-
of equation(21) oversandparticlesizesgivesthe total cle sizedistribution(p(d)in equations(22)and (23)).
dust emissionrate F inducedby saltation bombardment Anotheris the finecontentf, requiredin equations
(21)
of sand grains of all sizes and (24), whichcannotbe well represented by MDPSD
but can be estimated usingthe fully dispersedmethod.
Both the minimally dispersedparticle size distribution
F - F(d)p(d)Sd (22)
1 (MDPSD) and the fully dispersedparticlesizedistri-
bution (FDPSD) are analyzedby Griffith University
wheredl and d2 are the lowerand upperboundof sand
particle diameters. The ratio between the total vertical
Particle-Sizing
Laboratory.
Theyareexpressed
as
intervals(38 sizeclasses)
from 2 to 1159/•m. The tech-
dust flux and the total horizontal sand flux is
niques used to obtain these particle size distributions
aredescribed
by McTainshet al. [1997].The lowerand
upper limits for sand particle diameters dx and d2 can
= • (23) be set as 60/•m and 1159/•m for simplicity[Baghold,

Q 1 Q(a)p(a)5a 1941]or can usemorestrict aerodynamic-based


tions, as suggested
defini-
by Shaoet al. [1996]. The upper
limit for the dust particle size can be definedaccording
As an approximation,it is possibleto simplifythe above to aerodynamic conditions[Shaoet al., 1996],or de-
equation as finedas requiredfor comparison with observations(for
example,set to 20/•m). If high-resolution FDPSD is
available, f can be estimated as the sum of each dust
F=C•gfPb(O.24+C•u.••-•)
Q 2p(24) particle fractions,or as a singlefraction if the informa-
tion about dust emissionrate from the specificparticle
whichis a simplerearrangement of (21). size range is of interest.
Equation(24) revealsseveralqualitativelyimportant
relationships:(1) sinceQ is proportionalto u,,3 F must
be proportionalto u,i with i = 3 ,,• 4; (2) the saltation 3. Comparison with Experiments
bombardment efficiencyratio F/Q is linearlypropor- 3.1. Wind Tunnel Data
tional to the fraction of dusts containedin the parent
soil f and inverselyproportionalto soilsurfacehardness Rice et al. [1996a]carriedout wind tunnel studies
parameter
pk•withkl = 1 •- 1.5.Forlargep (hardsur- on crater volumes created by 250 -• 300/•m impact-
LU AND SHAO: MODEL FOR DUST EMISSION 16,833

ing particlesover a smooth and fine-grainedunaggre- Table 2. Predicted Volume Removal and Crater Depth
gated soil surface. They analyzedsevencraters and in Relationto Impact Velocity,Impact Angieand Saltat-
calculated the volume removal from measured contour ing Particle Size
profiles. For a fixed friction velocity u, = 0.4 m/s, d U c• = 10ø c• = 13ø c• = 15ø
they foundthe volumeof the cratersis in the rangeof
v Ym•x V Ym•x V Ym•x
0.1 •0 0.6 mm3 with the maximumcrater depth being
around 0.2 mm. The analysispresentedin the previ-
250 3 0.054 0.110 0.092 0.143 0.128 0.164
ous section can be compared with their experimental 3.5 0.081 0.129 0.141 0.167 0.198 0.192
data. To applyequations(8) or (11) to predictvolume 4 0.115 0.147 0.205 0.190 0.290 0.219
removal and the maximum crater depth, the horizontal
componentof plastic flow pressurep must be specified. 275 3 0.072 0.121 0.123 0.157 0.170 0.181
3.5 0.107 0.142 0.188 0.183 0.264 0.211
One
possibility
istoestimate
pbycombining/•
- V/iøKb 4 0.153 0.162 0.273 0.210 0.385 0.241

and• = 2Usina
•d givesp= 4raU2
• which
sin2
• ' For
3OO 3 0.093 0.132 0.160 0.171 0.221 0.197
saltators ranging between 250 and 300 pm, Rice et 3.5 0.139 0.154 0.244 0.200 0.342 0.230
al. [1996a]reportedthe meanvalueof impactenergy 4 0.198 0.176 0.354 0.229 0.500 0.263
!mU
2 2 = 2.31g cm2/s
2 for crust1 (a fragilecovering
of weakly bonded surfacecontaininguniform particles
as unaggregatedsoil [Rice et al., 1996a,Table II). It The unitsof d, U, V, andYmaxarein /•m, m/s, mma,
givesU = 3.51 m/s for d = 300 pm and U = 4 m/s if andmm, respectively. p = 967.57N/m2 is usedfor the
calculation.
d - 275 pm. In addition, using the same wind tunnel
and u. = 0.4 m/s, Rice et al. [1995]reportedcollision
data for particles ranging between 150 -• 250 pm and
300 -• 355 pm, respectively. For the 150 -• 250 pm
be due to the fact that the crater may have a nonsmooth
impactor, U = 3.80 4-0.65 m/s (mean-+-SD).For the
boundary which makesthe actual volume of the crater
300 -• 355 pm impactor,U - 3.30 + 0.62 m/s. There-
slightly larger than the model defined. It is causedby
fore it can be concludedthat the average impact ve-
the microheterogeneous nature of soil surface,in which
locity on unaggregatedsoil by 250 -• 300 pm particles
the particlesare removedin their entirety.
shouldbe well aroundU = 3.5 m/s. SinceYmax= 0.2
The presentmodelcan alsobe comparedwith the in-
mm and maximumsaltatordiameterd - 300pm [Rice
vestigationsof aeolianabrasionof rocksand minerals.
et al.,1996a],
wehaveX = 2Ymax
d __
-- 4
3' Setting
K = 2,
b = d, a = 13ø, U = 3.5 m/s, andX = 5, Dietrich[1977]concluded
4 we obtain eters that the fundamentalparam-
that control aeolian abrasion are the kinetic en-
p = 967.57N/m2.
Table 2 showsthe volume removal, V, and the max- ergy of the impacting grain and the bond strength of
the abraded material. This was confirmed by Greeley
imum crater depth, Ymax,both estimated using equa-
tions(8) and(10)withp = 967.57N/m2. Threediffer- and Iversen[1985]who investigatedthe susceptibility
of surfacesto abrasion, $a, defined as mass of mate-
ent valuesof U, a, and d within the experimentaldata
rial eroded per particle impact. They found that for
range are usedto accountfor the uncertainties.
The predictedvolumeremovaland crater depth shown a givenimpactparticlesize,$a is proportional to U2,
while for a given impact velocity,Sa is proportionalto
in Table 2 are in good agreementwith the data reported
da. Therefore
by Rice et al. [1996a].It is shownthat for givenim-
pact velocitiesand impact angles,both volume removal mU 2
and crater depth increasewith saltating particle size. Sa cr daU• cr 2
This is consistent with the wind tunnel observations of
Shaoet al. [1993]. Table 2 alsoshowsthat both the Since$a = ppV, theseobservations are fully consistent
volume removal and the crater depth increasewith im- with equation(11) for largep with the first term dom-
pact velocity and impact angle for a given impact par- inant over the secondterm, which is preciseenoughfor
ticle size. It is coherent with the measurement of Rice shallow abrasion.

