Sei sulla pagina 1di 25

Journal of Volcanology and Geothermal Research 98 (2000) 91115

www.elsevier.nl/locate/jvolgeores

A mechanical model for Merapi-type pyroclastic flow


T. Takahashi a,*, H. Tsujimoto b
a
Disaster Prevention Research Institute, Kyoto University, Gokasho, Uji Kyoto 6110011, Japan
b
Japan Weather Association, 16-11, Higashi-imazato 3, Osaka 5370011, Japan
Received 11 March 1999; received in revised form 3 December 1999; accepted 3 December 1999

Abstract
A numerical simulation model of Merapi-type pyroclastic flow for predicting a potential hazardous area is given, and its
applicability is examined by numerically reproducing actual phenomena that occurred in 1991 at Unzen Volcano, Kyushu
Japan. Merapi-type pyroclastic flow arises from collapse of the lava dome whereby large lava blocks are crushed into smaller
particles during movement down the steep slope. The early stage of flow (it may be called the debris avalanche stage), in which
the dominant particles are coarser than about 1 mm in diameter, is considered as a grain (granular) flow. In an inertial grain flow
particle collision stress plays an important role in the mechanism of flow, and the effect of gas emitted from the lava blocks is
minimal. Most of the particles composing the inertial grain flow deposit over a comparatively short range on steep slope due to
marked resistance within the flow. The remainder which is composed of mainly fine particles smaller than 1 mm continues to
run down by a support of the upward flow of gas ejected from the material itself, thereby the resistance to flow becomes small.
Thus, the pyroclastic flow stage in a narrow sense appears. The main body of the pyroclastic flow is composed of lower
insufficiently fluidized layer (bottom layer) and upper fluidized layer in which the entire weight of the particles is supported by
the upward gas flow. Sometimes, the fluidized layer cannot exist if gas emission is insufficient. As the slope down which the
pyroclastic flow moves becomes milder downstream, the bottom layer deposits some solids because the driving force due to
gravity within this layer becomes smaller than the resistance due to inter-particle contact. Then, a part of the fluidized layer, if
exists, changes to a part of the bottom layer because of the shortage of the upward gas flow that is caused by deposition of the
gas-emitting particles upstream. Thus, the entire main body stops when it arrives at a gently sloping area. Smaller particles and
gas escape from the main body, generating a hot ash cloud layer above the main body. The hot ash cloud layer can travel
independent of the main body, but its development or attenuation depend on the conditions of the supply of particles and gas
from the main body. When the hot ash cloud lacks the supply of particles and gas due to the stopping of the main body upstream,
or by its swerving from the course of the main body, it becomes soon weakened and stops. The entrainment of ambient air, the
escape of air and ash from the upper boundary as a buoyant plume, and the particle settling also affect the behaviors of the hot
ash cloud. Fundamental mechanics of the grain (granular) flow stage, the pyroclastic flow stage and the transition from the
former to the latter stages are discussed theoretically and experimentally, and mathematical formulae describing the phenomena
are obtained. These one-dimensional equations are extended to fit the planar two-dimensional system. Behaviors of the main
body as well as the hot ash cloud are simulated using the obtained system of equations. The results of simulation are very
satisfactory in comparison to the actual phenomena. 2000 Elsevier Science B.V. All rights reserved.
Keywords: pyroclastic flow; Merapi-type; mathematical simulation; granular flow; fluidized flow; hot ash cloud; Unzen

* Corresponding author. Fax: 81-774-32-6039.


E-mail address: takahasi@sabom.dpri.kyoto-u.ac.jp (T. Takahashi).

0377-0273/00/$ - see front matter 2000 Elsevier Science B.V. All rights reserved.
PII: S0377-027 3(99)00193-6
92 T. Takahashi, H. Tsujimoto / Journal of Volcanology and Geothermal Research 98 (2000) 91115

Nomenclature
b rate of gas emission from the unit mass of the solids
C volume concentration of solids
C solids concentration in a static bed
C mean solids concentration in the flow
Csmin threshold concentration that generates enduring inter-particle contact stresses
Cmf solids concentration that gives rise to fluidization
Cp solids concentration in the hot ash cloud layer
Cst solids concentration of granular flow on a flatter slope than the angle of repose
D uplift force due to upward gas flow
dp diameter of solid particles
dfc critical diameter of particles that produce fluidization
Es particle entraining velocity from the main body
e coefficient of restitution of the particle
F external force that operates on the deposition volume
f resistance coefficient
fb resistance coefficient at the lower boundary of the hot ash cloud layer
fcu ratio of the particle fraction transportable to the ash cloud layer from the main body
fi resistance coefficient at the upper boundary of the hot ash cloud layer
g acceleration due to gravity
g0 radial distribution function
H height of the upper boundary of the main body
Hf thickness of the main body
Hg granular flow depth
Hp thickness of the hot ash cloud layer
Hpa pressure in the hot ash cloud in hPa
h height of the bottom layer measured from the original ground surface
h0 height at which the driving and resistance forces balance
hf thickness of the bottom layer
In rate of increase in particle number
if rate at which the deposit surface rises
ig deposition speed in the granular flow stage
Kx x-wise coefficient of diffusion
Ky y-wise coefficient of diffusion
` mixing length
m mass of the volume deposited in the next time increment
mn numerical exponent
N frequency of particle collision
n weight ratio of sodium hydrogen carbonate to sand
nb deposited particle number per unit volume
np number of particles per unit volume
n particle number per unit volume of flow
ps static skeleton pressure directly transmitted from particle to particle
T granular temperature
Ta temperature of the ambient atmosphere, Kelvin
Tp temperature of the hot ash cloud layer, Kelvin
U0 velocity of the main flow part
T. Takahashi, H. Tsujimoto / Journal of Volcanology and Geothermal Research 98 (2000) 91115 93

Uf x-wise mean velocity of the main body


Ug cross-sectional mean velocity of the granular flow
Up x-wise mean velocity of the hot ash cloud
u local mean velocity
ud relative velocity of the deposition volume observed from the main flow
u shear velocity p
up shear velocity in the hot ash cloud ghp sin u;
Vf y-wise mean velocity of the main body
Vp y-wise mean velocity of the hot ash cloud
v velocity of the volume deposited in the next time increment
W weight of sand in the experimental pyroclastic material
w upward gas velocity
we rate at which the upper surface of the hot ash cloud layer rises
we1 air entraining velocity
we2 gas supply velocity from the main body
we3 detraining velocity of the ash cloud layer as a plume
wf rate of rise of the boundary between the main body and the hot ash cloud layer
ws settling velocity of a particle
z height measured perpendicular to the original ground surface
z0 height of the ground surface
zb surface height of the newly deposited bed
Greek symbols
as ratio of the quasi-static skeleton pressure to the entire operating pressure
b coefficient representing the fragileness of particles on collision
1p coefficient of eddy viscosity
u slope gradient
u bx0 slope angle of x-axis
u by0 slope angle of y-axis
k von Karman constant
m viscosity of the gas
n kinematic viscosity of the gas
na apparent kinematic viscosity of the material comprising the main body
r density of the gas
ra density of the gas phase
rm apparent density of the main body
r mp apparent density of the hot ash cloud layer
s density of the solids
t shearing stress
t bx x-wise shearing stress at the flow bottom
t by y-wise shearing stress at the flow bottom
tc shearing stress due to collision of particles
tg total shearing stress in the granular flow
tk kinetic shearing stress due to the particles in one layer plunging into the other layer
ts static shear stress due to enduring contact between particles
fp particle shape factor
fs internal friction angle in the solids
f sp angle of repose for the spherical particles
94 T. Takahashi, H. Tsujimoto / Journal of Volcanology and Geothermal Research 98 (2000) 91115

