ActaMechanica 63, 161--178 (1986) ACTA MECHANICA

| by Springer-Verlag 1986

Large Debris Flows: A Macro-Viscous Phenomenon

By

T. R. H. Davies, Christchurch, New Zealand

With 5 Figures

(Received October 8, 1985; revised November 26, 1985)

Summary
Field observations from a variety of sources suggest that destructive debris flows
occur when the density of the fluid-solid mixture exceeds about 1.5 T/m a, and that their
destructive ability is due to their pulsing nature and to their ability to carry large boulders.
If debris flows are treated as a macroviscous flow of large stones in a slurry of fine
solids in water, several of their obvious characteristics (boulder transport, deep bed erosion,
intermittent jamming) can be explained. Further, the amplification and translation in
a main channel of random surges due to jamming in tributaries explains the regular,
large pulses in Chinese debris flows as a roll-wave phenomenon.

Introduction
I n m a n y m o u n t a i n o u s areas the greatest threat to industrial or agricultural
developments is posed b y debris flows or m u d flows. These occur when, following
intense rainfall, river flows become transformed into a dense slurry of water,
fine solids, rocks and boulders, which can far exceed the channel c a p a c i t y and
d e v a s t a t e the land in the vicinity of the river. Such events p r o b a b l y occur in-
l r e q u e n t l y in m a n y m o u n t a i n valleys; in some regions, however, such as parts
of J a p a n a n d China, debris flows occur relatively often, and I k e y a [24], for
example, states t h a t in 1967 and 1972 respectively, 107 and 194 J a p a n e s e lives
were claimed by debris flows. The practical importance of further a t t e m p t s
to u n d e r s t a n d the mechanics of debris flows is emphasised by, for example,
[54], [18], [31], [12].
A high p r o p o r t i o n of the studies designed to clarify the w a y in which debris
flows occur deals only with p h e n o m e n a occurring on a relatively small scale.
However, the p h e n o m e n o n which directly threatens valleyfloor development
is n o t so m u c h the small debris-slide which moves down a steep hillside as a result
of the regolith becoming overloaded with water; it is r a t h e r the occurrence
of waves of m u d and boulders which burst forth from canyons, sometimes several

11 Acta Mech. 63/1--4

 162 T . R . H . Davies:

kilometres from the debris source, and devastate the valley floor beyond [43],
[33]. This study, therefore, concentrates on the characteristics and mechanics
of the larger and more destructive phenomenon experienced farther down the
channel system.
Current methods for assessing the area of land threatened by debris flows
rely largely on empirical data and on surveys of pre-existing deposits [25], al-
though some techniques make limited use of analytical and experimental results
[51], [50]. I t is clear, however, that the degree of understanding involved is
quite limited, and there is no implied explanation of the two most destructive
characteristics of debris flows; the ability of the flow to carry large boulders
and the occurrence in the flow of large waves of debris, sometimes several
metres high.
The present discussion proposes a theoretical framework for debris flow
behaviour which can explain these and other characteristics. The behaviour
of large debris flows is first outlined, to collate the flow behavioural and other
characteristics. Existing explanations of these characteristics are then discussed,
and their deficiencies outlined. A hypothesis is proposed which overcomes these
deficiencies, and its ability to explain debris flow occurrence and behaviour is
demonstrated.

Debris Flow Behaviour

Field Observations

B y the nature of the phenomenon, most field reports of debris flows result
from the observer, unprepared for the event, finding himself in a position to
record it. Notable exceptions are the work of the Japanese field study centre
at Mt. Yakedake [40], [46], the Chinese study programme at Jiangjia Ravine,
Chengdu [30], [33] and the observations of small-scale processes in Pennsylvania
reported by Johnson and l~ahn [28]. Pierson [41], [42] was able to study and
film a 3-day-long debris flow event close to Christchurch, New Zealand. In
these studies reliable data were obtained describing flow characteristics, compo-
sition and occurrence. The more adventitious reports, e.g. [13], [9], [27], [52], [39]
are necessarily less precise in their descriptions. All these reports, however, contri-
bute to a reasonably consistent picture of debris flow characteristics. This picture
is summarized below.

Occurrence

Not all rainstorms give rise to debris flows, even in regions where flows are
common. Debris flows are most often observed immediately following bursts
of intense rain during the course of long-duration storms [40], [41], [24], [39],
[13]. Intense rain, falling onto a previously saturated landscape with already
high river flows, triggers the phenomenon. Exceptions are the catastrophic
mudflows reported by [39] which originated in the collapse of a glacial moraine
 Large Debris Flows: A Macro-Viscous Phenomenon 163

and release of impounded lake water; while Sharp and Nobles [45] reported
snowmelt-initiated debris flows. ;~
A large volume of available sediment is also neeessary for debris-flow occur=
rence. Many flows are associated with deep beds of widely-graded alluvial material,
fed b y tributary streams flowing in steep, rapidly eroding ravines; consequently,
during heavy rain the sediment concentration in streams is high. Availability of
sediment in the channel and hillslopes thus influences whether or not a particular
rainstorm will give rise to a debris flow.