et al. [1996b]. With a cohesionless bed of sand-sized 3.2. Field Data


particles,Rice et al. [1996b]foundthat the numberof
ejectedgrainsper collisionincreasedwith increasingim- For practical purposes,it is usefulto establisha sim-
pact angie up to approximately 15ø. Above this angie ple relationshipbetweendust emissionrate and wind
speedfor givensoilconditions.Gillette[1977]presented
the number of ejectedgrains appearedto decrease.It is
predictedthat the smallestcratervolumeis 0.054mma, field measurements of vertical dust flux F and horizon-
and the largestcratervolumeis 0.5 mma. Both values tal sand flux Q as a function of friction velocity u, for
3
are slightlysmallerthan 0.1 mma and 0.6 mma, mea- nine differentsoils. It was shownthat while the Q cr u,
suredby Riceet al. [1996a].The smallunderestimation relationshipdescribeswell their observeddata, the F
may be resultedby the overestimationof p. It couldalso and u, relationshipshowslargescatteralthoughF gen-
16,834 LU AND SHAO: MODEL FOR DUST EMISSION

•..zrally
increases
with u,. Gilletteand Passi[1988]sug- parison between model and observationcan only make
gestedthat dust emissionrate is probably proportional sensewhen the simulated u,t is comparable with the
4
•O U,• measuredones. In this study, the calculationof u,t fol-
F - agu,4(1- U,t/U,) U, _•U,t (25) lows$hao et al. [1996]. It is basedon the formulaof
Greeleyand lversen[1985]for bare, dry soilsand takes
whereu,t is the saltationthresholdvelocityandag is the influencesof soil moisture, vegetativeresidue,and
a dimensional coefficient. surfaceroughnessinto account. Accordingto Gillette
NicklingandGillies[1989]reportedverticaldustemis- [1977]and Gillette[1988],both soil moisturecontents
sion rate at 13 sites in southern Arizona. A consider- and vegetative residue were very low and had little ef-
able degreeof scatter existsin the data if no distinction fects on total soil movement for soils 1 to 5. Our sim-
is made betweenthe soil surfacefeatures. They parti- ulationshowedthat u,t (henceQ) is not very sensitive
tionedthe data on the basisof surfacemorphologyand to the fractionof nonerodible
elements(seeTable4) for
land use, and on the basisof the percentageof silt and small values of vegetative residue but is sensitiveto soil
clay measured,respectively. The regressionalrelation- moisture.Thereforethe simulations
of u,t (henceQ) for
ship betweenvertical flux F and wind frictionvelocity soils 1, 3, 4, and 5 were done by varying soil moisture
i in which i con-
u, obtainedfrom their data is F c• u,, within the measuredrange, until the simulated u,t is
fined within i - 3 •- 4. Similar results are obtained closeto the measuredones. For soils2, 6, and 9 the mea-
in the dust emissionexperiment over Mali, West Africa suredsoilmoisturewasgivenby Gillette[1977].These
[Nicklingand Gillies, 1993]for five differentsoil sur- measurements were used in our simulation without ad-
face conditions. These F c• u,i relationshipsare well justment. As the particle sizedistributionsfor the seven
predicted by the dust emission model derived in thissoilshave a certain amount of particleswith a diameter
paper. d = 70 •- 80/•m, the estimated u,t, given in Table 4,
We now examinethe performanceof our model against is the threshold friction velocitiesfor d = 75/•m. The
the field measurements of Gillette [1977],usingequa- measuredu,t are asreportedby Gillette[1977],except
tions (22) and (21). Becausevery limited data points for soil 6. In the work of Gillette[1988],u,t for soil6
are available for soils 7 and 8, only soils 1, 2, 3, 4, 5 , is given as 0.6 m/s. However,in Figure 4 of Gillette
6, and 9 are usedfor comparison.Accordingto Gillette [1977],the smallestvalueof Q for soil 6 is shownas
[1977],soils1, 2, 4, and 5 havesandtexture,soil3 has u,t -• 0.52 m/s. Both data are listedin Table 4.
a loamy sand texture, soil 6 has a sandy loam texture, The (dry sieved)particlesizedistributionsof soil 3
and soil 9 has a clay texture. and soil 4, givenby Gilletteand Walker[1977],were
Since vertical dust flux F is proportional to horizon- usedto calculateQ (Figure2). Sincesoils1, 2, and5 are
tal sand drift flux Q, two kinds of comparisonsare at- sandy soilsand have similar features in soil aggregates,
tempted for F. One is fully simulated and the other soil moisture and surfaceroughness,as well as horizon-
is semisimulated.The fully simulated F is obtained by tal sandfluxes,their (dry sieved)particlesizedistribu-
first calculatingQ, usingthe saltation model of $hao et tion shouldnot differ greatly from that of soil 4. There-
al. [1996],and then relatingF to the predictedQ by fore we assumedthat soils1, 2, 4, and 5 have the same
usingequation(22). In order to useequation(24) to (dry sieved)particlesizedistributionbut differentdust
predict F, the value of p must be determined. The val-
particlecontents,whichweregivenby [Gillette,1977]
ues of p, given in Table 3, are calculated posteriorly by
(seeTable 3). To test the effectof particlesizedistri-
fitting measuredF, usingequation(24), and measured bution,we replacedthe particlesizedistribution(PSD)
Q. The fitted values of F are called semisimulatedF. of soil 1, 2, 4, and 5 with a high-resolutionMDPSD of
The discussion on the correctness of these fitted values an Australiansandysoildescribed
by Leyset al. [993].
of p and their comparison with the measurementsare Similarly, we replaced the PSD of soil 3, 6, and 9 with
given in section 4. a high-resolutionMDPSD of an Australian loamy sand,
One of the most important parameters for simulat- loam (crackingsesquioxide),
and clay (blackearth), re-
ing Q is the threshold friction velocity u,t. The corn- spectively. The particle size distributions of the Aus-