1. Introduction pyroclastic flow that can reproduce the phenomena


including the separation process of a pyroclastic
Merapi-type pyroclastic flows as defined by Escher flow from a debris avalanche and the formation
(Boudon et al., 1993) frequently have occurred due to process of a pyroclastic surge at the top of the main
collapse of the lava dome at Mt. Fugen of Unzen body.
Volcano in Kyushu Japan and caused many serious Numerous video records of pyroclastic flows, from
problems in nearby communities. Possible hazardous collapse of the lava dome to the stoppage of flow were
areas, within which no one was allowed to enter, even obtained (e.g. Sato et al., 1992; Ishikawa et al., 1994a;
if their homes were inside, needed to be delineated. Ui et al., 1999). Many field surveys have been done in
Under insufficient knowledge of pyroclastic flow the deposition areas (Ishikawa et al., 1993a, 1994b;
mechanics, a theory for granular flow proposed by Yamamoto et al., 1993; Fujii and Nakada, 1999;
Kanatani (1978), in which the fluidity of the material Miyabuchi, 1999). Our careful considerations in refer-
was determined solely by the inter-particle friction ence to these findings and the results of the field
coefficient, was applied in preparing the hazard map inspection and laboratory experiments by ourselves
used by the local government. In the process of come to a conclusion that the structure of the devel-
preparation of the map, an unrealistically small fric- oped pyroclastic flow (pyroclastic flow in a narrow
tion factor between particles was used to reproduce sense), which occurs after stoppage of avalanche-
the actual flow range. In turn, this small friction factor like motion of crumbled lave dome, is composed of
predicted an unreasonably thick deposit (Ishikawa et three parts: namely, the bottoma dense main body,
al., 1993a). Thus, the need for a reliable delineation a relatively dilute hot ash cloud layer on top of it, and
method was thrown into sharp relief during prolonged a violently billowing plume rising from the hot ash
volcanic activity at Unzen from 1990 to 1995. cloud layer. The main body can be divided into two
Using both the field and laboratory experimental sub-layers, namely, densely concentrated coarse parti-
data, Takahashi et al. (1995) proposed a new flow cles move downstream in a somewhat orderly manner
model that attributed the high fluidity of the pyroclas- colliding and jostling in the lowermost sub-layer, and
tic flow to upward gas flow originating in the degas- immediately above this, a sub-layer exists within
sing of the pyroclasts, themselves. This theory, which upward flow of gas seems to support the parti-
however, fits only a fully developed main body that cles; a fluidized layer (Wilson, 1980). In the latter sub-
is moving more or less steadily. Neither the processes layer turbulence is weak due to very dense concentra-
of origin and deposition of the main body nor the hot tion of particles. The hot ash cloud layer is composed
ash cloud layers formation, interaction with the main of intensely turbulent mixture of fine particles and gas.
body, and deposition process are discussed. There- This layer flows downstream as a density current and
fore, this theory was incomplete and not usable for travels a little farther even after the main body
the hazard mapping for a Merapi-type pyroclastic stopped. Valentine (1987) and Fisher (1995), for
flow, the flow modeling of the main body nevertheless example, call the main body as the pyroclastic flow
had been physically sound. A computer simulation and the hot ash cloud as the pyroclastic surge, but here
model for pyroclastic flows from dome collapse, we retain the term pyroclastic flow (in a narrow sense)
which is conscious of making hazard map, is available to indicate the overall pyroclastic current discriminat-
(Wadge et al., 1998). However, as the authors say, this ing the phenomena from the debris avalanche stage.
is not based on a rigorous physical model, but uses They do not refer to the existence of the fluidized
mathematical equations whose parameters must be layer. Theoretically, if the fragmentation of particles
evaluated by the inverse analyses similar to the does not proceed much, or if gas emission rate is too
method adopted for the hazard mapping at Unzen. small, the fluidized layer cannot appear. In this case,
There is no ground that accumulation of case studies the mechanics of pyroclastic flow (in a narrow sense)
is useful to determine the parameters. Thus, so far, is the same as the granular flow, and it is difficult to
there is no method to be used in hazard mapping flow on a gentle slope. If, on the other hand, the frag-
with sound physical background. We here propose a mentation proceeds as much as millimeter-size and
computer simulation model of a Merapi-type the gas emission is sufficient, there is a good chance
T. Takahashi, H. Tsujimoto / Journal of Volcanology and Geothermal Research 98 (2000) 91115 95

Fig. 1. Model of Merapi-type pyroclastic flow processes. Collapse of the lava dome induces a granular flow which changes to a fluidized flow.

to generate a fluidized layer. Only the existence of the ground surface is zero), the basal flow is a semiflui-
fluidized layer guarantees the reach of pyroclastic dized, in which highly concentrated particles move in
flow far downstream on gentle slope as is the case a laminar fashion with enduring contact making an
of many pyroclastic flows. ever-changing temporary skeleton structures. The
The development of a Merapi-type pyroclastic flow intensity of the upward gas flow increases in the
(in a wide sense including avalanche stage) is upward direction, so that immediately over the basal
modeled as in Fig. 1. This type of pyroclastic flow flow layer a fluidized layer appears, if the thickness of
is caused by the collapse of a lava dome, and is differ- flow is large and the upward gas flow is strong
ent from other types such as the ones due to collapse enough. The load of particles in the fluidized layer
of an eruption column or due to effluence of boiled is completely borne by the upward gas flow, and
pyroclastic material from the crater. The dome therefore the resistance to flow becomes very small,
composing rocks easily and rapidly break into small that causes the flow arrive to a mild slope reach. Note
particles. Although the instantaneous explosive frag- that the granular flow is able to exist only on very
mentation of a falling block produces plenty of fine steep slope near the angle of repose, although a
ash particles (e.g. Sato et al., 1992), most of the flow wide grain-sized population can move down substan-
space is occupied by the coarse particles. Gases tially milder slope (Lowe, 1976). With the develop-
violently escape from particles as they are broken, ment of fluidized layer within the main body, the
but the upward gas flow produced by emitted gas is smaller particles incorporated in upward gas flow
yet insufficient to fluidize the material. Therefore, the begin to escape from the main body, and a dilute
early stage of flow is a granular flow where coarse highly turbulent hot ash cloud begins to develop.
particles are suspended by repeated inelastic colli- Denlinger (1987) attributes the generation of ash
sions that cause the crushing of the particles. cloud to the turbulent diffusion of particles from the
Granular flow continues to run down depositing main body. However, because the particles are heavy
some parts of pyroclasts in response to gradual flatten- in the atmosphere, the elutriation of fine particles by
ing of slope, and within the flow fragmentation the upward gas flow would be more effective than the
proceeds. When fragments are largely millimeter- transport due to turbulent diffusion. The Rouse
sized particles, the upward flow of gas becomes strong number which controls the capacity of particle
enough to suspend particles in the upper part. This is suspension by turbulence is large for a slower pyro-
the beginning of the fluidized flow stage or the pyro- clastic flow as that at Unzen, and this condition makes
clastic flow stage in a narrow sense. Because the the turbulent diffusion small (Valentine, 1987). The
upward gas flow is insufficient in the lower part of small eddy diffusivity is also proved indirectly in case
the main body (the upward velocity at the original of Unzen by the fact that the hot ash cloud cannot
96 T. Takahashi, H. Tsujimoto / Journal of Volcanology and Geothermal Research 98 (2000) 91115

Fig. 2. Of the volume deposited for each 100 m reach in the pyroclastic flows from May 28 to June 4, the one on June 3 is the largest. Deposits in
the reach 0.51.6 km from the dome are mainly those of granular flow. The thinner deposits 1.63.2 km from the dome are those of fluidized
flow (Ishikawa et al. 1993a).

reach very far after separated from the main body. The eye, and therefore this must be produced by the gran-
fines within the hot ash cloud settle down fast rather ular flow process. Majority of the deposit at Mt.
than maintained in the flow by turbulence. Importance Merapi is also transported by an inertial granular
of fluidization on the mechanics of pyroclastic flow flow (Boudon et al., 1993). The distal deposit is by
was discussed by Sparks (1978). fluidized flow as discussed by Ishikawa et al. (1994b).
Deposition of the main body occurs in the basal Fig. 3 shows the arrival limits of the main body and
sub-layer by the decrease in the driving force corre- the hot ash cloud of the pyroclastic flow of June 3,
sponding to flattening of slope down which the main 1991 (Ishikawa et al., 1994b). Here the term surge is
body moves. If the gas-emitting particles contained in used as a synonym of flow of hot ash cloud. Fujii and
the basal sub-layer are left upstream, gas supply to the Nakada (1999)referred to the main body as block-
fluidized layer decreases to cause transfer of some and-ash flow and the hot ash cloud as ash-cloud
particles from the fluidized layer to the basal layer. surge, respectively. Fig. 3 also shows the directions
Thus, the fluidized layer gradually shrinks while of the surge estimated from the direction of felled
deposition continues. trees. The main body flowed along the gully called
Fig. 2 shows results of a survey of the deposited the Mizunashi River which originates at the summit of
volume at 100 m intervals along the valley (Ishikawa Mt. Fugen. It flowed along the confines of the river
et al., 1993a). The greater part of the material is gully. When it reached the unconfined valley
deposited within the upstream reach, a thin deposit debouchment, it continued for another 1 km with
layer extending downstream from there. The proximal lateral dispersion. The deposit has tongue shape, and
deposit is composed of numerous big boulders and a lateral cross-section convex at the central part. The
coarse particles as is evident from observations by hot ash cloud kept moving in the original direction

