You are on page 1of 5

CAN DISPERSIVE PRESSURE CAUSE INVERSE GRADING IN GRAIN FLOWS?

FRANCOIS LEGROS*
Institute of Earth Sciences Jaume Almera, Consejo Superior de Investigaciones Cientficas, c/ Llus Sole i Sabars s/n 08028 Barcelona, Spain
e-mail: fplegros@yahoo.com

ABSTRACT: A theoretical analysis is presented that suggests that dispersive pressure in a grain flow, modified grain flow, or traction carpet
cannot account for the upward segregation of large particles, as sometimes assumed. The analysis shows that rapid granular flows are selforganized in such a way that dispersive pressure at any level in a
shearing mass of grains is always equal to the applied normal stress.
An increase in dispersive pressure causes an immediate expansion of
the flow and a consequent decrease in dispersive pressure until it equals
the applied normal stress again. The gradient of dispersive pressure is
therefore hydrostatic, and only particles lighter than the bulk density of the flow are pushed upward. The inverse grading of heavy particles observed in a great variety of deposits is not caused by dispersive
pressure and must be explained by other mechanisms.
INTRODUCTION

Vertical inverse grading (also called reverse grading) is a common feature in many clastic deposits. It occurs in beach foreshore and berm overwash laminae, foresets of eolian and subaqueous dunes, and in the deposits of debris flows, turbidity currents, and pyroclastic flows and surges.
Deposits can be inversely graded throughout the bed or only in basal parts.
Since the 1970s, the most popular hypothesis in the geological literature
has been that inverse grading is caused by the segregation of large particles
away from the solid boundary because of the action of the dispersive pressure existing in a rapid granular flow. This mechanism has been invoked
in a great number of papers (it is not possible to list them all) for a variety
of deposits, including sand dunes (Lowe 1976), beach foreshore sands (Sallenger 1979; DeCelles and Cavazza 1992), deposits of hyperconcentrated
flows and debris flows (Schmincke 1967; Walton and Palmer 1988; Todd
1989; Best 1992; Cole and Stanley 1994; Sohn et al. 1999), turbidites
(Hiscott and Middleton 1980; Lowe 1982; Hiscott 1994; Sohn 1997), ignimbrites and pyroclastic-surge deposits (Sparks 1976; Sigurdsson et al.
1987; Cole and Scarpati 1993; Palladino and Valentine 1995; Gianetti
1998; Druitt 1998; Hughes and Druitt 1998), and in the deposits of rockslide avalanches (Cruden and Hungr 1986).
In contrast, the mechanism of size segregation by dispersive pressure is
not very popular amongst chemical engineers, and, in the engineering literature, other explanations for inverse grading are generally preferred. One
of them is that small particles fall through the holes that temporarily open
in an agitated granular mass more easily than large ones, and so tend to
concentrate at the base by kinetic sieving or random fluctuating sieving (Middleton 1970; Savage and Lun 1988; Cantelaube and Bideau
1995; Maske et al. 1997; Pouliquen et al. 1997). Some geologists have
suggested that inverse grading in some deposits may reflect a temporal
increase in the size of particles supplied to the progressively aggrading
deposit (Allen 1981; Branney and Kokelaar 1992; Kneller and Branney
1995; Hand 1997; Vallance 2000; Legros and Mart 2001).
Although the segregation of large particles due to dispersive pressure in
grain flows has been postulated for several decades in sedimentology, few
papers have attempted to propose a quantitative or even a qualitative physical treatment of the mechanism by which such a segregation could occur.
This paper presents an analysis, based on an idea originally proposed by
* Present address: Instituto Geofsico del Peru, Urb. La Marina B19, Cayma,
Arequipa, Peru
JOURNAL OF SEDIMENTARY RESEARCH, VOL. 72, NO. 1, JANUARY, 2002, P. 166170
Copyright q 2002, SEPM (Society for Sedimentary Geology) 1527-1404/02/072-166/$03.00

Middleton (1970), that shows that granular dispersive pressure cannot provide a force to lift heavy particles, so inverse grading must be explained
by other mechanisms.
ANALYSIS

Bagnold (1954) showed experimentally that a granular material sheared


between two plates develops a normal stress or dispersive pressure, P,
which is proportional to the shear stress and can be expressed by
P 5 f (C)s D 2