Table 3. SoilParametersUsedfor Calculationof VerticalDust Flux F for Comparison


with the
ExperimentalData of Gillette[1977]

Soil 1 Soil 2 Soil 3 Soil 4 Soil 5 Soil 6 Soil 9

5 5 5 5 2 1
Pb(kg/m3) 1000 1000 1000 1000 1000 800 700
Fine particlecontentsf (%) 4 4.5 18.5 3.2 6.9 22.3 72.0
Horizontal
pressure
p (105N/m2) 20 20 0.5 25 25 1.5 350
LU AND SHAO: MODEL FOR DUST EMISSION 16,835

Table 4. Soil ParametersUsed for Calculation of Sand Drift Q for Comparisonwith the Experimental Data of
Gillette[1977]

Soil 1 Soil 2 Soil 3 Soil 4 Soil 5 Soil 6 Soil 9

Soil moisture(%) 0.8 0.99 1.0 0.5 0.65 0.75 6.6


Nonerodibleelements(%) 0.45-0.8 0.01-0.1 0-0.05 1-1.5 0.8-1.3 0.05 0.01
Nonerodible elements Used 0.6 0.1 0.05 1.5 I 0.05 0.01
Observedu,t (m/s) 0.25 0.25 0.25 0.25 0.25 0.52-0.6 0.65
Simulatedu,t (m/s) 0.253-0.26 0.256-0.258 0.256-0.257 0.246-0.254 0.251-0.26 0.2453 0.9142
Simulated u** Used 0.2561 0.2576 0.2572 0.2542 0.2545 0.2430 0.9142

tralian soils are shown together with those of Gillette in our tests were the fraction for d < 50 •m given by
and Walker[1977]in Figure2. Gillette [1977]. The bulk densityof fine-texturedsur-
In all testswe usedCs - 1.37, c = 0.8 and pp = ßfacesoilsis in the rangeof 1000to 1300kg/ms and
2650kg/ma. Because the semisimulated p values(see coarse-texturedsurfacesoils are usually in the range of
Table3) areall about102-• 10a timeslargerthanpp, 1300to 1800kg/ms [Fothand Turk,1972].The bulk
the second
termin equation(24) is negligible
evenif Cs densityof loamysoils,according
to Gillette[1988],are
is set to as large as 10. While the fine particle contents in the rangeof 600 -• 950 kg/ms. For soils1 to 5,
givenby Gillette[1977]are for d < 50 •m, the emitted which are fine, loosesandor loamy sandand haveonly
dust particleswere measuredfor d < 20 •m and thus a small amount of aggregateslarger than 0.84 mm, we
the valueof f in equation(24) shouldbe set to rep- assumed
C, = 5 and pb = 1000kg/ms. The value
resent the dust Ëaction for d ( 20 •m. However,the C, - 5 implies that about 40% of the dust particles
highly erodible sandy soils are often bimodal with one contained in the removal crater will not be released if
mode closeto 2 •m and the other closeto 180•m. The we assumeU/u,-10 and c = 0.8. For soil 6, whichis
fraction of particles with diameter between 20 and 50 cloddy sandy loam with a dry aggregatemode around
•m was negligible.For this reason,the valuesof f used 100 mm and 36.7%loosematerial, we assumedC• = 2

SandySoil Sandyloam
350 ' ' ' • .... i ........ i ........ ! ........

15ø
t
ß

300

250

200 ß

150 .

100 .

50
ß

•0-• 10ø 10' 102 103 •0-' 10ø 10' 102 103
r--d/2 (]Jm) r=da (wn)

Loamy Sand SandySoil


200 ........ , ........ , ................. 250 ........ i ........ ! ........ i ........