Fig. 3. Area affected by the main body and hot ash cloud of a pyroclastic flow that occurred 3 June, 1991 at Unzen Volcano. The open circle on
arrow A shows the place where 43 people were killed by the hot ash cloud (pyroclastic surge) (Ishikawa et al. 1994b).
T. Takahashi, H. Tsujimoto / Journal of Volcanology and Geothermal Research 98 (2000) 91115 97

(arrow A in Fig. 3) and detached from the main body them, Takahashi and Tsujimoto (1997) have shown
which changed direction. The hot ash cloud killed 43 the following constitutive relationships:
people on top of a small hill (the circle on arrow A). r 
4 2 T 2u
These people stayed to watch descending pyroclastic tc C g0 sdp 1 e 1
flows but did not anticipate the development of a hot, 5 p 2z
low density ash cloud that could spill out of its chan- r 
nel and flow uphill. Similar divergent phenomena sd p T 2u
tk 2
occurred upstream: At the position of the first water- 3g0 p 2z
fall shown in Fig. 3, the main body changed direction
 2
with the valleys topography, whereas the hot ash dp2 1 2u
T 3
cloud kept its inertial motion for about 600 m (the 15 1 e 2z
arrow B in Fig. 3).
We discuss here the processes that occur from just ts ps tan fs 4
after collapse of the lava dome to the final deposition
downstream based on the model described earlier. The ps as CsgHg zcos u 5
quantitative discussion on the fragmentation in gran-
ular flow makes possible to explain transition from where C is the volume concentration of the solids, s
granular flow stage to the pyroclastic flow stage in the density of the solids, dp the diameter of the solid
the narrow sense. The mechanical models for particle, e the coefficient of restitution, T the granular
deposition of granular flow, the main body and the temperature that represents the energy contained in
hot ash cloud are also proposed. These models are the random motions of the particles, u the local
applied to an actual case shown in Fig. 3 to evaluate mean velocity, z the height measured from and
their validity. perpendicular to the bottom, Hg flow depth, u the
slope gradient, f s the internal friction angle in the
solids, g the acceleration due to gravity, g0 the radial
distribution function defined by:
2. Flow model in the granular flow stage
g0 {1 C=C 1=3 }1 ; 6
Immediately after the collapse of a lava dome the
material is comprised of huge blocks, boulders and ps the quasi-static pressure transmitted via a net-like
comparatively small amount of fine particles; the skeleton formed by contact of the particles in the same
effect of upward gas flow on the blocks and boulders instant, and a s the ratio of the quasi-static skeleton
is negligible. At this stage, concentration of solid pressure to all the operating pressure. Although the
grains is very large, and they are in contact or network pattern of the skeleton ever changes in the
frequently collide. This is granular flow defined by flow, it continues to exist as long as the solids concen-
Savage (1984). The mechanics of granular flow are tration is larger than a threshold value. Therefore, a s
controlled by gravitational driving actions and energy is assumed to give as following: zero static pressure
dissipation caused by inter-particle actions. The total where the solids concentration is equal or smaller than
shearing stress in subaerial granular flow, t g, is given the threshold value, all the pressure is static at the bed,
by the sum of the collision stress, t c; the kinetic stress and changing between the two extremes depending on
that arises when particles of one layer plunge into the solids concentration:
8
another layer, t k; and the quasi-static shearing stress; > 0; C Csmin
t s, that arises due to enduring contact between parti- >
> !mn
>
< C C
cles when the solids concentration exceeds the thresh- as smin
; Csmin C C 7
old value. >
> C Csmin
>
>
There are many reports on the stresses in inertial :
1; C C
granular flow (e.g. Savage, 1988), and they agree in
that the collision and kinetic stresses are proportional where C is the mean solids concentration in the flow,
to the square of the vertical velocity gradient. Among Csmin the threshold concentration that generates
98 T. Takahashi, H. Tsujimoto / Journal of Volcanology and Geothermal Research 98 (2000) 91115

enduring inter particle contact stresses, C the solids Merapi-type pyroclastic flow occurs under the
concentration when packed, and mn a constant. condition that the excess pore pressure of lava is
The equation for stress balance, in which the right smaller than the tensile strength (Sato et al., 1992),
hand side is the driving force due to gravity: therefore, the explosive collapse of dome and instan-
taneous fragmentation into fine particles are unlikely.
tg tc tk ts C sgHg zsin u; 8
To analyze the process by which lava blocks are frag-
is solved under the boundary condition u 0 at z 0; mented into smaller particles, the lava dome first is
on the assumption as a first approximation that there is considered an aggregate of large boulders assumed to
homogeneous solids concentration within the entire be crushed into smaller particles by inter-particle
flow depth: collisions during laminar flow. The frequency of colli-
( !3=2 ) sion in unit time per unit volume in a laminar flow is
u 2 Hg p z given by (Friedlander, 1983):
K1 1 1 9
u 3 dp Hg
4 3 2u 2
N d n 13
Therefore, the cross-sectional mean velocity is given 3 p 2z p
by:
where np is the number of particles in the unit volume. If
Ug 2 Hg p the rate of increase in particle number, In, is assumed to
K1 10
u 5 dp be proportional to N (i.e. the number of particles
produced per collision is assumed to be constant):
where
 s 48C 2 2u
C as tan fs f5 In b 14
K1 1 11 p2 dp3 2z
f2 f22 tan u f2
is obtained where b is the coefficient that represents the
and Ug is the cross-sectional mean velocity,
p p u
gHg psin u; f1 4hC 2 g0 ; f2 8hpC
2
g0 =5 p; f22 fragileness of the particles (or the particle production
1=3g0 p; f5 48h1 hC 2 g0 = p; h 1 e=2: rate) on collision. From Eq. (14) and the mass conserva-
To obtain the velocity and flow depth relationship tion equation, the change in particle diameter is
on a given slope from Eq. (9), the relationship obtained. The process of crushing, however, is consid-
between the slope gradient and macroscopic solids ered to end when the particle diameter becomes dfc, the
concentration must be given. Laboratory flume critical diameter at which fluidization occurs.
experiments for dry granular flow of the polystyrene
beads and natural sand (Takahashi and Tsujimoto,
3. Mechanics of the fluidized flow stage
1997) gave the following relationships:
9
C Cst constant; u u1 >
> 3.1. Outline of the experiment
 1=2 >
>
=
1 121 e 1 e
u tan 2 12 The experimental pyroclastic material was produced
5p C g0 1 ep >
>
>
> as follows: a given amount W 5 15 kg of fine
u f ; u u ; silica sand about 80 mm in diameter was heated to
1 sp 1
about 300C in a container, after which it was trans-
on rough bed: u1 fs ; ferred to the rotating drum where it was mixed with a
given amount 35 1000 g of sodium hydrogen
on smooth bed: u1 fs 5 carbonate. As soon as the latter compound was thrown
into the heated sand, H2O and CO2 gases were emitted
where f sp, the angle of repose for the spherical particles,
violently, and the interstitial fluid velocity becomes
is nearly equal to 26, and Cst has a constant value
sufficient to suspend the particles. Thus, fluidization
between 0.51 and 0.56. Fig. 4 gives a comparison of
occurs in the material. The chemical reaction was:
Eq. (12) with experimental results, showing that Eq.
(12) gives a fairly good approximation. 2NaHCO3 ! Na2 CO3 H2 O CO2 15
T. Takahashi, H. Tsujimoto / Journal of Volcanology and Geothermal Research 98 (2000) 91115 99

Fig. 5. Schematic diagram of the experimental set up.