1 2
dU
dy

(1)

where s is the density of particles, D their diameter, dU/dy is the velocity


gradient or shear strain rate, and f(C) is a positive function of the particle
concentration, which Bagnold showed to be proportional to l 2, where l
is the linear concentration defined as the ratio of the particle diameter to
the mean distance separating two particles. The presence of D 2 in the expression for the dispersive pressure has led many authors to assume that
large particles would tend to migrate towards the zone of least shear strain
in order to equalize the dispersive pressure within the sheared granular bed.
It is not clear, however, why dispersive pressure should equalize in the
granular bed and how this would occur. Moreover, the explanation of inverse grading by this mechanism relies on the assumptions that shear strain
is minimum at the top of the bed and that particle concentration is constant
vertically.
Some authors seem to have assumed that the dispersive pressure acting
on a particle is proportional to the square of its diameter and hence is
greater for larger particles, thus favoring their segregation and the formation of inversely graded layers (e.g., Sparks 1976; Sallenger 1979; DeCelles
and Cavazza 1992; Cole and Stanley 1994). This is not the case, however.
As pointed out by Middleton (1970), the dispersive pressure felt by a particle depends on the stress imparted by the other particles through inelastic
collisions, hence on the diameter of the other particles, not on its own
diameter. The experiments of Bagnold (1954) as well as many of the more
recent experiments on granular flows use monosized particles. Now, if we
introduce one particle of diameter Di in a monosized granular material with
particle diameter D Di, the dispersive pressure acting upon this particle
will be just the same as that acting upon any other. In a granular mixture
containing two or more types of particles with different diameters, the
dispersive pressure felt by any particle depends on the contribution of all
the particles, and may possibly be expressed by Equation 1 with a mean
value for D, but it does not vary with the diameter of the particle considered.
If we are concerned with the segregation of particles, we must actually
look at the forces acting upon them. The force acting upon one side of a
particle is equal to the pressure exerted multiplied by the area upon which
it is exerted, and thus it is proportional to the square of the particle diameter. For a particle to move upwards, the force acting upon its bottom
side must be greater than that acting upon its top. This requires the dispersive pressure to be less at the top than at the bottom of the particle. The
difference between dispersive pressure at the top and the bottom of the
particle is equal to the vertical gradient of dispersive pressure multiplied
by the diameter of the particle. Therefore, the upward force exerted upon
a particle by granular dispersive pressure is equal to the product of the
dispersive pressure gradient and the volume of the particle, and so it is

CAN DISPERSIVE PRESSURE CAUSE INVRESE GRADING?


proportional to Di3. Thus, larger particles experience greater dispersive
pressure forces.
In order to drive a particle upward, the dispersive pressure force must
exceed the particle weight, which is also proportional to Di3. The ratio of
the weight of a particle to the dispersive pressure force acting upon it is
therefore independent of the particle diameter. Because of this, one might
infer that dispersive pressure cannot cause size segregation, but this is not
true because the important variable is not the ratio but the resultant of the
two forces, i.e., the difference between the weight and the dispersive pressure force, which is also proportional to Di3. This case is similar to that of
particles suspended in a fluid, for which both the weight and the upward
Archimedes buoyancy force are proportional to Di3. The resultant of these
two forces is therefore also proportional to Di3, and size segregation can
occur because the resultant force that drives a large particle is greater than
that which drives a small particle. Whether the resultant force will drive
particles upward or downward depends on particle density and the pressure
gradient in the fluid. Similarly, in a granular flow, the absolute value of
the resultant force of the dispersive pressure force and the weight of a
particle is proportional to Di3, hence is always greater for larger particles.
This resultant force is directed upward when the gradient of dispersive
pressure is greater than the product of the particle density and the acceleration due to gravity (sig) and downward in the other case. The development of an inverse grading due to dispersive pressure therefore requires
that the vertical gradient of dispersive pressure be higher than the product
sig.
As shown by Equation 1, the vertical gradient of dispersive pressure
depends on the vertical gradients of velocity and concentration. These are
still poorly known in grain flows and may vary according to the type of
flow and boundary conditions considered. For example, by assuming a
constant particle concentration throughout the bed, Bagnold (1954), Lowe
(1976), and Takahashi (1980) theoretically deduced a concave-up velocity
profile for free-surface grain flows on inclines. Such concentration and
velocity profiles would imply an upward-decreasing dispersive pressure,
hence an upward-directed dispersive pressure force. Some numerical simulations of collisional grain flows show a nonshearing plug zone overlying
an expanded basal shear zone (Cleary and Campbell 1993; Straub 1996,
1997). A basal shear zone under a rigid plug has also often been postulated
for pyroclastic and debris flows on the assumption that such flows behave
macroscopically like a Bingham fluid, and this has been thought to be the
reason for the inversely graded basal layers commonly observed in their
deposits (e.g., Sparks 1976). On the other hand, some experimental grain
flows have roughly linear velocity profiles and upward-decreasing concentration (Hungr and Morgenstern 1984; Drake 1990). These trends are also
observed in some numerical simulations (Campbell et al. 1995) and theoretical analyses (Mills et al. 1999). For traction carpets, the situation is
different, because the energy is provided to the grain flow by the overlying
current. Hiscott (1994) has hypothesized that the velocity has a linear profile, as with a granular material sheared between two plates (Savage and
Sayed 1984), whereas a convex-up velocity profile with a linearly upwarddecreasing concentration was postulated by Sohn (1997) on the basis of
the theoretical work by Hanes and Bowen (1985) and the numerical simulations by Jiang (1995).
Because of these uncertainties on the vertical profiles of velocity and
concentration, it might seem impossible to determine with some confidence
the vertical profile of dispersive pressure through a shearing granular bed.
The solution to this apparently complex problem is, however, surprisingly
simple. The dispersive pressure must always equal the applied normal stress
at every level through the shearing granular bed. This is because sheared
granular materials in the collisional regime develop a dispersive pressure
that is a positive function of particle concentration (Bagnold 1954; Hungr
and Morgenstern 1984; Savage and Sayed 1984; Hanes and Inman 1985).
If the dispersive pressure increases and becomes greater than the applied
normal stress, there is a net upward force acting upon the granular bed. In