200
150

150
100
100

50
50

10-' 10
ø 10' 102 103 0-I 10ø 10' 102 103
r=-d/2 (!.un) r=d/2 (gin)

Figure 2. Particlesizedistributions(PSD) for sand(soils1,2, 4 and 5), loamysand(soil3)


comparison with thosegivenby Gilletteand Walker[1977](in the sameform givenby Gillette
and Walker[1977]). The dottedcurveis the particlesizedistributionof Gillette and Walker
[1977],and the solidcurveis AustraliansoilPSD assigned.AssumedPSDsfor sandyloamand
clay used in the simulation are also shown.
16,836 LU AND SHAO: MODEL FOR DUST EMISSION

10-'

Soil 51
'1
ISoil x '••'lO
--•
10-2

104 10-6

Observed
10-4 ß Observed ß- --. Semi-Simulated 10-7
Simulatedby measuredPSDs.

10-5
10-'
..... Simulatedby assignned
PSDs. .....
Fu
IIy
Simu
lated
bY
AM,
::Sg•,:edd•;S?•
s
Fully Simulatedby
....
' ' •
,,
10-"
10-•

Soil4•

[Soil
4 ?x•x•
?x
,,,?
x
10-2
,,
" •'x 10-6

10-3

10-4 :/4
,
/x
ß 10
-7
lO-'

10-5 , , , , , i
10-"
10-' , . . 10-•

ISoil3l lSoi131
10-2 10-7

•?•310
-3 lO•..•

10-4 10-7
,

10-s , i
10'-•
10-' . . . i 10-4

I Soil 2l ISoi121
10-2 104

10-6
ß

10-4 10-7

10-5 10-•
10-' 104

10-2 10-6

10-7

10-6

10
-4
10-•
O.l
10-ø

u. (m/s) u. (m/s)

Figure 3. Comparisonof simulatedusingPSDs of Gillette and Walker [1977](solid curve),


simulatedusingassignedPSDs, and observed(solidcircles)horizontalsand drift Q (left-hand
side)and that of fully simulated(solidcurve:usingpredictedQ by PSDs of Gillette and Walker
[1977];dotted,usingpredictedQ by assigned PSDs),semisimulated (usingobservedQ, dashed
curve plus solid circles) and observed(crosses)vertical dust flux F (right-hand side) versus
frictionvelocityu, for Gillette[1977]soils1 to 5. Parametersusedfor the comparison are listed
in Tables 4 and 3.
LU AND SHAO' MODEL FOR DUST EMISSION 16,837

andPb: 800kg/m3. Fortheclaytexturedsoil9, which for the aerodynamic effects, may not exist, as pointed
has only 9.3% looseparticles, we assumedCs = 1 and
out by Shao et al. [1996]. The wind tunnel experi-
Pb= 700kg/m3. ment of Shaoet al. [1993]showedthat saltationbom-
Figure 3 shows a comparisonof simulated and ob- bardment is the main mechanism responsiblefor dust
served horizontal sand fluxes, Q, and a comparisonof emission during wind erosion. It is therefore justified
the fully simulated(usingsimulatedQ to calculateF), to relate vertical dust emission flux to horizontal sand
semisimulated (usingobservedQ to calculateF) and drift intensity. It has been found in severalexperimen-
observedvertical dust flux F for soils 1 to 5. Although tal studies[Gillette, 1977; Nicklingand Gillies, 1989;
some of the parameters used for the simulation could Shaoet al., 1993]that the ratio F/Q doesnot depend
not be estimated accurately,the predicted Q and F and systematically on u,. Becauseof the complexitiesin-
observedQ and F are in good agreementfor all soils, volved in dust emission,it appears difficult to simply
except soil 2. In that case, the simulated Q is about expressF in terms of u, with a fixed exponent and ex-
5 times larger than observed,while the semisimulated pect the relationshipto be applicablefor all soil types.
F, is smaller than both measured and fully simulated. While Gilletteand Passi[1988]and Shaoet al. [1993]
This indicates that soil 2 may have a different particle suggestedthe F ocu4 and F ocu,3 relationshipthis
* ,

sizedistribution(the typical modemay not be closeto study suggestsa F oc u,i relationship


. The exponent
180 ktm)althoughit wasclassifiedas a sandysoil. It is i can be larger than 3 and dependson soil conditions,
interestingto observethat the simulation of Q is, some- especiallysoil hardness. Some secondaryfactors, such
how, not very sensitiveto the particle size distribution as surface roughness and soil moisture, also influence
and its resolution,while it dependsstrongly on u,t. It is the F and u, relationship.
surprisingto seethat the simulatedQ and henceF, us-
ingthe assigned
PSDs(of the Australiansoils),arevery
closeto thoseobtained usingthe PSDs givenby Gillette
4. Soil Erodibility and Plastic Pressure
and Walker [1977](Figure 3). We have alsorepeated For a given wind velocity, dust emissionis determined
the simulation by using nine other PSDs for sand and by the capacity of the surfaceto releasedust particles,
loamy sand soils from the Australian Mallee area. The which is largely affected by the physical and chemical
same observation can be made. For soil 6 and 9 the propertiesof the soil [Zobeck,
1991]. It is well known,
fully simulated dust fluxes do not agree well with the for instance,that surfacecrust and soil aggregationsup-
observations,becauseof the poor performanceof the presswind erosionand reduce dust emissionrate. Also,
saltationmodel for aggregatedsoils(Figure 4). How- a high percentage of fine particles in a soil does not
ever, the semisimulatedvertical dust fluxes for these necessarilyimply that the soil must be a strong dust
two soilsstill agreewell with the observeddata. source. Soils with more than 10% clay are, for exam-
Despite the advantagesof parameterizing the dust ple, not easily erodible unlessthe surface is broken up
emissionrate as a function of wind speed, a universally by mechanicaldisturbance. Sandy and loamy soils are
valid expressionof F in terms of u,, which accountsonly usually more important dust sourcesbecausethey are