because each cell lacked walls in front and in the


rear, something like a simple book shelf. When the
steady flow part arrived, the sheath was rapidly
Fig. 4. Comparison of the calculated volume fractions using Eq. dropped down to enclose the material that flowed
(12) with experimental results obtained in chutes with polystyrene onto each shelf. The solids volume trapped in each
beads and sand under various bed conditions. Open symbols show cell was divided by the capacity of the cell to obtain
results of the experiments with polystyrene beads (4.5 mm in the concentration at the height of each cell. In the
diameter) under various bed conditions. Closed symbols show the
results of Kitou et al. (1993) and Matsumura et al. (1997) for sand.
experimental case, the solids concentration in the
Arrows show the extent of variation in the solids concentration. main body was almost uniformly distributed, between
40 and 45% in volume, but decreased drastically
upward in the hot ash cloud layer.
An experimental flume 4 m long, 5 cm wide and Various combinations of channel slope, u 5:5
32 cm deep (Fig. 5) was used. As soon as the hot 24; weight ratios of sodium hydrogen carbonate to
fluidized pyroclastic material was poured into the sand, n 0:007 0:067; and total amount of sand,
flume from the rotary mixing drum upstream, it W 5 15 kg; were used. The majority of experi-
formed a tongue-shaped front, kicked up a cloud ments were conducted under the fixed smooth bed
of dust from itself, and flowed very fast. The condition, but some were done under an erodible
comparatively steady flow that followed the front bed condition with the same silica sand as the bed.
consisted of a low, dense main body layer and No erosion by flow occurred in these experiments
an upper, thinner hot ash cloud layer. It closely probably because the lowermost layer is not fluidized
resembled the structure of the actual flow that and the flow is laminar adjacent to the bed. Experi-
occurred at Unzen. mental observations showed that entrainment of air
The experimental flow was recorded from the side from the flow front was negligible; therefore, the
through a transparent wall by a high-speed video and reasoning of Wilson (1980) about the formation of a
by normal videos. Because the particle diameter of the fluidized bed in a pyroclastic flow that the entrapped
material was too small to trace its motion, freeze-dried air from the flow front expands to generate as
instant coffee particles were thrown into the flow as strong an upward flow as causing fluidization is
the tracer in order to measure vertical velocity distri- doubtful. As an evidence of this, an experimental
bution in the main body. run using mere hot sand without sodium hydrogen
The solids concentration distribution normal to the carbonate could not flow and was deposited near
bottom was measured by a device (Fig. 5) composed the supply point. As far as the experimental runs
of multi-layered cells and a sheath. The device had a are concerned, no difference in the structures of
rectangular tube shape 1.2 cm wide and 3 cm deep flow for different n values was evident; i.e. the
and was installed in the flume in advance without material was well fluidized even at the smallest
the sheath. It allowed the flow front to pass freely value of n 0:007:
100 T. Takahashi, H. Tsujimoto / Journal of Volcanology and Geothermal Research 98 (2000) 91115

3.2. Structure of the main body

Within the main body, upward gas flow occurs due


to gas emission. Because the uplift force exerted by
the gas flow is not sufficient to sustain all the weight of
the particles in the lower part of the main body, some
part of the external body force is balanced with the
quasi-static skeleton force. This lower part is here
called the bottom layer. In this layer particles
move longitudinally and are in enduring contact.
The solids concentration in the bottom layer should
decrease upwards from the stable bed value, C, to
that of the initiation of fluidization, Cmf, at the upper
boundary. Beyond this boundary, upward gas flow is
sufficient to support all the weights of the particles,
and the fluidized layer develops. Our experiments
showed that the solids concentration throughout the Fig. 6. Structure of the main body. The hatched section shows the
freshly deposited layer that is separated from the main body.
entire fluidized layer remained almost constant and
was approximately equal to Cmf. For irregular parti-
cles, C also is approximately equal to Cmf, therefore z-direction is balanced with gas production rate at
the entire main body has a nearly uniform vertical height z, and Eq. (17) means the load of particles is
distribution of its solids concentration, 4045% in supported by the sum of the static pressure and the lift
volume as obtained in our experiments (Takahashi force due to upward gas flow.
et al., 1995). For the uplift force, Erguns formula is applicable
The schematic structure of the main body, in which (Bird et al., 1960):
plane z zb is the surface of the deposit left by the  2
main body and z 0 is the original ground surface, is dD 150mw C
18
shown in Fig. 6. The layer beneath height z zb is dz f2p dp2 1 C
now stable, but it emits gas and may be eroded by the
where m is the viscosity of the gas and f p the particle
subsequent flow. The amount of gas emitted from the
shape factor.
deposit layer may affect the fluidity of the flow. The
Eliminating w and dD=dz from these equations
meaning of the height h0 in the figure is explained
under the boundary condition z 0; w 0 gives:
later.
 3
The height of the upper surface of the bottom layer, dps 150bns C
Csg cos u z 19
h, is obtained as follows: dz f2p dp2 1C
The conservation of mass for the emitted gas is
written: where n is the kinetic viscosity of the gas. If we
assume that at z h; dps =dz 0; Eq. (19) gives:
d
{r1 Cw} bsC 16 f2p 1 C3 gdp2 cos u
dz h 20
150C bnC
and the balance of force in the z direction is:
and at the solids concentration of Cmf:
dp dD
sCg cos u s 0 17 Cmf
dz dz 11:1f2p 21
1 Cmf 3
where r is the gas density, b the rate of gas emission
is satisfied (Shirai, 1973), Eq. (20) becomes:
from the unit mass of the solids, w the upward gas
velocity, and D the uplift force due to upward gas gdp2 cos u
flow. Eq. (16) means the change in gas flow rate in h 22
1650bnC
T. Takahashi, H. Tsujimoto / Journal of Volcanology and Geothermal Research 98 (2000) 91115 101

3.3. Stress and velocity distributions in the main body and


s
 
If the solids concentration in the main body is z 2
assumed to be constant, the operating shearing stress Fu 1 1 30
h
in the flow due to gravity is given by:
for the bottom layer.

t sCgH zsin u 23 The solution of Eq. (29) for the bottom layer under
the condition z 0; u 0 is:
The internal shearing stress in the bottom layer would 8 9
  s
 
be the sum of the turbulent mixing stress and the u < 1 z z 2=
u cos 1 1 1 31
quasi-static Coulomb friction stress, and it is given as: k: h h ;
 2
 2 du
t sC` ps tan fs 24 The velocity at the boundary between the bottom
dz layer and the fluidized layer, uh, is from Eq. (31):
 
and that in the fluidized layer would be given by only u p
the mixing stress: uh 1 32
k 2
 2
 2 du
t sC` 25
The velocity distribution in the fluidized layer is
dz given by solving Eq. (29) under the conditions Fu
1; and z h; u uh ; as:
where ` kz zb is the mixing length, and k the  
u z p
von Karman constant. Collision stress is neglected u ln 1 33
because in the fluidized flow Hf =dp is so large that k h 2
the effect of particle collision becomes small. The mean velocity of the entire main body, Uf,
Integrating Eq. (19) under the conditions z therefore is:
h; ps 0; and using Eq. (20) gives: ( ! )
  Uf 1 p2 p h h p
z 2 1 ln 34
1 
ps sCgh 1 cos u 26 u k 8 2 H H 2
2 h
We assume Eq. (34) is approximately applicable even
The height h0, at which the driving and resistance when zb is not negligible, therefore the equation for
forces balance, is given as the solution of the equation: the general case is:
( ! )
C sgH z ps tan fs 27 Uf 1 p2 p hf hf p
1 ln 35
u k 8 2 Hf Hf 2
Substitution of Eq. (26) into Eq. (27) gives:
" s
  # Fig. 7 compares the velocities in the experiments
tan u tan u tan u H with those of Eqs. (31) and (33), in which C 0:42
h0 h 1 2 1
tan fs tan fs tan fs h and b 0:004 s1 : The theory well explains the
28 experimental results. In this figure, Eq. (9) gives the
granular flow corresponding to the experimental
If h0 0; the flow continues to deposit material, and condition H=dp 500: This result clearly shows that
if h0 0 erosion may occur, depending on the bed the granular flow model is not applicable for the flui-
condition. In our experiments no erosion took place. dized layer. The normal turbulent fluid model such as
For the sake of simplicity if h0 0 is assumed, Eqs. that for plain water might be considered applicable.
(23)(25) and (28) give: To explain the experimental tendency by the ordinary
du u logarithmic velocity distribution function that is
Fu 29 applicable to the normal turbulent flow, however,
dz kz
p u H=na 10 should be adopted, where n a is the
where u gH sin u; Fu 1 for the fluidized layer apparent kinematic viscosity of the material. This
102 T. Takahashi, H. Tsujimoto / Journal of Volcanology and Geothermal Research 98 (2000) 91115

Fig. 7. Velocity distributions in the main body obtained experimentally and theoretically. The various symbols represent experimental cases. Eq.
(9), which is applicable to granular flow, can not explain the velocity of the fluidized pyroclastic flow.