167

response, the bed expands, as observed in experiments, causing a decrease


in particle concentration and, consequently, in dispersive pressure, until the
dispersive pressure is just able to balance the applied normal stress and no
more expansion can occur. If the dispersive pressure becomes less than the
applied normal stress, the reverse process occurs, with a net downward
force acting upon the granular bed and compacting it, thus increasing particle concentration and dispersive pressure. Therefore, in the collisional
regime, the dispersive pressure is always forced to equal the applied normal
stress at any level through the granular bed. The applied normal stress
increases with depth in the granular bed owing to the weight of the overburden, so the vertical gradient of dispersive pressure is simply equal to
the static pressure gradient within the bed.
The situation can therefore be compared with that of a fluid suspension,
where the hydrostatic pressure gradient imparts an upward-directed Archimedes buoyancy force to particles. Particles tend to sink if their weight is
greater than the Archimedes force, which occurs when particle density is
higher than fluid density. Similarly, in a shearing, dry granular bed, the
gradient of dispersive pressure is equal to the static pressure gradient, which
is itself equal to the product of the bulk density of the bed and the acceleration due to gravity (rbg), so particle weight is overcome by the upwarddirected dispersive pressure force only for light particles, with density lower than the bulk density of the bed. If all particles have the same density,
the bulk density of the bed is always less than the density of an individual
particle, so the weight always overcomes the dispersive pressure force and
particles tend to sink. Because the resultant between weight and dispersive
pressure force is proportional to Di3, the tendency to sink is stronger for
large particles than for small ones. Dispersive pressure in the collisional
regime can therefore not explain the upward migration of large, heavy
particles. A summary of the demonstration is given in the appendix.
Natural flows are generally not pure grain flows, and the relative importance of dispersive stress compared to other stresses in such flows is
difficult to assess. When a fluid is present between particles, it supports
part of their weight. In order to push a particle upwards the gradient of
dispersive pressure should exceed (sirf)g, where rf is the density of the
fluid (which, in practice, may comprise water plus fines in natural flows).
However, the gradient of dispersive pressure that balances the static, buoyancy-reduced granular load is (rbrf)g, so particles still need to be lighter
than the bulk density of the flow in order to become segregated upwards.
The gradient of dispersive pressure also decreases when part of the static
granular load is supported by frictional rather than collisional particle contacts. The presence of an interstitial fluid may also drive a rapid granular
flow in the macroviscous regime if the viscous stress due to the fluid dominates the stress due to grain collisions (Bagnold 1954). Bagnold showed
that, in this regime, collisions between particles still transmit a dispersive
pressure that is positively correlated to the particle concentration. The reasoning applied above to the collisional regime is therefore also valid in the
macroviscous regime. As soon as the dispersive pressure becomes greater
than the normal static stress due to the immersed weight of particles, the
granular bed expands, thus reducing particle concentration and dispersive
pressure, until dispersive pressure equals normal static stress. In such conditions, the upward-directed dispersive pressure force acting upon a particle
cannot overcome its immersed weight, unless the particle density is less
than the bulk density of the particle-fluid mixture. Dispersive pressure in
the macroviscous regime can therefore not explain the upward migration
of large, heavy particles.
DISCUSSION