Flow Character

Most reports of large debris flows emphasise the unsteady, "pulsing" nature
of the flow [41], [27], [39], [33]; periods of relatively low flow rate, or zero flow
[33], separate the passage of large (up to several metres high) waves of heavily
debris-laden slurry. Kang and Zhang [30] and Li and Luo [32], however, noted
that some debris flows were continuous, or non-pulsing, in nature; while the
events reported b y Okuda et al. [40] apparently consisted of only a single wave
or " h e a d " followed by a declining continuous flow. The moraine-burst mudflows
reported b y Niyazov and Degovets [39], although caused b y an event which
initially generated a single wave, were recorded at downstream stations as com-
prising several distinct waves.
The composition and motion of the flow in continuous or non-pulsing events,
and in the between-pulse phases of pulsing events, differ markedly from those
of the pulses themselves. Table 1 lists these differences. The debris-flow pulses
are waves of high-density, high-viscosity, solid-fluid slurry of laminar appearance,
with both fine and coarse solids distributed throughout the flow; they move faster
than the intervening flow, and are strongly erosive, causing deep scour of the

Table 2. Comparison o/pulses with between-pulse/low in debris/lows

Characteristic Pulse Non-pulsing or l~eferences

between-pulse

Flow density High Low; [41], [30], 33]

> 2.1 T/m3 < 2.6 T/m 3
Sizes present above bed Bimodal; Unimodal; [30], [33]
fines and gravel fines dominant
Location of coarse load Throughout Bedload only [42], [32]
flow depth
Appearance Laminar Turbulent [42], [27]
Viscosity High Lower [423
Velocity High; ,~ 3--10 m/s Low; ~ 1 2 m/s [41], [39]
Effect Strongly erosive Depositional [42], [33]

11"
164 T. 1~. It. Davies:

channel bed. Separating the pulses are intervals of slower, lower-density, lower-
viscosity flow of turbulent appearance with a high concentration of fines in
suspension and coarser sizes moving as bedload, associated with bed aggradation.
I t should be noted that flow m a y be laminar in the hydraulic sense -- i.e. domi-
nated b y viscous rather than inertial forces [35] -- even though it is highly
disturbed b y surface waves and boundary roughness. The criterion for laminar
flow is not the absence of turbulence, but the rapidity with which an introduced
turbulent eddy is damped out.

Previous Explanations of Debris-Flow Behaviour

The most significant characteristics of large debris flows are their pulsing
nature and their ability to carry large particles at the surface of, or within, the
flow which do not sink to the bed. Previous attempts to explain these phenomena
are now discussed.
Pulsing Flow
The intermittent nature of large debris flows is also seen in small debris flows
and debris slides. Observation of the latter has shown [4=1], [27], [9], [45] that a
pulse can be generated b y the entry of a debris slide to a stream channel, forming
a temporary dam; subsequent mobilisation of this material due to rising water
level results in a surge-wave travelling downstream. I t seems unlikely, however,
that this mechanism can explain pulsing in large debris flows, due to the volume
of material which would be needed to dam a large channel. Nor does it seem
likely that entry of realistically-sized debris slides to the channel would occur in
such a time sequence that sufficiently large waves could be generated b y super-
position of smaller ones. The pulsing of large debris flows seems more likely to be
caused b y some intrinsic instability of slurry flow in an open channel; certainly
Niyazov and Degovets [39] are of the opinion that mudflow pulsing results from
channel processes, and their report of several large pulses resulting from a single
initiating event supports this contention, as does the identification by Li et al. [33]
of critical values of channel flow and fluid density beyond which pulsing flow
occurs.
An alternative explanation of pulsing was applied to the Jiangjia Ravine
flows b y Li et al. [33], having also been observed in laboratory flmnes [30], [17].
I t was proposed that an initially continuous flow halts because of its shear strength
if the flow depth becomes less than some critical value, and will only be remobilised
when sufficient material has accumulated behind the stationary front. The remobi-
ised front then moves downstream, and the flow depth some way behind it
again decreases to the critical value, causing flow to cease again. Quoted values of
slurry shear strength [33], however, suggest that only if large grains in the flow
can interact (interlock) will the shear strength be sufficient to stop a flow, and
then only in narrow headwater gullies where sideways interlocking can help the
grain mass to jam [2].
 Large Debris Flows: A Macro-Viscous Phenomenon 165

Support o] Large Grains

Three mechanisms have been proposed to explain the ability of debris flow to
carry large rocks which are not in direct contact with the stationary bed of the flow :
i) the cohesive strength of the fine material in the flow;
ii) buoyancy due to excess pore-water pressure within the body of the flow;
iii) dispersive pressure caused by intergranular contact stresses resulting from
grains shearing past each other.
These mechanisms are now examined.

Cohesive Strength
It has been shown that stationary debris flow material has substantial strength
(that is, it can resist an applied stress without flowing) due to the cohesion of the
clay particles present [42], [27], [33] and can support large grains. The excess
weight of the grain is transferred to the ground below through the rigid structure
of the stationary material. When the material is flowing, however, then by
definition its strength has somewhere been exceeded, and a continuous rigid
matrix no longer exists to transfer the grain's Weight to the ground. If general
shearing is taking place within the flow, then no part of the flow can possess
strength, and the imposition of an additional stress, such as placing a boulder in
the flow, will result only in changes in the rate of shear; a grain of excess density
cannot be continuously supported by the flow material. Even if part of the flow
is non-shearing, forming a rigid plug that moves with the flow [27], this plug and
any boulders it contains cannot be supported by any strength of the flowing
material between it and the bed. A flowing fluid cannot simultaneously" act as a
solid. Only if all the material forms a rigid plug, which is sliding along the under-
lying ground without a n y deformation, can the strength of the material support
solids of excess density.
Hampton [20] apparently demonstrated the ability of a flowing Koalin-in-
water slurry to support larger heavy grains in the absence of both suspension by
turbulence and dispersive stresses between the larger grains. The volume concen-
tration of the larger grains was about 10~o, however, so the claim that dispersive
stresses were avoided eannot be sustained; Bagnold [2] states that when grain
concentrations reach about 9%, grain collisions become "a certainty", and
dispersive stresses will therefore be present. It seems likely that the "competence"
of a flowing slurry to support heavy grains demonstrated by Hampton [20], [38]
resulted from dispersive pressure between these grains rather than from any
strength of the slurry (Hampton [21, p. 754]).
Casual experiment with slurry material shows that, when it is stationary,
heavy stones will rest on the surface without sinking; as soon as the body of the
slurry is continuously disturbed, however, the stones will sink to the bottom.
It is concluded that the concept of grain support by the cohesive strength of
flowing debris is unrealistic.
166 T . R . H . Davies:

Buoyancy
I-Iampton [21] and Pierson [42] both envisaged that unexpectedly dense grains
could be supported b y buoyancy forces if the pressure of pore fluid within the
flowing material increased with depth at a rate greater than hydrostatic. Pierson
[42] measured pore pressures in tanks containing stationary debris flow material,
and also the rate of change of pore pressure following mixing; he concluded that
since excess pore fluid pressures dissipated only slowly, over several hours, grains
could be supported b y this mechanism for the duration of a debris flow. Hampton
[21] sampled the heavy grains present in the upper portions of stationary slurries
and concluded that the results supported the buoyancy concept.
This idea again relies on transferring a principle, legitimate in a stationary
slurry, to a flowing body of debris. Certainly, in stationary debris flow material,
connections between voids will be small due to the grain-size distribution, and
excess pore fluid pressures will dissipate only slowly; overlying heavy solids cart
be partly supported b y this excess pressure since their excess weight is transferred
to the bed by a confined stationary column of high-pressure fluid. In a shearing
body of debris flow material, however, solid grains are continuously moving
relative to one another, and voids and their intereonneetions will be continuously
changing in location and geometry. Excess pore pressures will dissipate rapidly
lof the order of seconds or less) because pore fluid and solid grains are able to move
easily in response to any pressure gradient. Again, the flowing material is unable
to resist an applied stress; and any non-hydrostatic pressure gradient will cause
additional flow, and hence dissipate.
If a rigid plug of debris is moving with the flow, because shear stresses within
it are less than its yield strength [27], then excess pore pressures can exist within
it and will dissipate only slowly. This pressure, however, still cannot be trans-
ferred to the lower flow boundary through the shearing layer that exists between
it and the plug.
I t is concluded that the concept of grain support b y buoyancy due to excess
pore fluid pressures in a moving debris flow is not realistic.

Dispersive Pressure
Debris flows can be explained by the concept that when solid grains are
present within a shearing fluid, contacts between grains result in a tendency for
grains to be forced apart, away from each other [3], [4], [34], [47], [48]. This
dispersive pressure certainly exists [1], [5] and the process implies that, under
inertial grain shearing conditions, large grains will be forced towards the low-shear
region at the surface of the grain layer. Lowe [34] and Takahashi [47], [48], [49]
further confirm the realism and relevance of the concept.
Most studies have assumed that grain interaction in a debris flow will be
inertial in nature. According to Bagnold [3], grain shearing will be inertial or
viscous respectively according as the dimensionless grain flow parameter G2 is
 Large Debris Flows: A Macro-Viscous Phenomenon 167

greater than 1500 or less than 100; G2 is given by

(~D2T
G2 = ~ (1)
2~]2

in which T, the solid-transmitted shear stress, is given by

T = ((~ - - Q) g D cos rich. (2)

Here (~ -~ grain density, 9 ~ intergranular fluid density, g ---- gravity acceleration,
D = grain diameter, ri = channel slope, c = volumetric grain concentration,
h = depth below granular surface, ,~ = linear grain concentration ( = grain
diameter/mean radial separation distance) and ~ = dynamic viscosity of the
intergranular fluid. In inertial grain shearing, momentum is transferred among
grains entirely b y impulsive contacts; that is, no grain momentmn is transferred
to the intergranular fluid between successive grain contacts. In viscous grain
shearing, grain momentum resulting from a collision is entirely dissipated by
fluid friction before the next collision. Bagnold [3] shows b y further analysis that
grains will disperse at uniform concentration throughout the flow if G~ =< 100,
which, if the intergranular fluid is water and flow depth is about 0.2 m, sets an
upper limit of D ~ 50 microns for grains of quartz density. This explains the
existence of slurries of clay particles in water, which do not show a concentration
gradient with depth [3], [10]. The large grains conspicuous at the surface of
moving debris flows and shown b y deposits to be present throughout the flow
depth could not possibly experience viscous grain shear if the intergranular fluid
were water, and the dispersion of large grains is not predicted b y the analyses of
Takahashi [47], [48] or Lowe [34] (although it was a s s u m e d b y Bagnold [4]). If,
however, the interstitial fluid were the slurry of much finer grains which makes up
low-density non-pulsing debris flows -- as the granulometry of the Jiangjia flows
[33] certainly suggests (Fig. 1) -- then, as shown below, viscous grain shearing
becomes a possibility, and dispersive stress is able to explain how large grains are
maintained within the body of a moving debris flow.
In the following, the role of dispersive stresses and the instability leading to
pulsing flow are examined in more detail.

Hypothesis
In an attempt to explain the behaviour of large debris flows, the following
two-part hypothesis is proposed:
a) The dense, highly dispersed, erosive nature of debris flow pulses results
from mobilisation of bed material under viscous grain shearing conditions and
laminar fluid flow.
b) Pulsing in debris flows occurs if a surface disturbance is amplified by the
instability of open channel flow and consequent development of roll waves.
168 T . R . H . Davies:

10 2
IN
~176 ..... LOW DENSITY FLOWS 1 . 1 1 < y < I . / , 0 I I m3

so
CLASS
/,\ HIGH DENSITY F L O W S 1 , 5 6 < y < 2 . 2 6 T / m 3

/
t X
/ . ~ Y--1.o Y--21

,,"A ,, AA
I\",, IX'\

i GRAIN SIZE [mm] ~

10~ ....... I ........ L ........ I ........ r ......
10 .3 10 .2 10-1 10 0 101 8.101
Fig. 1. Grain-size distributions of low-density (broken line) and high-density (full lines}
debris flows, plotted on non-eumuIative logarithmic axes. Dat~ from debris flows in
Jiangjia Gully [33].
The high-density flows have the same grain-size distribution as the low-density ones up to
about 0.1 mm