condition requires the apparent viscosity of the mate- 4. Deposition of the main body
rial to be ma 0:6 Pa s; 600 times that of water.
Clearly, that is too big because the experimental pyro- 4.1. Experimental observations
clastic material in the container has virtually no resis-
tance when stirred with a stick. Fig. 8 shows temporal variations in the surface
The flow model of the main body discussed above was heights of the main body and the deposit 250 cm
obtained on the assumption that a fluidized upper layer downstream from the supply point. This figure and
exists on the bottom layer. Actually, under some condi- other experimental results demonstrate the following
tions of gas emission and the particle diameter such as the characteristics:
case of inactive fragmentation, h can be larger than H, 1. After about the moment of maximum flow depth,
evidence that skeleton stresses exist throughout the the bottom of the flow begins to rise at approxi-
entire main body. In this case instead of Eq. (18): mately constant speed.
!3 2 2. The rise speed for the deposit surface is small when
150b ns 
C H z2 the gas emission rate is large.
ps C sgH zcos u 
fp d p
2 2
1C 2 3. The rise speed of the deposit surface is small when
36 the channel slope is steep.
4. The bulk of the deposit at the end of an experimen-
is obtained. The mean velocity then is: tal run is thicker when the gas emission rate is
! small and the channel slope is flat.
Uf 1 p2
1 37
u k 8
4.2. Mechanical model of deposition
The height of the stress equilibrium plane in this
case is: Deposition takes place when the driving force
becomes smaller than the resistance force. Because
gdp2 f2p cos u 1 C  3 tan u

height h0 in Fig. 6 is the position on the plane of
h0 H 1
75nb C 2 tan fs equilibrium between these two forces, the part
38 beneath h0 should be deposited in time. As stated
earlier, deposition does not occur suddenly, it
and h0 is smaller than H. proceeds gradually; i.e. it is an inertial motion
Even in this case a hot ash cloud layer will develop process. Consider the motion in Fig. 9 of the hatched
immediately on the bottom layer (Fisher, 1995), but volume that begins to deposit. The equation of motion
the thickness of it will be small. The arrival distances for this volume is:
of the main body as well as the hot ash cloud will be
short in comparison to the pyroclastic flow having a dy dm
fluidized layer. m F u 39
dt dt d
T. Takahashi, H. Tsujimoto / Journal of Volcanology and Geothermal Research 98 (2000) 91115 103

Fig. 8. Example of temporal changes in the heights of the upper (H) and lower (zb) boundaries of the main body in an experimental case for the
channel slope u 5:5; amount of pyroclastic material 7.5 kg, and the weight ratio of sodium bicarbonate to sand n 0:0167; 250 cm
downstream of the supply point. The vertical line at t 2:7 s shows the time of the end of passage of the main body at that position, after
which about a 1.3 cm-thick deposit remains.

where m is the mass of the hatched volume, v the surface and the apparent density of the hatched
velocity of the mass, F the external force that operates 
volume, sC;
on the volume, and ud the relative velocity of the
dm dz
depositing part observed from the main flow. sC b Dx 41
dt dt
The external force is the sum of the shearing force
working on top of this volume, the gravity force oper- If v is assumed to be nearly equal to the velocity of
ating on the volume, and the friction force at the the main flow, U0, and the particles separated from the
bottom: volume are deposited immediately on the bottom,
  then v becomes equal to U0. Moreover, if the velo-
 z 2
F 12 sCgh 1 b cos u tan fs Dx city of the volume is assumed to change linearly until
h 40 it stops when zb coincides with h0, the rate at which the
sH zb Cg sin uDx depositing surface rises, if, is:
  
1 zb 2
Change in the mass of the hatched volume is if 1
gh cos u tan fs 1
given by the product of the rise rate of the deposit 2U0 2 h
 42
gH zb sin u

Similarly for h H; the entire main body is


composed of the bottom layer:

H zb
if g sin u g tan fs cos u
2U0

150nb tan fs C 2 H zb
 3 43
f2p dp2 1 C 2

If deposition also occurs in the granular flow stage due


Fig. 9. Model for the deposition process of the main body. The to imbalance between the driving and resistance
hatched section shows the volume that begins to be deposited. forces, the deposition speed can be similarly
104 T. Takahashi, H. Tsujimoto / Journal of Volcanology and Geothermal Research 98 (2000) 91115

Fig. 10. Calculated results for temporal changes in the various boundaries of the main body.

estimated. In the granular flow stage, however, diluted gaseous body. A three-dimensional solution
because some parts of the particle load are supported of the equations for the conservation of mass,
by the dynamic collision force, Eqs. (42) or (43) will momentum and heat energy is necessary to accu-
give too large a deposition speed. Then, referring to rately simulate the hot ash cloud. Because of insuffi-
Eq. (7): cient knowledge about the structures and
thermodynamics of this layer, at present such a
i g as i f 44
calculation is impossible. Therefore a planar two-
is used where ig is the deposition speed in the granular dimensional model is introduced in which three-
flow stage. dimensional characteristics are taken into account as
Fig. 10 shows results calculated for a position much as possible. The frame of reference used in
300 cm downstream of the supply point, at which constructing the model includes:
point the channel slope is 14.5, and there a constant
flow with Hf 3:0 cm and U0 170 cm=s for 5 s. 1. Adoption of the planar two-dimensional model to
Because the flow in the experimental flume is make possible delineation of a hazardous area.
unidirectional, a one-dimensional version of the simu- 2. Resistances at the boundary between the main
lation model that appears later in this paper was used body and the hot ash cloud layer, and between
for the calculations to obtain Fig. 10. Comparison of the ambient air and ash cloud are considered
this figure with Fig. 8 leads to a conclusion that the effective.
deposition of the main body is well explained by this 3. The supply of gas from the main body and
model. entrainment of air from the circumambient
atmosphere are taken into account in the
conservation of mass equation.
5. Modeling of the hot ash cloud layer 4. The hot ash cloud travels farther after the main
body stops, but soon loses horizontal momen-
The hot ash cloud layer is formed by a supply tum, then ascends vertically and diffuses
of hot gas and fine particles from the main body. because of the lack of particles supplied by
Entrainment of ambient air also is an important the main body. Therefore, if the bulk density
factor, because the apparent density of the hot of the hot ash cloud becomes smaller than the
ash cloud layer is close to that of ambient air ambient atmosphere, plumes are considered to
and the upper boundary of this layer is very decouple from this layer.
much turbulent. The video images taken at Unzen 5. Under the law of the conservation of particles,
Volcano confirm the violently billowy rising the supply from the main body, settling due to
plume. The cause of this plume is the ascending the effect of gravity, and dispersion caused by
flow produced by the lightness of the heated and turbulence are considered. Only the particles
T. Takahashi, H. Tsujimoto / Journal of Volcanology and Geothermal Research 98 (2000) 91115 105

whose diameter is smaller than 0.1 mm are thought where Hf is the flow depth of the main body, Uf and Vf,
to be supplied by the main body through respectively, are the x and y components of the mean
elutriation. velocity, u bx0 and u by0 the slope angles of the x and y
6. The hot ash cloud is considered to be a continuous axes (laid on the original ground surface), zb is the
fluid, and, for the sake of simplicity, the solids height of the deposit surface, r m the apparent density
concentration in this layer is assumed uniform in of the main body, t bx and t by are the shearing stresses
the vertical direction. at the flow bottom operating in the x and y directions,
n is the particle number per unit volume and is equal
to nb in the depositing process, if the depositing velo-
6. Two-dimensional simulation city, and In given by Eq. (14). The left-hand sides of
Eqs. (46) and (47) are the terms of momentum change,
6.1. Fundamental equations for the granular flow the first terms of the right-hand sides are the driving
stage and for the main body in the fluidized flow force due to gravity, the second terms are the pres-
stage sures in the direction of flow due to surface slope of
flow, and the last terms are the bed friction.
The fundamental equations for planar two-dimen- Note that momentum conservation Equations (46)
sional flow are and (47) neglect the horizontal stress term induced by
differences in Uf and Vf in the y- and x-directions,
Conservation of mass:
respectively. These terms are small and the applicabil-
2Hf 2Uf Hf 2Vf Hf ity of these momentum equations for water flood
if 45 flows (Takahashi and Nakagawa, 1987) and for debris
2t 2x 2y
flows (Takahashi, 1991) is obvious.
The shearing stresses at the bottom of flow are
Conservation of momentum for the x-direction:
given by
2Uf Hf 2Uf2 Hf 2Uf Vf Hf Granular flow stage: dp dfc