10o 10-3

[Soil Soil
10-4

10-2
10-5
10-3
10-6
10-4

10-7

10
-8,•
10-•

10-2
Soil 10-6

10-3 10-7

10-8

10-5
O.
ß Observed
Simulated
0.2
• , , ,

1•.0
I xObserved
-- - ß Semi-simulated

2.00.1
F,ully
0.2
simulatead
, ,

2.0
10-9

u, (m/s) u, (m/s)

Figure 4. As Figure3 but for Gillette[1977]soils6 and9. Because


no measured
PSDsfor these
soils are found in publications,only assignedPSDs are used for the simulation.
16,838 LU AND SHAO' MODEL FOR DUST EMISSION

less cohesive,although these soils have a small percent- It is conventionally called soil penetrometer resistance
age of dust particles. The mechanical stability of the (definedas the mean maximum penetrationpressure
soil is therefore important. (MMPP), or meanmaximumforceper unit area).This
Chepiland Woodruff[1963]definedthe mechanical techniqueis originally derived for investigatingthe ef-
stability of a soil as the resistance of the soil to the fect of soil surface strength on the seedlingemergence
breakdown by a mechanical agent, such as tillage, wind and erosioncausedby raindrop impact. For the caseof
shear, or abrasion from aeolian particles. Mechanical saltation bombardment, particle size has an important
stability depends largely on interparticle cohesion in effect when it is of the same order as the indent di-
the soil system. During wind erosion, saltating par- ameter. A decreasein grain size is accompaniedby an
ticles strike the soil surface and cause a certain amount increasein local soil strength as the dislocationsgener-
of disruption around the impact point. The amount of ated by the indenterare blockedby the grainboundaries
disruption dependson the cohesiveness of the solid par- [Riceet al., 1997]. It is thereforecriticalto choosea
ticles in the removed volume by bombardment. More suitable test indenter with a scale comparableto salta-
precisely, it depends on the total interparticle tensile tion particle size and with a load comparableto the
forcepresentedin that volume of soil which is disturbed stressbrought by the saltators.
by the impactof the saltatingsandgrain. Chepil[1955] In section3.1, we estimatedp - 967.57N/m2 for
suggestedthat the modulus of rupture, a measure of the unaggregated
soilbed usedby Rice et al. [1996a].
the cohesivestrength of dry briquettes, varied inversely This value is comparablewith the soil penetrometerre-
with the diameterof the soilparticlesa c< 2. sistanceobtainedby Rice et al. [1997]for the same
On the basisof the sameidea, Sinalley[1970]sug- soil texture with light spray fine-particle soil surface.
gestedthat soil erodibility can be measuredin terms of They found that MMPP for sprayedsurfacefalls be-
its tensile strength that is related to the packing den- tween1000and 106N/m2. The modelestimated
p is
sity, the coordinating number of the particle, and the slightly lower than the lower limit of their estimates,
interparticle bond strength. He showedthat the tensile which representsa situation when the dedicatedcrust
strength in a simple soil system to be inversely pro- soil surfaceby light spray is disturbed by previouspen-
portionalto d3. Both Chepiland Woodruff [1963]and etration. Therefore, it is reasonableto believethat the
Sinalley[1970]suggest that veryfinesoilsarelesserodi- estimated valuep = 967.57N/m2 is not too far from
ble than coarsesoils.Rice et al. [1997]suggested that the true value of p for the original unaggregatedsoil.
surfaceerodibility can be characterizedby using a mod- In section3.2, we estimatedp for sevensoilsstudied
ulus of elasticity. Wind erosion tends to take place in by Gillette[1977]and foundthat for sandand loamy
dry, less cohesivesoils that normally have small mod- sandsoils, p iswithintherangeof 10•025x 105N/m2.
ulus of elasticity. The elasticity can be neglectedfor Thesep valuesare comparablewith the field measure-
totally loose soil and only pure plasticity needs to be ment of surface MMPP of sandy soils by using a nee-
taken into account. Another frequently used indicator dle penetrometer [Callebautet al., 1985]but are about
of soil erodibility is the aggregatecontent. It is found 100 times larger than the valuesof modulusof rupture
that the moredry the aggregate(< 0.84 mm) content, measuredby Gillette[1988]. The modulusof rupture
the lesserodiblethe soil [Chepiland Woodruff,1963; definedby Gillette[1988]is the forceappliedto break
Gillette,1977].However,a precisemeasurementof soil certain length and thicknessof crust. For a givensoil
erodibility is not yet available becauseof the difficult surface,measurements
often showthat the modulusof
nature of the problem itself. rupture(whichis a typeof tensilestrength)canbe 10to
A common feature exists among the above men- 100 timessmallerthan penetrometerresistance(which
tionedmeasuresof soilerodibility(soilcohesion,
tensile is a type of compressive strength).However,the mod-
strength,modulusof elasticity,and aggregatecontent). ulus of rupture is of the nature of macrohardnessand
They are soil mechanicalproperties and closelylink to is not measurablefor loose,or weakly crusted soil sur-
soil stress-strainrelationshipsor yield conditions. How- faces. As wind erosionmainly occurson loosesoil sur-
ever, there is a great diversity in the measurementof soil faces,the modulusof rupture may haveonly limited use
mechanicalproperties, suchas thoseobtained in triaxial as an indicator of soil erodibility.
tests, and shear strength by vane tests, tensile strength Accordingto Rice et al. [1997]the soilsurfacepen-
by modulus of rupture tests, and compresslyestrength etrometer resistance
(hencep) varies
between
103N/m2
by penetrometer resistancetests. This is becausefor for the lightsprayfinesoiland107N/m2 forthe deep
material like soil, there is no uniqueyield conditionand wetted soil with the same texture contents. This shows
the yield is a function of the test method and the forces t aat p is not only a functionof particlesize. Evenlarger
imposed. variation existsfor a wide range of soil types. Figure 5
In our dust emissionmodel, the horizontal component showsthe calculatedcrater volume versusimpact veloc-
of plastic flow pressurep is a parameter that represents ity for four differentvaluesof p, corresponding
to loosely
soil erodibility. It can be determined by using a nee- packedunaggregated, lightly sprayedthen dried (two
cases)and deepwettedthen dried soil surface,which
dle type penetrometer with suitable penetration depth
load and speed[Riceet al., 1997]. consistsof the samefine particles(_< 53/•m) [Rice et
and well-controlled
LU AND SHAO' MODEl. FOR DUST EMISSION 16,839