As shown in the preceding section, the fact that s and D 2 appear in the
equation of the dispersive pressure (Eq. 1) does not mean that the dispersive
pressure force acting upon a particle is proportional to its density and to
the square of its diameter. On the basis of this erroneous idea, Sallenger
(1979) developed the concept of dispersive equivalence. According to

168

F. LEGROS

Sallenger, two particles are in dispersive equivalence, i.e., they behave in


a collisional granular flow as if they were identical, if the products of their
respective density and squared diameter (siDi 2) are equal. Exploiting this
concept, some sedimentological studies showed that light and heavy particles were approximately in dispersive equivalence in various inversely
graded deposits (Sallenger 1979; DeCelles and Cavazza 1992; Cole and
Stanley 1994). This means that, at any level in these deposits, light particles
were on average larger than dense ones by a factor equal to the square root
of their density ratio. It was concluded that the inverse grading was caused
by dispersive pressure. These results and conclusions are further examined
below.
DeCelles and Cavazza (1992) studied thin, inversely graded laminae in
nearshore-marine sequences. They measured the size of micas (density ;
2850 kgm23) and quartzofeldspathic grains (density ; 2650 kgm23) at
various levels in the laminae and showed that the two species were within
15% of dispersive equivalence. Because the density ratio between micas
and quartzofeldspathic grains is only 1.08, however, the size ratio expected
for dispersive equivalence is only 1.04, i.e., micas are expected to be 4%
smaller than quartzofeldspathic grains at the same level. Given the large
scatter in the data (15%), the claim of DeCelles and Cavazza (1992) that
micas and quartzofeldspathic grains are in dispersive equivalence is not
substantiated. Their results just indicate that micas and quartzofeldspathic
grains are roughly of the same size at any level in the deposit. In addition,
the significant shape difference between the two types of grains may also
have played a role in the segregation process.
Cole and Stanley (1994) studied deposits of subaqueous volcaniclastic
sediment gravity flows, in which they calculated that vitric shards were in
dispersive equivalence, rather than settling equivalence, with pumice
grains. In their calculations, they used an arbitrary density of 1010 kgm23
for pumice grains. However, pumice-grain density may vary significantly,
in particular with size and, in a subaqueous deposit, with the amount of
absorbed water. If a density of, say, 2000 kgm23 had been assumed for
the water-logged pumice grains, the size ratio predicted between vitric
shards and pumice grains would have been much different for both settling
and dispersive equivalence. Furthermore, there is a continuum between
vitric shards and pumice grains. Vitric shards are on average necessarily
smaller than pumice grains, not because of segregation processes but by
definition, inasmuch as vitric shards are fragments of bubble wall and so
must be smaller than bubbles, whereas pumice grains contain bubbles and
so must be larger. The claimed dispersive equivalence of Cole and Stanley
(1994) is therefore unconvincing.
Sallenger (1979) measured the size of heavy grains (mean density ;
3200 kgm23) and light grains (density ; 2650 kgm23) at various levels
in both a beach foreshore lamina and a frozen, subaerial grain flow, experimentally generated from the same beach material. Both deposits were
inversely graded, with the coarse, light particles concentrated at the top
and the small, heavy ones concentrated at the base. With a density ratio of
1.21, heavy grains should be 10% smaller than light grains found at the
same level for dispersive equivalence to be satisfied. The 400 grains of
each type measured by Sallenger at various levels in the deposits give fairly
accurate data that show that heavy grains are indeed significantly smaller
than light grains at every level, but on average by 13% and in some instances by up to 18%, instead of 10%. This discrepancy was hypothesized
by Sallenger to result from shape differences between heavy and light
grains. This seems a reasonable hypothesis, but if it is admitted that shape
effects can account for a size difference of up to 8%, they may also account
for the whole size difference observed between light and heavy grains.
Another problem with Sallengers attempted demonstration of dispersive
equivalence arises from the fact that the overall populations of heavy and
light grains in the deposit did not have the same mean size. The mean sizes
of the overall populations are not given by Sallenger but, from the data
presented, it seems that the mean size for heavy grains is about 1.5 times
less than that for light grains, which means that the two grain populations