2t 2x 2y q
2Hf zb t tbx as rm gHf cos ux tan fs rm fUf Uf2 Vf2 50
gHf sin ubx0 gHf cos ubx0 bx
2x rm
q
46 tby as rm gHf cos uy tan fs rm fVf Uf2 Vf2 51
Conservation of momentum for the y-direction:
Fluidized flow stage: dp dfc
2Vf Hf 2Uf Vf Hf 2Vf2 Hf q

2t 2x 2y tbx rm fUf Uf2 Vf2 52

2Hf zb tby q
gHf sin uby0 gHf cos uby0
2y rm tby rm fVf Uf2 Vf2 53
47
where u x and u y are the x- and y-components of the
Conservation of particle number per unit volume: slope angles of the surface of the deposit, and f is the
2n Hf 2n Uf Hf 2n Vf Hf resistance coefficient of the main body that is given by
earlier discussions as
2t 2x 2y
Granular flow stage: dp dfc (from Eq. (10))
if nb In Hf 48
 
25 K dp 2
Variation in ground surface height: f 54
4 C Hf
2zb
if 0 49 Fluidized flow stage: dp dfc (from Eq. (35) and
2t
106 T. Takahashi, H. Tsujimoto / Journal of Volcanology and Geothermal Research 98 (2000) 91115

Eq. (37)) upper surface of the hot ash cloud, we2 the gas
8 ( ! )2 supply velocity from the main body, and we3
>
> p2 p hf hf p
>
> k2
1 ln ; hf Hf the decoupling velocity of the ash cloud layer
>
< 8 2 Hf Hf 2
f
as a plume that becomes lighter than the ambient
> !2
>
> p2 air.
>
>k 1
2
; hf Hf
: 8 The momentum conservation equations for the x-
and y-directions are:
55
where
p 2Up Hp 2Up2 Hp 2Up Vp Hp
4 151 eC 2 g0 1
K p p 56 2t 2x 2y
75 p1 e 3g0 15p1 e
gHp2 2rmp
The average solids concentration for the granular flow gHp sin ubx0
2rmp 2x
stage is given by Eq. (12), and for the fluidized flow
stage approximately 0.42.  
2zb Hp 2 2Up Hp
The deposition velocity, if, is given by Eqs.(42), gHp cos ubx0 1p
2x 2x 2x
(43), or (44) depending on the flow stage and degree  
of fluidization. 2 2Up Hp
1p fi Up2 fb Up Uf Up
The representative diameter, dp, in the main body is 2y 2y
obtained from the relationship:
!1=3 Uf 60
6C
dp 57
pn
2Vp Hp 2Up Vp Hp 2Vp2 Hp

6.2. Fundamental equations for the hot ash cloud 2t 2x 2y
layer
gHp2 2rmp
The schematic structure of the hot ash cloud layer gHp sin uby0
2rmp 2y
and the notation are shown in Fig. 11.
The mass conservation equation is: 2zb Hp
gHp cos uby0
2Hp 2Up Hp 2Vp Hp 2y
ve vf 58
2t 2x 2y    
2 2Vp Hp 2 2Vp Hp
where we is the rate at which the upper surface of the 1 1
2x p 2x 2y p 2y
hot ash cloud layer rises, and wf the rate of rise of the
boundary between the main body and hot ash cloud fi Vp2 fb Vp Vf Vp Vf (61)
layer. Because the bulk of the main body is far thinner
than that of the ash cloud layer, in the discussion of where 1 p is the coefficient of the eddy viscosity, fi the
the motion of the ash cloud layer, wf is considered to resistance coefficient at the upper boundary of the
be zero. layer, fb the resistance coefficient at the lower bound-
The rising velocity of the upper surface of the ash ary, and r mp the apparent density of the hot ash cloud
cloud layer should be written: layer:

we we1 we2 we3 59 rmp Cp s 1 Cp ra 62

where r a is the density of the gas phase and Cp the


where we1 is the air entraining velocity at the solids concentration in the hot ash cloud layer.
T. Takahashi, H. Tsujimoto / Journal of Volcanology and Geothermal Research 98 (2000) 91115 107

Fig. 11. Structure of the hot ash cloud and notations.

Conservation of solids phase is: modeled by referring to previous investigations, and


the other quantities are obtained in the process of
2Cp Hp 2Up Cp Hp 2Vp Cp Hp calculations.

2t 2x 2y
 cu maxws we2 ; 0Cp 6.2.1. The air entraining velocity, we1
maxwe2 ws ; 0Cf
Fukushima and Kaneko (1997) give the formula:
   q
2 2Cp Hp u
maxwe3 ws ; 0Cp K ve1 a Up2 Vp2 64
2x x 2x 90
 
2 2Cp Hp where u is the gradient of the slope () on which the
Ky 63
2y 2y pyroclasts flow (as shown in Fig. 12) and a is a coef-
ficient, which they give as a 0:1 0:4: Here,
where fcu is the ratio of the particles fraction transpor- however, by the reason explained later, we consider
table to the ash cloud layer from the main body, Kx u to be the gradient of the surface of the hot ash cloud
and Ky are the coefficients of diffusion in the x and y in the direction of the velocity vector:
directions, and ws is the setting velocity of the particle.
Note that the transport of particles by the eddy diffu- Uf sin ubxh Vf sin ubyh
tan u q 65
sivity in the vertical direction is neglected in compar- Uf2 cos2 ubxh Vf2 cos2 ubyh
ison to the transport by the elutriation as explained
earlier. Because Hf and zb are considered far smaller than
To obtain Up, Vp, Hp, Cp from these equations, the Hp, tan ubxh 2z0 Hp =2x; and tan ubyh
values of we1, we2, we3, 1 p, fi, fb, fcu, ws, Kx and Ky must 2z0 Hp =2y are approximately satisfied, when
be given. Of these, we1, we2, we3, 1 p, fcu, and ws are z0 is the height of the ground surface.
108 T. Takahashi, H. Tsujimoto / Journal of Volcanology and Geothermal Research 98 (2000) 91115

ignored, the vertical motion of air is described by:


Dw 1 2p
g 68
Dt ra 2z
where w is the vertical velocity of the air and p the
pressure. Separating the variables into those of the
fundamental field and disturbance:

w w w 0 ; ra r r 0 69
Fig. 12. Slopes used to calculate air entrainment in Fukushima and and assuming that static equilibrium conditions are
Kaneko (1997) and in this paper.
satisfied in the fundamental field:
1 2p dw
The reasons why such a modification in u is used g 0; 0 70
r 2z dt
are that the actual forefront of the hot ash cloud is like
a very steep cliff, and the entrainment of air at the Then, Eqs. (69) and (70) and the equation of state
forefront seems to be very large. give:

Dw 0 Ta Tp
6.2.2. The gas supplying velocity, we2 g 71
The gas production rate per unit time for a unit area Dt Ta
of the main body is C sHf b; therefore the gas supply where Ta is the temperature of the circumambient atmo-
velocity is given by: sphere (K). From this equation we3 is approximated
C sHf b by:
we2 66 s
ra
2gHp 0:5Tp Ta
we3 72
where r a depends on the temperature Tp (K) and pres- Ta
sure Hpa (hPa) in the ash cloud layer:
In this equation 0.5Tp is used instead of Tp to account
1:293 Hpa for the upward temperature drop.
ra : 67
1 0:00367Tp 273 1013 If the average temperature of the hot ash cloud is
200C and that of the atmosphere is 20C, Eq. (71)
gives the upward acceleration, 6 m/s 2. According to
6.2.3. The decoupling velocity, we3 the VTR analysis of Ishikawa et al. (1994a), it is about
Eqs. (62) and (67) suggest that, if the temperature 1 m/s 2. The simplified theory introduced here
of the hot ash cloud is assumed to be 700C, the provides a reasonable approximation. In the numeri-
density of the ash cloud becomes almost the same as cal calculation, we3 is considered only when r mp
the ambient atmosphere when the solids concentration becomes smaller than the ambient air value.
is about 0.03% in volume. There are few data on the
density of hot ash clouds, but Ishikawa et al. (1994a) 6.2.4. Eddy viscosity, 1 p
estimated it, from a VTR of the Unzen pyroclastic Horizontal eddy viscosity is assumed to be similar
flow, as being less than 0.01% after stoppage of the to the vertical eddy viscosity of ordinal fluid flow
main body. Therefore, the ash cloud is likely to (Lane and Kalinske, 1941):
disperse violently into the sky especially after the
kup Hp
main body ceases to move. In fact cumulonimbus- 1p 73
like behavior even after longitudinal motion ceased 6
was conspicuous. The coefficients of dispersion Kx and Ky are assumed
If the viscosity and turbulent Reynolds stress are to have the same value as 1 p.
T. Takahashi, H. Tsujimoto / Journal of Volcanology and Geothermal Research 98 (2000) 91115 109