102 i i i !
to use such empirical relationshipswithout considering
• Unaggregated,
n = 2.696,p= 967.57Nlra2 their limitations.
-- -- Lightcrust,
n = 2.383,p = !5000Nlra2
-- - Lightcrust,n -- 2.259,p = 50000Nlra2 The penetrometerresistance(or p) is directlyrelated
....... Deep
wetted,
n= 2.062,
p= 1500000
Nlra2
x Unaggregated(Riceet. al. 1996a) to interparticle bonding forces(or energy). Both of
OCrust I (Rice et. al., 1996a)
•l•Cmst 2, (Riceet. al., 1996a)
them are affectednot only by the particle size but also
ß Crest 3 (Rice et. al., 1996a) by wetting and drying processes,as shownby Rice et al.
10ø
[1997],by chemicalor watersoluble material,andby the
contentsof organic material. Becauseit is very difficult
to describe all the processesinvolved mathematically,
• 10-•
theoretical investigation of soil strength properties has
only been carried out for very simple, ideal cases. Al-
though some qualitative insight in the mechanism of
• 10-2
o cohesionhas been achieved by researchersin colloidal
>
science,it is unlikely that the average values of these
strength properties can be calculated with reasonable
I=• 0_3
accuracyif all the processesare considered.Practically,
these soil strength properties are obtained by experi-
mental measurementfor given conditionsunder specific
10-4
requirement. Obviously,the measurementof p is easier
than that of interparticlebondingforces(or energy).
10-5
5. Conclusions

In this paper, we have presenteda new dust emission


1 2 3 model based on the understanding that dust emission
4 5 10

ImpactVelocityU (m/s) is mainly causedby saltation bombardment. The pre-


Figure 5. Calculatedcrater volumeversusparticle im- diction of dust emissionrate is achievedthrough mod-
pact velocity for four differentvaluesof p, correspond- eling the ploughingprocessof individual saltating sand
ing to looselypackedunaggregated,lightly sprayedthen grains and the resulted volume removal of the surface
dried (two cases)and deepwettedthen dried soil sur- soil. It has been found that the crater volume removal is
faces,whichcontainfine particles(_< 53/•m) [Rice et proportionalto the impactingparticlevelocityV c• U•;
al., 1997].The diametersof impactparticleandimpact the vertical dust emissionrate is proportionalto friction
angle are set to the averagevalues d = 275 /•m and
velocity, F c• u,•+• , with n beingaround2 to 3. For
c• = 13ø, as indicatedby Rice et al. [1997]. The val-
ues of volume removal for unaggregatedsoil and three practical purposesthe vertical dust flux can be calcu-
averagevaluesof volume removal in relation to impact lated usingequation(23) or equation(24) for simplicity.
velocity,estimatedfrom Rice et al. [1996a]experiment The theoretical predictionof dust emissionrate is well
data, are also shown. supportedby the field observationsof Gillette [1977],
Gillette and Walker [1977], and Nickling and Gillies
[1989, 1993]. The analysispresentedin this paper
providesa clarification of the long-standingargument
al., 1997].The averageimpactparticlediameterd was amongresearchers wh•ther F is proportionalto u,
4
275/•m and the averageimpact angle c• was 13ø. The [$haoet al., 1993]or to u, [Gilletteand Passi,1988].
measured values of crater volume removal versus im- The theory providesan explanation for the wide diver-
pact velocity for unaggregatedsoil are retrieved from genceof the power relationship between F and u,. It
Figurell of Rice et al. [1996a].The impactvelocities has beenshownthat soil surfacestrength(represented
are estimated from the kinetic energy lost to the bed by penetrometer
resistance)
playsan importantrole in
by setting the ratio of ricochetto impact velocityequal determining the velocity dependenceof dust emission
to 0.57 of the meanvalue [Riceet al., 1996a]. Three rate.