are in hydraulic settling equivalence. The mean sizes of heavy and light
grains segregated at any level must be influenced by the overall size distribution of each population in the deposit. Consider, for example, the top
level, enriched in coarse grains. If we hypothesize that grains are segregated
according to size only, and that density has no effect, the top level should
be composed almost entirely of light grains, because the coarsest heavy
grains available in the deposit are smaller than the coarsest available light
grains. Of course, all particles segregated in the top level cannot have
exactly the same size. If grains as small as the coarsest heavy grains available in the deposit are present in the top level, then some of them should
be heavy grains. Thus, the mean size of heavy grains in the top level should
be less than that of light grains, although density has been assumed to play
no role. This remains true at every level in the deposit because heavy grains
always have a greater amount of small grains and a lower amount of large
grains available in the overall population from which they are segregated
than light grains. This effect is probably able to explain most of the size
differences observed by Sallenger.
In the three examples above, dispersive equivalence could not be properly tested because heavy and light grains had different shapes, different
sizes, or variable densities. A simple way to avoid these problems is to
generate experimental granular flows with particles of different densities
but same size and shape. According to Sallenger (1979), we should expect
the heavy particles to become segregated to the top. Alonso et al. (1991)
carried out such experiments and observed that the heavy particles tended
to sink. These authors also showed that, in order to compensate for the
floating effect of a greater diameter, a particle must have a higher density, just the opposite of what is expected from the hypothesis of dispersive
equivalence.
Takahashi (1980) calculated theoretically the velocity of upward migration of large particles due to dispersive pressure forces in a granular flow
and compared it with experimental measurements. In order to match the
theoretical and observed velocities, he had to adjust freely the friction term
in the equation of momentum. Furthermore, he assumed an unrealistic vertically constant particle concentration in the flow. As discussed previously,
a dispersive pressure force sufficient to overcome the weight of a heavy
particle implies a dispersive pressure gradient stronger than the static pressure gradient of the bed, hence bed expansion and decrease of the gradient
of dispersive pressure. The upward migration of large particles in a granular
flow is better explained, and even quantitatively modeled, by the process
of kinetic sieving (Savage and Lun 1988). This process, by which large
particles accumulate at the top owing to the easier percolation of small
ones, seems to dominate particle segregation as long as the size ratio is not
too large (Savage and Lun 1988; Cantelaube and Bideau 1995; Pouliquen
et al. 1997; Thomas 2000). For size ratios greater than 5, Thomas (2000)
shows that, in experimental granular flows containing two types of particles
of different size but same density, large particles are able to force their
way down by squeezing out small particles. Depending on the size ratio
and the relative abundance of the two species, large particles can accumulate at the top, the base, or even the middle of the flow, a behavior
difficult to explain if segregation is due to dispersive pressure forces.
The segregation processes in granular mixtures containing a wide range
of particle sizes are not easy to infer from experiments that use only two
particle sizes (Thomas 2000). Furthermore, many pyroclastic-flow, debrisflow, and turbidity-current deposits are inversely graded only at their bases,
the rest of the deposit being ungraded or normally graded, which may be
difficult to explain by a simple segregation process. One hypothesis is that
the deposit is not the frozen image of the flow, but that it forms by progressive aggradation of particles at the base of a steady or unsteady flow
(Branney and Kokelaar 1992). In this model, the material deposited from
the leading part of the flow, which will form the basal layer of the deposit,
may contain less coarse particles because they are mainly transported in
the slow, basal part of the flow and thus lag behind the flow head (Hand
1997; Legros and Mart 2001). Similarly, in debris-flow deposits, it has

CAN DISPERSIVE PRESSURE CAUSE INVRESE GRADING?