6.2.5. Particle entraining velocity from the main body, the upstream part was smoothed to a certain extent
Es using the formula:
If fine particles are assumed to be transported from
the main body to the ash cloud layer at a velocity bai; j bai; j 0:5 {bai 1; j bai 1
equal to the upward gas velocity, the following
bai; j 1 bai; j 1} 0:125
formula would be satisfied:
76
Es we2 C fcu 74
where ba(i,j) is the altitude of the grid (i, j) based on
The mean particle diameter is about 1 mm in the
digital data.
deposit of the pyroclastic flow at Unzen Volcano,
(2) Parameters and coefficients: Some of the many
whereas in the pyroclastic surge it is smaller than
parameters and coefficients in the equations could be
0.1 mm (Ishikawa et al., 1994b; Miyabuchi, 1999).
estimated from field investigations, others had to be
Particle size analysis of the pyroclastic deposit
determined by trial and error. The values estimated
shows that the content of 0.1 mm or finer particles is
from field investigations were the mean diameter of
only a few weight percent. Therefore, fcu would be of
particles in the ash cloud layer, dpf 0:1 mm; the
about the same order.
critical particle diameter that produced fluidization,
dfc 1:0 mm; and the temperature of the ash cloud,
6.2.6. Settling velocity of the particles
Tp 400C (Taniguchi et al., 1996). The values esti-
The setting velocity of the particles is given by
mated from results of laboratory experiments were the
Stokes law as:
kinematic viscosity of the volcanic gas, n
2   0:0002 m2 =s; the gas emission rate, b 0:0001=s;
gdpf s
ws 1 75 fs 38:5; C 0:65; as 0:3; Csmin 0:51; and
18n ra
mn 0:5: The values assumed for the trial case
where dpf is the mean diameter of the particles in the were k 0:7; b 0:02; fi 0:1; fb 0:1; and fcu
hot ash cloud layer, and n the kinematic fluid viscosity 0:005:
in the hot as cloud layer. (3) Boundary conditions: The mechanism of the
collapse of the lava dome is not clear, but we
6.3. Application to actual phenomena assumed that pyroclastic material having a net
volume of 500,000 m 3 was supplied within 100 s
The system of equations described was used to
at the volume concentration of 0.43 (bulk volume,
reproduce an actual case of Merapi-type pyroclastic
1:16 106 m3 ) from a grid near the dome that used
flow that occurred on 3 June, 1991, at Mt. Fugen,
in the calculation (width, 50 m). The discharge is
UnzenVolcano, Kyushu Japan and whose arrival
assumed to increase linearly to Qmax 23 250 m3 =s
limit is shown in Fig. 3. The volume of the collapsed
at 50 s then to decrease linearly to zero at 100 s.
lava dome in that pyroclastic flow was estimated to be
Although abundant fine ash material seems to be
about 500,000 m 3 (Nakada and Fujii, 1993).
produced on collapse, the initial particle diameter
just after collapse of the dome that would be
6.3.1. Conditions for calculation
effective for the granular flow stage is assumed to
(1) Topographical data: The topography of the
be 5 m.
Mizunashi River down which the pyroclastic flow
ran had been changing from day to day due to frequent
lava dome collapses and resulting deposition from 6.3.2. Results of calculation
pyroclastic flows. Therefore, there is no detailed topo- Temporal changes in the arrival limits of the main
graphical data available for that day. We used the body and hot ash cloud are shown in Fig. 13. Sixty
digital altitude data from the 50 m grids made by seconds after collapse of the dome the hot ash cloud
the Geographical Survey Institute, Japan before the has not yet developed. Presumably there was a short-
eruption. To reflect topographical changes before age of the fine material in the main body required to
that particular pyroclastic flow, the topography of form an ash cloud because up to about 30 s the
110 T. Takahashi, H. Tsujimoto / Journal of Volcanology and Geothermal Research 98 (2000) 91115
T. Takahashi, H. Tsujimoto / Journal of Volcanology and Geothermal Research 98 (2000) 91115 111

Fig. 14. Numerically simulated relationships between the time elapsed after the collapse of the lava dome and its position (shown by elevation),
velocity, and the mean diameter of the front of the main body.

granular flow stage continued. At 90 s, an ash cloud maintained at 2030 m/s until it arrives at the position
has developed on and around the main body and it of the first waterfall, after which it gradually
leads 200 to 300 m in advance of the main body. decreases. Actual velocity variation was not measured
This lead occurred due to mechanism of motion of in this particular pyroclastic flow, but on other occa-
the hot ash cloud independent of the main body. At sions similar figures were obtained (Ishikawa et al.,
150 s, the main body has decelerated in a winding part 1994a).
of the gully where the first waterfall is located (Fig. 3), Temporal variations in the velocity, thickness of the
whereas the ash cloud has continued to develop, one flow and position (altitude) of the front of the hot ash
part going over the right-bank ridge in the direction of cloud layer are shown in Fig. 15. The cloud occurs just
Akamatsudani and the other over the left-bank ridge after the end of the particle-crushing process, after
toward Oshigadani (arrows in the figure). At 240 s, the which movement gradually accelerates until it reaches
main body has virtually stopped at the debouchment the debouchment of the gully at the altitude of about
of the Mizunashi River, but the ash cloud has contin- 250300 m, at which point the maximum velocity is
ued on and spread in a fan-shape involving the left 7080 m/s, thereafter decelerating markedly.
bank community of Kitakamikoba where many For this particular pyroclastic flow, the velocity of
houses were destroyed by fire. After 240 s, the front the hot ash cloud was estimated from a video (Yama-
of the ash cloud no longer proceeds, and the width of moto et al., 1993) and from the positions of toppled
the flow gradually decreases. This is because of lack concrete electricity poles and trees (Ishikawa et al.,
of particles from the main body and detachment of the 1993b). On the basis of these data, the velocity ranged
material as a plume. from 20 to 60 m/s, confirming the validity of the
Fig. 14 shows temporal changes in the velocity, calculations. The calculated thickness of the devel-
maximum particle diameter, and position (altitude) oped hot ash cloud layer, about 50 m, agrees with
of the front of the main body. At the front, the maxi- values for other pyroclastic flows (Ishikawa et al.,
mum diameter becomes about 1 mm within 30 s, and 1994a).
the crushing process ends by that time. This may be The average particle concentration in the front part
too early, which suggests that a b value smaller than of the ash cloud and the total solids volume of the ash
0.02 should have been adopted. The front velocity is cloud layer are shown in Fig. 16. Except for the early

Fig. 13. Temporal changes in the arrival limits of the main body and hot ash cloud obtained by numerical simulation. The shaded area shows the
actual path of the main body. Closed circles show the calculated flow area of the main body, open circles represent the flow area of the hot ash
cloud at that time.
112 T. Takahashi, H. Tsujimoto / Journal of Volcanology and Geothermal Research 98 (2000) 91115

Fig. 15. Numerically simulated relationships between the time elapsed after the collapse of the lava dome and the position (shown by elevation),
velocity, and thickness of the flow of the hot ash cloud layer.

stage, the solids concentration is nearly constant, clastic flow. The actual area affected by the hot ash
between 0.02 and 0.03%. The reason for this is that cloud also is shown. The calculated hot ash cloud
if the concentration becomes smaller than this the ash spreads more laterally at the debouchment of the
cloud is detached as a plume. The maximum amount gully than does the actual one. As is evident in the
of solids is 16 000 m 3; 3.2% of the collapsed lava figure, however, the general tendencies of width and
dome material. Field data on particle size distributions length are well reproduced. Moreover, the deposit
in the deposit of the main body show that only a few to area of the main body in the calculated and actual
10% of the material constitutes the fraction of parti- results (shown by shading in Fig. 13) almost comple-
cles smaller than 0.1 mm. This indicates that both the tely overlap.
particle transportation model from the main body to Longitudinal distributions of the deposit and flow
the hot ash cloud and the adopted parameter values are thicknesses at 360 s, just before the end of pyro-
reasonable. clastic flow, are shown in Fig. 18. The majority of
Fig. 17 shows traces calculated for the main body the collapsed materials are deposited within the
and the hot ash cloud just after stoppage of the pyro- granular flow reach. The rest move as a pyroclastic

Fig. 16. Temporal changes obtained by simulation in the elevation of the front position, average particle concentration, and total solids volume
in the hot ash cloud.
T. Takahashi, H. Tsujimoto / Journal of Volcanology and Geothermal Research 98 (2000) 91115 113

flow producing a comparatively thin deposit. This


tendency is in accord with the actual situation.
Compared with Fig. 2, however, the deposit reach
for the granular flow obtained by calculation is too
short. The reasons for this are the error inherent in
Eq. (44) and the too large b value, as mentioned
before. Another reason may be the formation of a
very steep slope due to growth of the lava dome,
which has not been taken into consideration. As
shown in Fig. 2 virtually no deposition takes
place near the dome. To evaluate this last factor,
a calculation was made in which the slope
upstream from the altitude of 8501250 m was made
steeper than the original slope. This showed that the
deposition rate there was significantly decreased.
Because exact altitude data on the day of the pyro-
clastic flow is not available, no further calculations
can be made.