a) Grain Behaviour Under V%cous Shear

I t was s t a t e d earlier t h a t w i t h w a t e r as the i n t e r g r a n u l a r fluid, viscous g r a i n
shear occurs o n l y if D < 50 microns. E q u a t i o n (1) shows t h a t for G 2 = 100,
D varies d i r e c t l y w i t h ~; thus in o r d e r for viscous g r a i n s h e a r to occur w i t h
grains of s a y 0.05 m d i a m e t e r (5 • 104 microns) ~ m u s t be increased b y a f a c t o r
of a b o u t 5 • 104/50 : 1000. ~ for w a t e r is a b o u t 10 -~ k g m -1 s -1, so the v i s c o s i t y
n e e d e d is a b o u t I k g m -1 s -1. Viscosities of this m a g n i t u d e or greater, which is the
s a m e as t h a t n e e d e d for l a m i n a r fluid flow in a n a t u r a l channel, h a v e been m e a -
sured in mudflows b y L i et al. [33]. T h a t such viscosities occur in fluids w i t h a
high c o n c e n t r a t i o n of fine s e d i m e n t was also shown b y Chu [11], whose experi-
m e n t a l l y - j u s t i f i e d t h e o r e t i c a l e q u a t i o n p r e d i c t s a n increase of v i s c o s i t y of a b o u t
a t h o u s a n d - f o l d w i t h v o l u m e c o n c e n t r a t i o n s of fine particles (up to 5 m m d i a m -
eter) of the o r d e r 50% or so. l ) a i et al. [14] f o u n d a n e x p e r i m e n t a l r e l a t i o n s h i p
b e t w e e n t h e v o l u m e c o n c e n t r a t i o n of 10 m i c r o n d i a m e t e r s e d i m e n t s in w a t e r a n d
the coefficient of r i g i d i t y (or B i n g h a m viscosity), which g a v e a b o u t a 100-fold
increase of v i s c o s i t y as v o l u m e e o n e e n t r a t i o n increased f r o m 1 to a b o u t 3 5 % ,
a g a i n suggesting t h a t a 1000-fold increase w o u l d occur a t a v o l u m e c o n c e n t r a t i o n
or t h e o r d e r of 50%. Thus, we m a y a s s u m e t h a t the i n t e r g r a n u l a r fluid of a debris
flow p u l s e is t h e s l u r r y of fine s e d i m e n t s in w a t e r which e h a r a c t e r i s e s t h e i n t e r -
p u l s e p h a s e ; t h e large p a r t i c l e s which a r e also p r e s e n t d u r i n g t h e p u l s e [30] m a y
well be e x p e r i e n c i n g viscous g r a i n s h e a r conditions as t h e y m o v e r e l a t i v e to t h e
 Large Debris Flows: A Macro-Viscous Phenomenon 169'

intervening slurry. Note that the increased density of the intergranular fluid also
decreases G~ by a factor of about two or three as compared to water; this effect is
minor compared to that of viscosity, but helps to promote viscous grain shear [3].
Bagnold [3, p. 287 et seq.] shows that grains experiencing viscous shear in a
laminar flow of limited depth occupy the whole flow depth at uniform concen-
tration. This effect was earlier reported in experiments using grains of very low
excess density [2]. Thus the dispersion of large grains throughout a debris flow
agrees with the suggestion of viscous grain shearing. Further, Bagnold [3, p. 291
et seq.] shows that if the gradient (rate of change with depth) of the tangential
shear applied to a bed of grains underlying a fluid flow exceeds the gradient of the
bed yield strength, then the bed will shear to a indefinite depth and the whole bed
will flow; if Ge < 100, the bed grains will also become dispersed throughout the
whole flow depth from the free surface to the underlying bed. The gradient of the
applied tangential stress must be extremely high for this to occur, but if it can be
achieved the predicted dispersion of the bed material and the consequent deep
scour of the channel are certainly in agreement with the observed behaviour of
debris flows (Table 1). The observations b y Okuda et al. [40] and Li et al. [33]
that the granulometry of debris flow deposits suggests their origin in the bed
material of channels upstream, also supports this view, as does the common
observation that debris flows scour much more strongly vertically than hori-
zontally, leaving a characteristic deep U-shaped channel. The existence of
extremely high shear stress at the base of a dispersion of large grains is shown b y
Hirano and Iwamoto [29], who report slip at this boundary and hence a very high
velocity gradient.

Rheology

I t has been suggested [27], [19] that the behaviour of debris flows can be
modelled as a Bingham plastic, that is, a substance which has a yield stress below
which it acts as a solid and above which it deforms at a strain rate proportional
to the amount by which the yield stress is exceeded; the constant of propor-
tionality is the inverse of the Bingham viscosity or coefficient of rigidity. This
behaviour is illustrated in Fig. 2.
Johnson [27] and l~odine and Johnson [44], among others, have shown that
a Bingham plastic can simulate debris flow behaviour quite reasonably. The
dispersed grain-flow behaviour suggested b y Bagnold [2] and applied herein to
debris flows has a rheology quite similar to that of a Bingham plastic, and is
properly described as a dilatant plastic (Fig. 2). When the linear concentration of
grains ~ is greater than about 22, the grain body behaves as a solid in that a certain
level of stress (the yield stress) is needed to cause grains to dilate and shear past
each other. When ~ is less than about 14, the grain body behaves as a Newtonian
fluid, which has no yield stress and which deforms at a strain rate proportional to
the applied shear stress. As ~ reduces from 22 towards 14, the grain body behaves
170 T . R . H . Davies:

Zv-YIELD STRESS J_s~ S

~/X=22 s s ~ ~
i ~ "~ / " NEWTONIAN
"C~,].~y~~ / / / ---- BINGHAM
/" .... DILATANT PLASTIC
/ ........ O NULAR

STRAINRATE
Fig. 2. Rheology of Newtonian fluid, Bingham plastic and dilatant plastic (as described
by Bagnold [2] for grain shearing) materials. The granular curve is a possible feature of
dilatant grain flow rheology analogous to Engelund and Wan's [17] analysis of the
properties of Bentonite

us a non-Newtonian fluid, with viscosity (shear stress divided b y velocity gradient)

decreasing with grain concentration [2].
Davies [16] showed that the inferred rheology of the Sherman Glacier rock
;avalanche [37], in which grains sheared with air as the interstitial fluid, was in
qualitative agreement with that of the dispersive stress model.
The rheology of a dispersed grain system is similar to that of a Bingham
plastic. Being based on theoretical and experimental consideration of material
behaviour rather than simple assumption, dispersive stress models m a y prove to
be more realistic and satisfactory than the simple Bingham model.
Engelund and W a n [17] show that a clay slurry will flow in pulses if its yield
stress is greater than t h e minimum shear stress during flow. I t seems quite likely
that a body of grains will also show this characteristic since it is well known that
a larger stress is needed to overcome static friction between two solid bodies than
to maintain their sliding ]notion. The rheology will then be as shown b y the
granular curve of Fig. 2, and pulsing of flow can result from halting of grain flow
b y solid friction and subsequent remobilisation of the grains due to flow build-up
behind them [2], particularly in the steep, confined upper stream channel.
We now consider the conditions required for a surge or pulse, once initiated, to
.amplify rather than to attenuate as it passes down a channel.

~) Instability o/ OpenChannelFlow
Due to an inherent imbalance between the gravity force driving the flow and
the resistance provided b y the channel boundaries, open channel flow is unstable
a n d develops into a series of roll waves if the Froude number, F, exceeds some
 Large Debris Flows: A Macro-Viscous Phenomenon 171

critical value Fc [22]. F is defined as

F = V/(gd) 1~2, (3)

where V and d are mean flow velocity and depth respectively, and gis the acceler-
ation due to gravity. Fe depends on the cross-sectional shape of the channel and
on the Reynolds' number of the flow Re

Vale
Re = ~ (4)
r]
where ~ is the fluid density.
Experimental and theoretical studies have investigated the occurrence of roll
waves, with the following conclusions:

i) I n highly turbulent flow (Re > 105), Fc is close to 2 in a wide channel [22].
(ii) As Re decreases towards the turbulent limit of about 1500, Fc decreases
towards about 1.3 [6].
iii) With 600 < Re < 1500, flow instability decreases [36], [26]. Slug waves
m a y occur, which result from the development of slugs or patches of turbulent
flow separated b y laminar flow [36], [7]. These move more slowly than the under-
lying flow [36].
iv) With Re < 600, in laminar flow, F~ is a b o u t 0.6 in an infinitely wide
channel [6], [26]. As the width-to-depth ratio becomes smaller, however, Fc in-
creases; with a width-to-depth ratio of 10, Fr is about 0.8. Fig. 3 summarises the
flow conditions needed for roll wave development.
v) With F > 1.3 (Re > 1500) the small initial roll waves travel more slowly
than the underlying flow [36], [7]; as the waves grow, however, the intervening
flow becomes shallower and slower and m a y vanish altogether [23].
vi) Roll waves in laminar flow always travel faster than the flow on which
they are superimposed [7], [36].
vii) The transverse spacing and regularity of roll waves increases down-
stream [36].
Because F in natural channels rarely, if ever, exceeds 1.0 when averaged over
substantial channel lengths, and because Re in natural channels is usually of the
order of 105 or so, it is d e a r t h a t roll-wave instability is unlikely to occur. Roll
waves do occur in artificial channels where flow boundaries are smooth and slopes
steep [8], [23]. Thus, if flow in a wide natural channel becomes laminar, due to the
presence of a high concentration of fine sediment in suspension, conditions are
suitable for roll waves to occur if F > 0.6, I n addition, if the water surface is
strongly disturbed, roll waves would grow in amplitude very rapidly [36]. Such
waves can travel long distances while retaining their identity, and will tend to
persist as a channel becomes wider and shallower downstream since high width/
depth ratios are more conducive to the maintenance of instability [6].
I t seems unlikely t h a t debris flow pulses can be caused b y slug flows [36], [7]
172 T . R . H . D~vies:

since the slug itself is turbulent, b y contrast with the viscous appearance of
debris flow waves, and turbulence would cause the dispersed grains to fall to t h e
bed [2].
The mechanism of roll wave generation and development can thus explain
the very large debris flow waves reported b y K a n g and Zhang [30], Li et al. [33]
and Niyazov and Degovets [39] in wide valleys a considerable distance downstream
of the debris source. Certainly, roll waves in debris flows are visually and gee-
metrically v e r y similar to flow roll waves in water.
One might reasonably question whether roll waves can in fact occur in a. fluid
which contains a high concentration of solid particles, since the studies mentioned
above referred only to the flow of water. I~oll waves have, however, been observed
in the flow of sand down a slope through a body of still water [15], so there seems
no reason to suppose that the presence of solid grains would prevent roll wave
formation. A plug flow, in which the upper p a r t of the flow is rigid, would however
prevent a n y free surface instability, and the presence of roll waves thus indicates
that plug flow probably does not occur (and, in fact, that the yield stress ~y of
the flow is small).
Li et al. [33] note that mudflows in the Jiangjia Ravine, China, begin to pulse
when the previously continuous flow rate reaches a b o u t 5 m a s 1 and the fluid
density reaches a b o u t 1.6 T m s, equivalent to a solids concentration of about
35% b y volume. I n the 25 m wide channel they describe, a viscosity of about
500 times that of water would be required for flow to become laminar at Re = 500,
and this is indeed the order of increase which would be expected with a solids
concentration of the order of 35% [11], [14]. Thus, a measure of support is given

FROUDE i!i!::!i!i!i!! LOWER LIMIT OF ROLLWAVES~..