been proposed that basal inverse grading could be the result of the lower
competence of the leading part of the flow, which mixes with river water
(Vallance 2000). Basal inverse grading may also reflect an increase in flow
competence due to the temporal increase of discharge rate of a waxing flow
(Allen 1981; Kneller and Branney 1995).
Naylor (1980) proposed another mechanism for the formation of inverse
grading at the base of debris-flow deposits. He assumed that the high shear
rate likely to occur in the basal part of the flow reduces the strength of the
muddy material enough to allow the coarse particles to settle out from the
basal layer. Naylor suggested that these settled clasts would form lithic lag
breccias upslope of the inversely graded deposit. Because the upper part
of the flow is faster than the basal part, however, the basal layer is continuously formed at the flow head with material from the upper part of the
flow that still contains coarse particles. Since coarse particles are postulated
to settle once they enter the basal layer, there should be a continuous
coarse-particle breccia all along the base of the final deposit, instead of the
inversely graded basal layer that this model is supposed to explain. The
mechanism proposed by Naylor would require that the basal layer, which
has dropped its coarse particles, progress downstream at the same velocity
as the rest of the flow, in contradiction with the assumption of basal shearing on which the model is based.
CONCLUSIONS

The hypothesis that inverse grading in deposits of grain flows or modified grain flows can develop because of dispersive pressure is not correct.
The gradient of dispersive pressure necessary to generate inverse grading
cannot be maintained because it causes an immediate expansion of the grain
flow and hence a decrease in particle concentration and dispersive pressure.
In this manner, the gradient of dispersive pressure is forced to be equal to
the static pressure gradient in the flow, and only particles with density
lower than the bulk density of the flow can be pushed upward. The concept
of dispersive equivalence used in sedimentology is flawed, as shown by
the present analysis and by experiments that use particles of same size and
different densities (Alonso et al. 1991). Other plausible explanations for
inversely graded deposits include size segregation by the kinetic sieving
process or the formation of the deposit by progressive aggradation with a
temporal increase in the size of particles supplied from the current to the
deposit.
ACKNOWLEDGMENTS

Mike Branney and Gerry Middleton provided constructive and helpful reviews.
The author is supported by a Pierre and Marie Curie grant (ERBFMBICT983445).
REFERENCES

ALLEN, P.A., 1981, Sediments and processes on a small alluvial stream-flow dominated Devonian alluvial fan, Shetland Islands: Sedimentary Geology, v. 29, p. 3166.
ALONSO, M., SATOH, M., AND MIYANAMI, K., 1991, Optimum combination of size ratio, density
ratio and concentration to minimize free surface segregation: Powder Technology, v. 68, p.
145152.
BAGNOLD, R.A., 1954, Experiments on a gravity-free dispersion of large solid spheres in a
Newtonian fluid under shear: London, Royal Society, Proceedings, v. 225, p. 4963.
BEST, J.L., 1992, Sedimentology and event timing of a catastrophic volcaniclastic mass flow,
Volcan Hudson, Southern Chile: Bulletin of Volcanology, v. 54, p. 299318.
BRANNEY, M.J., AND KOKELAAR, P., 1992, A reappraisal of ignimbrite emplacement: progressive
aggradation and changes from particulate to non-particulate flow during emplacement of
high-grade ignimbrite: Bulletin of Volcanology, v. 54, p. 504520.
CAMPBELL, C.S., CLEARY, P.W., AND HOPKINS, M., 1995, Large-scale landslide simulations: global deformation, velocities and basal friction: Journal of Geophysical Research, v. 100, p.
82678273.
CANTELAUBE, F., AND BIDEAU, D., 1995, Radial segregation in a 2D drum: an experimental
analysis: Europhysics Letters, v. 30, p. 133138.
CLEARY, P.W., AND CAMPBELL, C.S., 1993, Self-lubrication for long-runout landslides: examination by computer simulations: Journal of Geophysical Research, v. 98, p. 21,91121,924.
COLE, P.D., AND SCARPATI, C., 1993, A facies interpretation of the eruption and emplacement
mechanisms of the upper part of the Neapolitan Yellow Tuff, Campi Flegrei, southern Italy:
Bulletin of Volcanology, v. 55, p. 311326.