averagevaluesof the volume removal in relation to im- The good agreementof the new dust emissionmodel
pact velocitycalculatedfrom the Rice et al. [1996a] with experimental data further demonstratesthat the
experiment data are alsoshownin Figure 5. It is shown exact dislodgementmechanismof dust emisionby salta-
clearly that p has a profoundinfluenceon dust emission tion bombardmentis indeed surfaceploughingand cra-
rate. tering. It raisesa questionover the basic assumption
It is noteworthythat the empiricalrelationshipsof made by energy-baseddust emissionmodels; namely,
field dust emissionrate versusu. are often derived by the energy of impact saltator lost to the bed is mainly
least squaresregressionfor differentsoiltexture and sur- used for rupture interparticle bonds of those ejected
faceconditions,and thus large differencesin p and large dust particles. It can be shownthat the impacting en-
data scatter are inevitable. It is therefore dangerous ergyislargelyconsumed by plasticdeformation(plough-
16,840 LU AND SHAO: MODEL FOR DUST EMISSION

ing, relocatinggrainsand frictionbetweengrains).The forces and hard to estimate theoretically. It may in-
fraction of energyusedto rupture interparticlebondsis creasewith impact velocity accordingto severalexperi-
actually negligible. mentalobservations [Gilletteand Walker,1977;Gomes
We have pointed out that there is a fundamental dif- et al., 1990;Alfaro et al., 1997].
ference in the forces involved in the entrainment of sand
particles by wind and in the dust emissionby saltation Acknowledgments. This work is supported by Aus-
bombardment. Unlike sand saltation, dust emissionis tralia Research Council. The authors are grateful to G. H.
McTainsh and K. Tews at Griffith University for providing
indirectly derived by wind force. It involvesthe rela-
us with the particle size distributions of several Australian
tionship betweenwind and saltation as an intermediate soils. We also wish to thank two anonymous reviewers for
process.The mechanismresponsiblefor the breakdown their constructive commentswhich greatly improved this pa-
of interparticle bondsby particle impact is fundamen- per.

tally different from that responsiblefor the lift-up of


References
sand particlesby aerodynamicforces. While the thresh-
old friction velocity u.t is the key parameter for char- Alfaro,S.C., A. Gaudichet,L. Gomes,and M. Maille, Mod-
acterizingthe surfacecapacityto resist sand drift, the elingthe sizedistributionof a soilaerosolproducedby sand-
horizontalplasticflow pressurep whichcan be measured blasting,J. Geophys.Res., 102, 11,239-11,249,1997.
Anderson,R. S., and P. K. Haft, Wind modificationand bed
as surfacepenetrometerresistance,emergesas the most response
duringsaltationof sandin air, Acta Mech.,Suppt.,
important quantity in characterizingthe soil physical 1, 21-51, 1991.
propertiesrelated to dust emission.It is predictedthat Bagnold, R. A., The Physics of Blown Sand and Desert
the dustemission
rate is inverselyproportional
to pk, Dunes, 265 pp., Methuen, London, 1941.
with k - I -• 1.5. For hard surfaceswith a large p, Callebaut,F., D. Gabriels,W. Minjauw, and M.D. Boodt,
Determinationof soil surfacestrengthwith a needletype
the dust emissionrate is more likely to be proportional penetrometer, Soil Till. Res., 5, 227-245, 1985.
to u,,3 whilefor softsurfaces
with a smallp, the dust Chepil, W. S., Factors that influence clod structure and
4
emissionrate is more likely to be proportionalto u,. erodibilityof soilby wind, IV, Sand,silt and clay,Soil Sci.,
Becauseof the large variation in soil conditions,it is 80, 155-162, 1955.
thereforenot surprisingthat a large scatterexistswhen Chepil, W. S., and N. P. Woodruff, The physicsof wind
erosionand its control,Adv. Agron., 15, 211-302,1963.
field measurementsof dust emissionrate are plotted Dietrich, R. V., Impact abrasionof harder by softermateri-
againstu., and both F c• u.3 andF c• u.4 relationships als, J. Geol., 85, 242-246,1977.
are observed. Finnie, I., and D. H. McFadden,On the velocitydependence
The new dust emissionmodel does, however,have of the erosionof ductile metals by solid particles at low
the followinglimitations: (1) In reality, especiallyfor anglesof incidence,Wear, J8, 181-190, 1978.
Foth, H. D., and L. M. Turk, Fundamentalsof Soil Science,
agriculturalsoils,dust particlesare not only ejectedby 436 pp., John Wiley, New York, 1978.
sand particles impacting the surface but also are re- Gillette, D. A., Fine particulate emissionsdue to wind ero-
leaseddue to the breakingdown of aggregatesduring sion, Trans. Am. Soc. Agric. Eng., 20, 890-987, 1977.
saltation and impact. The latter dust emissionmecha- Gillette, D. A., and T. R. Walker, Characteristicsof airborne
nism is not consideredin the present model. The com- particlesproducedby wind erosionon sandysoil,highplains
of west Texas, Soil Sci., 123, 97-110, 1977.
bination of these two processesand the random nature Gillette, D. A., Productionof dustthat may be carriedgreat
of soilaggregates,microtopography
(ripplesand ridges distances,Spec. Pap. Geol. Soc. Am., 186, 11-26, 1981.
that influencethe localimpactangle)makeit difficult Gillette, D. A., Thresholdfriction velocitiesfor dust pro-
to precisely predict the dependencyof dust emission ductionfor agriculturalsoils,J. Geophys.Res., 93, 12,645-
12,662, 1988.
rate on the impactvelocity,(2) the ratio of the vertical
Gillette, D. A., and R. Passi,Modelingdust emissioncaused
and horizontal plastic pressureK and the contact area by wind erosion,J. Geophys.Res., 93, 14,233-14,242,1988.
betweenthe ploughingparticleand the surfacesoilmay Gomes,L., G. Bergametti,G. Coude-Gaussen, andP. Rognon,
vary during collisionand cannotbe preciselydescribed, Submicrondesert dusts: A sandblastingprocess,J. Geo-
(3) because the elasticforceshavebeenneglected in the phys. Res., 95, 13,927-13,935, 1990.
Greeley,R., and J. D. Iversen,Wind As a GeologicalProcess
model,it may not be suitablefor highlycrustedsoil,for on Earth, Mars, Venusand Titan, CambridgeUniv. Press,
whichthe elasticstrainsare comparablewith the plastic New York, 1985.
stains. The erosionbehaviorof crustedsoil may not be Hutchings, I. M., Deformation of metal surfacesby the
the sameas looselypackedsoil surfaces,and additional obliqueimpact of squareplates, Int. J. Mech. Sci., 19,
paremetersmay be required in characterizingthe soil 45-52, 1977.
Joussaume, S., Three-dimensional simulation of the atmo-
physicalpropertieswhichcontroldustemission,(4) it sphericcycleof desertdust particlesusinga generalcircu-
is expectedthat the surfacedeformingproperty(or p) lation model, J. Geophys.Res., 95, 1909-1941,1990.
will exhibit a large spatial heterogeneityand temporal Leys, J., P. Butler, and C. McDonough,Wind erosionre-
variation, andp is moredifficultto measurefor practical searchat Borrika in the South Australian Murray Mallee,
purposes Tech. Rep., Dep. of Conserv.and Land Manage.,N.S.W.,
than parameterssuchasu.t, (5) uncertainties 1993.
may be introducedby parametercN; the fraction of dust
Linsey,T. L., The entrainmentof fine grainedsedimentby
particlesbecomesuspended from the volumeejectedby wind, Ph.D. 277 pp., thesis,QueenMary College,Univ. of
saltationbombardment.It is a functionof interparticle London, 1989.
LU AND SHAO: MODEL FOR DUST EMISSION 16,841