N U M B E R ~iiiiii!iiii; u ~ [~
1.5 ~:-:-:,:.:-:~ n ~
~::::::::::i
t~
U
F ~ . : ! ? ! : i : ! : L , /
, i , : . : . : . : - : . : ~
i::::::::::::~
r.::::::::::9 ~.~
I::::::::::::~
~i-:.:.:.:.:~ Q~ J

iil I
::::::::::1
- SLUG LOWS
9 ~::::::::::::l
":i:i:i:i:i:i Re
~ " ~i:i:i:i:i:iI REYNOLDS' NUMBER ==r
0.5 . . . . . . . . . l ........ , ........ I ........

10 2 10 3 I0 ~ 10 5 10 6
H HUNSHUI (INCIPIENT PULSING) J2=JIANOJIA(NO PULSING)
J1 JIANGJIA (INCIPIENT PULSING) Q =QUINSHUI(NO PULSING)

:Fig. 3. Theoretical lower limit of roll-wave development in w0~ter [6] (~ull line) compared_
with field data for incipient pulsing and non-pulsing debris flows [32], [33], [53]. Shading
indicates possible imprecision of d~ta
 Large Debris Flows: A Macro-Viscous Phenomenon 173

to the hypothesis that the onset of pulsing is caused b y the development of roll
waves as the flow becomes laminar.
Field data from China [32], [33], [53] allows a rough test of the hypothesis that
debris flow pulses are caused b y a roll-wave instability. Fig. 3 shows the lower
limit of roll-wave development in water flows [6], and it is assumed that this
applies also to debris flows. The conditions for the onset of pulsing in Jinangjia
Gully are known [33], as are those for flows apparently at the limit of pulsing in
I~Iunshui Gully [53], and these both plot close to the theoretical limit for roll-wave
development. Non-pulsing flows in Jiangjia Gully and Quinshui Gully have also
been measured [33], [32] and plot well within the stable region. The shading in
:Fig. 3 represents the degree of confidence in the assumptions of flow viscosity and
velocity which had to be made with these data. The field data thus do not contra-
dict the roll-wave hypothesis, which seems worthy of further investigation.
The present hypothesis suggests that the introduction of large quantities of
fine material can cause the flow to become laminar, and create conditions favour-
able for the development of roll waves. Grain shearing conditions also become
viscous, so that material of all sizes becomes dispersed throughout the flow; and
if bed shear stresses become extremely high, the underlying bed material can
become incorporated in the flow, and similarly dispersed.

Debris Flow Development

In the light of the foregoing discussion the following sequence of events is
suggested whereby a large, pulsing debris flow develops (Fig. 4).
a) Following persistent rain, flow and sediment concentration in upland
streams are high, but flow is still turbulent and bedload moves at the base of the
flow as usual (:Fig. 4 a).
b) Slope erosion processes caused by intense rain wash large quantities of
fine sediment into the stream; flow becomes laminar and bedload grain shearing
becomes viscous. Bedload grains disperse throughout the flow depth and in a very
steep channel, bed material is eroded and dispersed at a high rate (Fig. 4b).
c) The coarse grain concentration m a y become so high that, in smaller flows,
intergranular friction causes the flow to slow or stop; it is remobilised b y the
pressure of material accumulating behind it and then moves downstream, causing
intense bed scour and a characteristic U-shaped channel (Fig. 4 c, 4d).
d) The remobilised jams from m a n y small tributaries enter the main channel
as surges, causing a random series of disturbances to the flow therein. Under
suitable flow conditions (Fig. 3) these disturbances amplify as roll waves, becoming
organised in the downstream direction as an increasingly regular series of pulses
(Figs. 4 e, 5).
e) The additionalbed shear stress caused by roll waves incorporates more coarse
bed material into the flow, it is dispersed and increases the bulk flow density
which increases the shear stress still more; the pulse becomes highly erosive.
174 T. 1% H. Davies:

Q SUSPENDED LOAD
BEDLOAD
....~._,. -.?
9L ~ T ~ . %, " -

~o:o:~,~ -..~..~9

!.~i.'.i.-~!,i-ii~!:::? ~
-~:,,:-':f'Y. ~ .: v : , ...~

BED EROSION INCREASES C

TRIBUTARY CROSS-SECTION TRIBUTARY CROSS-SECTION
HIGH STREAMFLOW R e >> B O1:1 F O L L O W I N G RAIN B U R S T
a) G ~ > I~00 b) Re~<500 G~<I00 C~0.7

. ==r ,
: ~ ~ ' . ' : ~ /-- STATIONARY

". ""~:(~
..::,9 ~.:.
~ " .~ ..... ~
...o...~.. : . . 9 : =
: -"G :~. ,-v . ~ , ~ -
o 0 ~ ' . ~ ; ~,
. . . . "'~a::.'n:o.'.).:C:q..~::@: :~.::..~:q.:. ~b: ' w o ? : ~ - ~ : . ~ . W , ~ _ _ ~
. . . . . . . . . . . :'~ : ~ ~: ~.:::'~:.6::::~:.~:!o:?:~~: ; ~ " ~ : ? : " .~
. . . . . . . . ,.~..:.-~.:.:.~.:~_::.:~,
TRIBUTARY LONG-SECTION
JAMMING AS C >0.7
e)

~.:o:~..:b~.:~.:.~-.~.,o!
~..:.~-
. . . . . ...... '.~- ~--.-
~ .......... ~ .~.::..
. . . . "0~, : ~
,........~:~
• ..... ~.~_? j ~ ~ .o_.:;!,.
~u~ : ~ ..%
: ~ .~. . .:0: . . . ., ~. . . . . ~-
....~,.~ ::.,.~.~.:~:~...,.~.~.~.:~!:
,, ~.~..,.:..~v:.:.~-.~...~.~,..,,
~ . ~ ~ . ~ . ~ o ~ . ~ , ~ . ~~. . . , ~ . . . . ~ ,
~!tr9: ~ i ~ : ? : @ ;~:"~::.&:::....:~::
.......... 9~:,:'~!~.:.~.::!i-~?!~),:~:::.!:!%~r ~.
...... ":'.:'D:j,:7.~:

TRIBUTARY LONG-SECTION
JAM RELEASED BY MATERIAL BUILD-UP CAUSING SURGE
d) C~<0.7

TRIBUTARY SURGES I~ ROLL WAVES IN M A I N C H A N N E L ~

i
I

0 W S

IN C H A N N E L

e)

Fig. 4:. Scenario for development of large pulsing debris flows

 Large Debris Flows: A Macro-Viscous Phenomenon 175

1,2M

>
Z
(.')
rn
1.8M

o
2.4M

F~
-'4
3.7M

. ~"~..~.-..~/X 4.9 M

~ 6 . 1 M

~__fM,,,~/~ 7.0 M
I, I
1SEC
Fig. 5. Evolution of surface-tension roll-waves in water from random input disturbances,.
showing increase in regularity, wavelength and amplitude downstream (after [36])

Conclusions

i) According to the hypothesis investigated herein, large debris flows will

develop surge waves only if the flow conditions are suitable for roll wave develop-
ment; they will carry large material by dispersive stress only if the shearing of
these grains in the intergranular slurry is viscous ; and if these two conditions are
fulfilled, the flows m a y be capable of scouring the channel bed to extreme depth.
ii) Available data support the present hypothesis, which is therefore proposed
as a reasonable framework on which future work might be based. I n order to test
the hypothesis properly, however, more and better data on debris flows and on
viscous grain shearing is required, from both field and laboratory work.

Acknowledgements
The writer gratefully acknowledges financial support from the Eidgen6ssische Anstalt fiir
das forstliche Versuchswesen, Birmensdorf, during the preparation of this manuscript, and
the provision of facilities and services by the Versuehsanstalt fiir Wasserbau, I-Iydrologie
und GlaziologieETH-Zfirich, and Lincoln College, Canterbury, New Zealand. g. Bfihler made ~
valuable comments and suggestions on an earlier draft of the manuscript.
176 T . R . H . Davies:

References

[1] Bagnold, R. A.: Experiments on a gravity-free dispersion of large solid spheres

in a Newtonian fluid under shear. Proceedings Royal Society of London 225, 49--63
(1954).
[2] Bagnold, R. A. : Some flume experiments on large grains but little denser than the
transporting fluid, and their implications. Proe. Inst. Civ. Eng. pp. 174--211 (1955).
[3] Bagnold, R. A. : The flow of eohesionless grains in fluids. Phil. Trans., Royal Soe.
London A 249, 235--297 (1956).
[4] Bagnold, R. A.: Deposition in the process of hydraulic transport. Sedimentology 10,
45--56 (1968).
[5] Bailard, J. A., Inman, D. L.: A reexamination of Bagnold's granular-fluid model
and bed load transport equation. J. Geoph. Res. 84, C 12, 7827--7833 (1979).
[6] Berlamont, J. E., Vanderstappen, N.: Unstable turbulent flow in open channels.
ASCE, J. ttydraul. Div. 107, 427--449 (1981).
~7] Binnie, A. M. : Instability in a slightly inclined water channel. J. Fluid. Mech. 5,
4, 561--570 (1959).
i[8] Brock, R. R.: Development of roll waves in open channels. Report No. KH-R-16,
p. 226. W. N. Keck Lab., Calif. Inst. Techn., Pasadena, 1967.
[9] Broscoe, A. J., Thompson, S.: Observations on an alpine mudflow, Steele Creek,
Yukon. Can. J. Earth Sc. 6, 219--229 (1969).
[10] Carter, R. M.: A discussion and classification of subaqueous mass transport with
particular application to grain-flow, slurry and fluxoturbidities. Earth-Sc. Rev. 11,
145--177 (1975).
[11] Chu Junda: The viscosity of sediment-water mixture. Proc. Int. Symp. River Sed.,
Beijing, China, 1, 205--211 (1980).
[12] Costa, J. E.: Physical geomorphology of debris flows, in: Developments and appli-
cations of geomorphology (Eds.: Costa, J. E., Fleischer, P. J.) Springer 1984.
[13] Curry, R. R.: Observation of alpinemudflows in theTenmile Range, Central Colorado.
Bull., Geol. Soc. Amer., 77, 771--776 (1966).
[14] DaiJilan, Wan Zhaohui, Wang Wenzhi, Chert Wukui, Li Xijun: An experimental
study of slurry transport in pipes. Proc. Int. Symp-River Sed., Beijing, China, 1, 195
to 204 (1980).
[15] Davies, T. R. It.: The investigation of avalanche processes by laboratory experi-
ments; exploratory tests. Intern. Rep., p. 31. Laboratory of Hydraulics, Hydrology
and Glaciology, ETH-Zurich, 1979.
[16] Davies, T. R. It.: Spreading of rock avalanches by mechanical fluidization. Rock
mech. 15, 9--24 (1982).
[17] Engelund, F., Wan Zhaohui: Instability of hyperconcentratedllow. ASCE, J. Hydraul.
Div. 110 ttY3, 219--233 (1984).
[18] Eisbacher, G. H.: Mountain torrents and debris llows. Episodes 4, 12--17 (1982).
[19] Enos, P.: Flow regimes in debris flows. Sedimentology 24, 133--142 (1977).
[20] Hampton, M. A.: The competence of fine-grained debris flows. J. Sed. Petrol. 45, 4,
834--844 (1975).
[21] Hampton, 1VLA.: Buoyancy in debris flows. J. Se4. Petrol. 49, 3, 753--758 (1979).
[22] Henderson, F. M. : Open channel flow. Macmillan 1966.
[23] Holmes, W. H.: Travelling waves in steep channels. Civil Eng. 6, 367--368 (1936).
[24] Ikeya, I:[. : Introduction to Sabo works, p. 168. Japan Sabo Association 1976.
[25] Ikeya, H.: A method of designation for area in danger of debris flow, in: Erosion and
sediment transport in Pacific rim steepland (Davies, T. R. H., Pearce, A . J . , eds.)
IAHS Pnbl. No. 132, 576--588 (1981).
 Large Debris Flows: A Macro-Viscous Phenomenon 177