169

COLE, R.B., AND STANLEY, R.G., 1994, Sedimentology of subaqueous volcaniclastic sediment
gravity flows in the Neogene Santa Maria Basin, California: Sedimentology, v. 41, p. 37
54.
CRUDEN, D.M., AND HUNGR, O., 1986, The debris of the Frank Slide and theories of rockslide
avalanche mobility: Canadian Journal of Earth Sciences, v. 23, p. 425432.
DECELLES, P.G., AND CAVAZZA, W., 1992, Constraints on the formation of Pliocene hummocky
cross-stratification in Calabria (southern Italy) from consideration of hydraulic and dispersive
equivalence, grain-flow theory, and suspended-load fallout rate: Journal of Sedimentary Petrology, v. 62, p. 555568.
DRAKE, T.G., 1990, Structural features in granular flows: Journal of Geophysical Research, v.
95, p. 86818696.
DRUITT, T.H., 1998, Pyroclastic density currents, in Gilbert, J.S., and Sparks, R.S.J., eds., The
Physics of Explosive Volcanic Eruptions: Geological Society of London, Special Publication, p. 145182.
GIANETTI, B., 1998, Geology of proximal, small-volume trachytetrachyandesite pyroclastic
flows and associated surge deposits, Roccamonfina volcano, Italy: Journal of Volcanology
and Geothermal Research, v. 80, p. 113136.
HAND, B., 1997, Inverse grading resulting from coarse-sediment transport lag: Journal of Sedimentary Research, v. 67, p. 124129.
HANES, D.M., AND BOWEN, A.J., 1985, A granular-fluid model for steady intense bed-load
transport: Journal of Geophysical Research, v. 90, p. 91499158.
HANES, D.M., AND INMAN, D.L., 1985, Observations of rapidly flowing granular-fluid materials:
Journal of Fluid Mechanics, v. 150, p. 357380.
HISCOTT, R.N., 1994, Traction-carpet stratification in turbidites: fact or fiction?: Journal of
Sedimentary Research, v. A64, p. 204208.
HISCOTT, R.N., AND MIDDLETON, G.V., 1980, Fabric of coarse deep-water sandstones, Tourelle
Formation, Quebec, Canada: Journal of Sedimentary Petrology, v. 50, p. 703722.
HUNGR, O., AND MORGENSTERN, N.R., 1984, Experiments on the flow behaviour of granular
materials at high velocity in an open channel: Geotechnique, v. 34, p. 405413.
HUGHES, S.R., AND DRUITT, T.H., 1998, Particle fabric in a small, type-2 ignimbrite flow unit
(Laacher See, Germany) and implications for emplacement dynamics: Bulletin of Volcanology, v. 60, p. 125136.
JIANG, Z., 1995, The motion of sedimentwater mixtures during intense bedload transport:
computer simulations: Sedimentology, v. 42, p. 935945.
KNELLER, B.C., AND BRANNEY, M.J., 1995, Sustained high-density turbidity currents and the
deposition of thick massive sands: Sedimentology, v. 42, p. 607616.
LEGROS, F., AND MART, J., 2002, Formation of inversely graded basal layers in ignimbrites by
progressive aggradation: Journal of Volcanology and Geothermal Research, v. 111/14, p.
2533.
LOWE, D.R., 1976, Grain flow and grain-flow deposits: Journal of Sedimentary Petrology, v.
46, p. 188199.
LOWE, D.R., 1982, Sediment gravity flows: II. Depositional models with special reference to
the deposits of high-density turbidity currents: Journal of Sedimentary Petrology, v. 52, p.
279297.
MASKE, H.A., HAVLIN, S., KING, P.R., AND EUGENE STANLEY, H., 1997, Spontaneous stratification
in granular mixtures: Nature, v. 386, p. 379382.
MIDDLETON, G.V., 1970, Experimental studies related to problems of flysch sedimentation, in
Lajoie, J., ed., Flysch Sedimentology in North America: Geological Association of Canada,
Special Paper 7, p. 253272.
MILLS, P., LOGGIA, D., AND TIXIER, M., 1999, Model for a stationary dense granular flow along
an inclined wall: Europhysics Letters, v. 45, p. 733738.
NAYLOR, M.A., 1980, The origin of inverse grading in muddy debris flow deposits: a review:
Journal of Sedimentary Petrology, v. 50, p. 11111116.
PALLADINO, D.M., AND VALENTINE, G.A., 1995, Coarse-tail vertical and lateral grading in pyroclastic flow deposits at the Latera Volcanic Complex (Vulsini, Central Italy): Origin and
implications for flow dynamics: Journal of Volcanology and Geothermal Research, v. 69,
p. 343364.
POULIQUEN, O., DELOUR, J., AND SAVAGE, S., 1997, Fingering in granular flows: Nature, v. 386,
p. 816817.
SALLENGER, A.H., 1979, Inverse grading and hydraulic equivalence in grain-flow deposits: Journal of Sedimentary Petrology, v. 49, p. 553562.
SAVAGE, S.B., AND LUN, K.K., 1988, Particle size segregation in inclined chute flow of dry
cohesionless granular solids: Journal of Fluid Mechanics, v. 189, p. 311335.
SAVAGE, S.B., AND SAYED, M., 1984, Stresses developed by dry cohesionless granular materials
sheared in an annular shear cell: Journal of Fluid Mechanics, v. 142, p. 391430.
SCHMINCKE, H.-U., 1967, Graded lahars in the type section of the Ellensburg Formation, southcentral Washington: Journal of Sedimentary Petrology, v. 37, p. 438448.
SIGURDSSON, H., CAREY, S.N., AND FISHER, R.V., 1987, The 1982 eruptions of El Chichon volcano, Mexico (3): physical properties of pyroclastic surges: Bulletin of Volcanology, v. 49,
p. 467488.
SOHN, Y.K., 1997, On traction-carpet sedimentation: Journal of Sedimentary Research, v. 67,
p. 502509.
SOHN, Y.K., RHEE, C.W., AND KIM, C.B., 1999, Debris flow and hyperconcentrated flood-flow
deposits in an alluvial fan, northwestern part of the Cretaceous Yongdong basin, Central
Korea: The Journal of Geology, v. 107, p. 111132.
SPARKS, R.S.J., 1976, Grainsize variations in ignimbrites and implications for the transport of
pyroclastic flows: Sedimentology, v. 23, p. 147188.
STRAUB, S., 1996, Self-organisation in the rapid flow of granular material: evidence for a major
flow mechanism: Geologische Rundschau, v. 85, p. 8591.
STRAUB, S., 1997, Predictability of long runout landslide motion: implications from granular
flow mechanics: Geologische Rundschau, v. 86, p. 415425.
TAKAHASHI, T., 1980, Debris flow on prismatic open channel: ASCE, Proceedings, Journal of
the Hydraulics Division, v. 106, p. 381396.