Marticorena, B., and G. Bergametti, Modeling the atmo- Rice M. A., C. E. Mullins, and I. K. McEwan, An analysis
spheric dust cycle, 1. Design of a soil-derived dust emission of soil crust strength in relation to potential abrasion by
scheme,J. Geophys. Res., 100, 16,415-16,430, 1995. saltating particles, Earth Surf. Process. Landforms, 22,
Marshall, E. R. and M. C. Shaw, Forcesin dry surfacegrind- 869-883, 1997.
ing, Trans. ASME, 74, 51-59, 1952. Shao, Y., M. R. Raupach, and P. A. Findlater, The effect of
McEwan,I. K., B. B. Willetts, andM. A. Rice,The grain/bed saltation bombardment on the entrainment of dust by wind,
collisionin sand transport by wind, Sedimentology,39, 971- J. Geophys. Res., 98, 12,719-12,726, 1993.
981, 1992. Shao, Y., M. R. Raupach, and J. F. Leys, A Model for pre-
McTainsh, G. H., A. W. Lynch, and R. Hales, Particle-size dicting aeolian sand drift and dust entrainment on scales
analysis of aeolian dusts, soils and sediments in very small from paddock to region, Aust. J. Soil Res., 34, 309-42,
quantities using a Coulter multisizer, Earth Surf. Process. 1996.
Landforms, 22, 1207-16, 1997. Smalley, I. J., Cohesion of soil particles and the intrinsic
Nickling, W. G., and J. A. Gillies, Emission of fine-graned resistanceof simple soil systemsto wind erosion, J. Soil Sci.,
particulates from desert soils, in Paleoclimatology and Pa- 21, 154-161, 1970.
leometeorology: Modern and Past Patterns of of Global At- Tegen, I., and I. Fung, Modeling of mineral dust in the at-
mospheric Transport, Edited by: Leinen M. and M. Sarn- mosphere: Sources,transport, and optical thickness, J. Gem
thein, pp. 133-165, Kluwer Acad., Norwell, Mss., 1989. phys. Res., 99, 22,897-22,914, 1994.
Nickling, W. G., and J. A. Gillies, Dust emissionand trans- Willetts, B. B., and M. A. Rice, Inter-saltation collisions,in
port in Mali, West Africa, Sedimentology,J0, 859-868, 1993. Proceeding of International Workshop on Physics of Blown
Owen, R. P., Saltation of uniform grains in air, J. Fluid Sand,Memoirs8 (Edited by O. E. Barndorff-Nielsen
et al.),
Mech., 20, 225-42, 1964. pp. 83-100, Dept. of Theor. Stat., Aarhus Univ., Denmark,
Raupach, M. R., Saltation layer, vegetation canopies and 1985.
roughnesslengths, Acta Mech. Suppl., 1, 135-144, 1991. Zobeck, T. M., Abrasion of crustedsoils,influenceof abrader
Rice M. A., B. B. Willetts, and I. K. McEwan, An experi- flux and soil properties, Soil Sci. Soc. Am., 55, 1091-1097,
mental study of multiple grain-sizeejecta producedby colli- 1991.
sionsof saltating grains with a fiat bed, Sedimentology,42,
695-706, 1995.
Rice M. A., B. B. Willetts, and I. K. McEwan, Wind erosion H. Lu and Y. Shao School of Mathematics, The Univer-
of crusted soil sediments,Earth Surf. Process. Landforms, sity of New South Wales, Sydney2052, Australia. (e-mail:
21, 279-293, 1996a. Y.Shao@unsw.edu.au)
Rice, M. A., B. B. Willetts, and I. K. McEwan, Observations
of collisions of saltating grains with a granular bed from (ReceivedJuly 12, 1998; revisedJanuary4, 1999;
high-speed cine-film, Sedimentology, J$, 21-31, 1996b. acceptedFebruary25, 1999.)
16,842

You might also like