[26] Ishihara, T., Iwagaki, Y., Iwasa, Y.: Discussion of "Roll waves and slug flows in
inclined open channels". ASCE, J. Hydraul. Div. 86, 45--60 (1960).
[27] Johnson, A. M. : Physical processes in geology, p. 577. Freeman Cooper and Co. 1970.
[28] Johnson, A. M., Rahn, P. H. : Mobilization of debris flows. Zeit. Geomorph., Suppl. 9,
168--186 (1970).
[29] Hirano, M., Iwamoto, M.: Mechanical characteristics of debris flow. Proc. XVII
IUFRO Congr. Kyoto, Japan, 1981.
[30] Kang Zhicheng, Zhang Shucheng: A preliminary analysis of the characteristics of
debris flow. Proc. Int. Symp. River Sed, Beijing, China, 1, 213--226 (1980).
[31] Kroafellner-Kraus, G. : Uber den Geschiebe- und Feststofftransport in Wildbgchen.
Ost. Wasserw. 34, 213--226 (1980).
[32] Li Jan, Luo Defu: The formation and characteristics of mud/low and flood. Zeit.
Geomorph. 25, 4, 470--484 (1981).
[33] Li Jan, Yuan Jianmo, Bi Cheng, Luo Defu: The main features of the mudflow in
Jiang-Jia Ravine. Zeit. Geomorph. 27, 3, 325--341 (1983).
[34] Lowe, D. R. : Grain flow and grain flow deposits. J. Sed. Petrol. 46, 1, 188-- 199 (1976).
[35] Massey, B. S.: Mechanics of fluids, 5th ed., p. 625. Old Wokingham, Surrey: Van
Nostrand-Reinhold U. K. Ltd. 1983.
[36] Mayer, P. G. : Roll waves and slug flows in open channels. ASCE, J. Hydraul. Div.
8~, 99--141 (1959.).
[37] McSaveney, M. J.: Sherman glacier rock avalanche, in: Roeksiides and avalanches,
Vol. 1 (Voight, B., ed.) Dev. Geotech. Eng. 14 A, pp. 197--258. Elsevier 1978.
[38] Middleton, G. V., Hampton, ~I. A.: Subaqueous sediment transport and deposition
by sediment gravity flows, in: Marine sediment transport and environmental manage-
ment (Stanley, D. J., Swift, D. J. P., eds.), pp. 197--218, 1976.
[39] Niyazov, B. S., Degovets, A. S.: Estimation of the parameters of catastrophic mud-
flows in the basins of the lesser and greater Almatinka Rivers. Soy. Hydrol. 2, 75--80
(1975).
[40] Okuda, S., Suwa, It., Okunishi, K., Yokoyama, K., Nakano, M.: Observations on the
motion of a debris flow and its geomorphological effects. Zeit. Geomorph., Suppl.
85, 142--163 (1980).
[41] Pierson, T. C. : Erosion and deposition by debris flows at Mr. Thomas, North Canter-
bury, New Zealand. Earth Surf. Proc. 5, 227--247 (1980).
[42] Pierson, T. C. : Dominant particle upport mechanisms in debris flows at Mr. Thomas,
New Zealand, and implicatio ,s for flow mobility. Sedimentology 98, 49--60 (1981).
[43] Rodine, J. D. : Analysis of the mobilization of debris flows. Ph. D. Thesis, Stanford
University, 1974.
[44] Rodine, J. D., Johnson, A. 5~. : The '~bility of debris, heavily freighted with coarse
elastic materials, to flow on gentle slopes. Sedimentology 23, 213--234 (1976).
[45] Sharp, R. P., Nobles, L. H.: Mudflow of 1941 at Wrightwood, Southern California.
Bull. Geol. Soc. Amer. 64, 547--560 (1953).
[46] Suwa, H., Okuda, S.: Dissection of valleys by debris flows. Zeit. Geomorph., Suppl.
85, 164--182 (1980).
[47] Takahashi, T.: Mechanical characteristics of debris flow. ASCE, J. Hydraul. Div.
104, 1135--1169 (1978).
[48] Takahashi, T. : Debris flow on prismatic open channel. ASCE, J. Hydraul. Div. 106,
381--396 (1980).
[49] Takahashi, T.: Debris flow. Ann. Rev. Fluid Mech. 13, 57--77 (1981a).
[50] Takahashi, T. : Estimation of potential debris flows and their hazard zones; soft
countermeasures for a disaster. Natural Disaster So. 13, 57--89 (1981b).
[15] Takahashi, T., Ashida, K., Sawai, K.: Delineation of debris flow hazard areas, in:

12 Acta Mech. 63/I--4

178 T . R . H . Davies: Large Debris Flows: A Macro-Viscous Phenomenon

Erosion and sediment transport in Pacific rim steeplands (Davies, T. R. H., Pearce,
A. J., eds.) IAHS Publ. No. 132, 589--603 (1981).
[52] Woodruff, J. F. : Debris avalanches as an erosional agent in the Appalachian Mountains.
J. Geol. 70, 399--406 (1971).
[53] Zhang, X., Liu, T., Wang, Y., Luo, J.: The main features of debris flows and control
structures in Hnnshui Gully, u Province, China. Proc. Int. Symp. on Erosion,
Debris Flow and Disaster Prevention, Tsukuba, Japan, 1985.
[54] Ze]ler, J.: Die Sehwierigkeit einer teohnisch korrekten Festlegung der Wildbaeh-
ge~ahrenzonen. Sonderdruek, I00 Jahre Fachveranstaltungen, Wien, pp. 169--187
1972.
T. R. H. .Davies
Department o/Agricultural Engineering, Lincoln College
University o/Canterbury
Christchurch
New Zealand
presently at: E.A./. V., CH- 8903 Birmensdori, Switzerland