170

F. LEGROS

THOMAS, N., 2000, Reverse and intermediate segregation of large beads in dry granular media:
Physical Review E, v. 62, p. 961974.
TODD, S.M., 1989, Stream-driven, high-density gravelly traction carpets: possible deposits in
the Trabeg Conglomerate Formation, SW Ireland and some theoretical considerations of
their origin: Sedimentology, v. 36, p. 513530.
VALLANCE, J.W., 2000, Lahars, in Sigurdsson, H., ed., Encyclopedia of Volcanoes: Academic
Press, p. 601616.
WALTON, A.W., AND PALMER, B.A., 1988, Lahar facies of the Mount Dutton Formation (OligoceneMiocene) in the Marysvale Volcanic Field, southwestern Utah: Geological Society
of America, Bulletin, v. 100, p. 10781091.
Received 25 July 2000; accepted 13 April 2001.

(A2)

The total dispersive pressure force acting upon the particle, F, is therefore
F 5 Fz 2 Fz1D 5 (Pz 2 Pz1D )D 2 5

dP 3
dP
D 5 V
dz
dz

(A3)

where V is the particle volume. F is directed upward if dispersive pressure increases


with depth in the granular flow. In order to drive the particle upward, F must
overcome the weight of the particle, W, given by
W 5 sVg

(A4)

where g is the acceleration due to gravity. Because the dispersive pressure is always
forced to be equal to the static pressure at any level in the flow, we can write

APPENDIX

Consider a particle of dimension D and density s in a granular flow. For simplicity, we consider here that the particle is cubic and that its lower and upper faces
are perpendicular to the dispersive pressure gradient. The lower face is at a level z
and undergoes an upward force (Fz) caused by the dispersive pressure at that level
(Pz), equal to
Fz 5 PzD 2

Fz1D 5 Pz1DD 2

(A1)

The upper face, at a level z 1 D, undergoes a downward force (Fz1D ) caused by


the dispersive pressure at that level (Pz1D ), equal to

dP
5 rb g
dz

(A5)

where rb is the bulk density of the flow. The resultant force acting upon the particle,
R, is therefore equal to
R 5 F 2 W 5 (r b 2 s)gV

(A6)

We see that R is positive, hence directed upward, only for s , r b, i.e., for particles
with a density lower than the bulk density of the flow.

You might also like