7. Conclusion

Merapi-type pyroclastic flow processes are consid-


ered to begin as a granular flow, followed by fragmenta-
tion of particles within that flow and emission of gas
from the material itself, with a gradual change in the
flow structure to a fluidized flow in which the resistance
to flow is minimal. Upward gas flow, which becomes
significant in the final stage of the granular flow, trans-
Fig. 17. Calculated and actual flow limits for all the pyroclastic flow
processes.
ports fine material upward, and the fine material now
became millimeter-sized continues to run down leav-
ing the granular flow deposit upstream. Thus, the main

Fig. 18. Thickness of the deposits and flowing layer at time 360 s. Only a very thin flowing layer is present downstream of 1500 m at this time.
Note that the thickness depicted is 20 times the actual one.
114 T. Takahashi, H. Tsujimoto / Journal of Volcanology and Geothermal Research 98 (2000) 91115

body of the pyroclastic flow in a narrow sense is the parameter values are still difficult to determine
formed. The finer particles elutriate from the main theoretically; e.g. the adopted b value seems to be
body and by the drag force of the upward gas flow too large. Theoretically, although the difference in b
and the flow soon is separated into a lower dense main values between 10 4 and 10 5 does not produce large
body an upper light hot ash cloud layer. The hot ash differences in the solids concentrations, the adopted
cloud entrains ambient air, and when it becomes b 0:0001=s could be larger than the actual value by
lighter than the circumambient atmosphere, it about one order, because for degassing to continue at a
detaches from the ash cloud as a plume. Meanwhile, constant rate within about 200 s, the water content in
in the lower part of the main body deposition pyroclastic material with b 0:0001=s must be about
proceeds, and when that body reaches the gently slop- 2 wt%, whereas in the actual material it is at most
ing area it stops. The hot ash cloud layer travels inde- 0.5 wt% (Kusakabe et al., 1999). Further investiga-
pendently for a while after the main body stops, but it tions that focus on these problems therefore are
soon ceases due to the lack of material from the main needed.
body and it detaches as a plume.
The fundamental mechanisms of granular flow: i.e.,
Acknowledgements
constitutive relations, resistance to flow, equilibrium
particle concentration, and the fragmentation process
The critical comments and reference papers
caused by inter-particle collisions were discussed in
supplied by reviewers Prof. H. Sato and Prof. R.V.
terms of our recent investigations.
Fisher are highly appreciated. The editor Prof. Y.
Characteristics of the fluidized flow stage: i.e.,
Ida also gave useful comments. Prof. K. Ishihara
separation into an incompletely fluidized bottom
kindly supplied the copies of some reference papers.
layer and an upper fluidized layer, the stress and velo-
Y. Satofuka, M. Kondo and R. Takeuchi contributed
city distributions in the main body, and the deposition
for some parts of the investigations including the diffi-
process were discussed theoretically. The theories
cult and dangerous experiments.
were examined satisfactorily using laboratory experi-
ments specially devised to emit gas from the material
itself. References
Based on the theoretical models for granular flow
stage and fluidized flow stages, a horizontally two- Bird, R.B., Stewartt, W.E., Lightfoot, E.N., 1960. Transport
dimensional numerical simulation model for the Phenomena. Wiley, New York, 780 pp.
Boudon, G., Camus, G., Gourgaud, A., Lajoie, J., 1993. The nuee-
main body was introduced. Next, particle and gas ardente deposits of Merapi volcano, Central Java, Indonesia:
transportation from the main body to the hot ash stratigraphy, textural characteristics, and transport mechanism.
cloud layer, air entrainment from the circumference, Bull. Volcanol. 55, 327342.
the settling of particles, and detachment from the ash Denlinger, R.P., 1987. A model for generation of ash clouds by
cloud as a plume were considered in a numerical pyroclastic flows, with application to the eruptions at mount
St. Helens, Washington. J. Geophys. Res. 92 (B10), 10 284
simulation model for the hot ash cloud layer. In the 10 298.
actual computation, the models for the main body and Fujii, T., Nakada, S., 1999. The 15 September pyroclastic flows at
ash cloud can be coupled. Unzen volcano (Japan): a flow model for associated ash-cloud
This newly developed numerical simulation model surges. J. Volcanol. Geotherm. Res. 89, 159172.
that is faithful to the actual structure of the Merapi- Fisher, R.V., 1995. Decoupling of pyroclastic currents: hazards
assessments. J. Volcanol. Geotherm. Res. 66, 257263.
type pyroclastic flow and having theoretical back- Friedlander, S.K., 1983. Science of Aerosol (Japanese translation).
ground as described by many formulae in this paper Sangyo tosho, Tokyo, 312 pp.
was used to reproduce an actual pyroclastic flow that Fukushima, Y., Kaneko, Y., 1997. Decelerating inclined thermals
occurred at Unzen Volcano, Kyushu Japan. All the with suspended solid particles. Annu. J. Hydraulic Eng., JSCE
processes were fairly well reproduced. 41, 537542 (in Japanese with English abstract).
Ishikawa, Y., Yamada, T., Yajima, S., 1993. Comparison of depos-
The model lacks the process of the beginning of the its obtained by numerical simulation with the actual conditions
collapse of the lava dome, and it can not predict when for pyroclastic flow at Unzen volcano. Civil Eng. J. 35-6, 2733
and how a large pyroclastic flow will occur. Many of (in Japanese).
T. Takahashi, H. Tsujimoto / Journal of Volcanology and Geothermal Research 98 (2000) 91115 115

Ishikawa, Y., Yamada, T., Yajima, S., Shimoda, Y., 1993. Hot ash and generation of pyroclastic flows at Unzen volcano. Nature
cloud damage from pyroclastic flows at Unzen volcano. J. 360, 664666.
JSECE 46-4, 39 (in Japanese with English abstract). Savage, S.B., 1984. The mechanics of rapid granular flows. Adv.
Ishikawa, Y., Yamada, T., Yajima, S., 1994a. Kinematic character- Appl. Mech. 24, 289376.
istics of the main body and hot ash cloud in a pyroclastic flow at Savage, S.B., 1988. Streaming motions in a bed of vibrationally
Unzen. Spec. Rep. on Volcanic Sabo Res. Mt. Unzen,.1, JSECE, fluidized dry granular material. J. Fluid Mech. 194, 457478.
pp. 2430 (in Japanese). Shirai, T., 1973. Fluidized Beds. Kagaku-gijutsu-sha, Tokyo,
Ishikawa, Y., Yamada, T., Kawashita, S., Matsumoto, T., Takaha- 236 pp. (in Japanese).
shi, T., Otsuzuki, H., 1994b. Sedimentary structure and physical Sparks, R.S.J., 1978. Gas release rates from pyroclastic flows: An
properties of pyroclastic flow deposits produced at Mt. Unzen assessment of the role of fluidization in their emplacement. Bull.
volcano on June 3, 1991, Spec. Rep. on Volcanic Sabo Res. Mt. Volcanol. 41, 19.
Unzen, 1, JSECE, pp. 7178 (in Japanese). Takahashi, T., 1991. Debris Flow. Balkema, Rotterdam, 165 pp.
Kanatani, K., 1978. A theory for the flow of granular materials. Takahashi, T., Nakagawa, H., 1987. Hazard zone mapping in
Proc. Fifth Symposium of Japanese Earthquake Engineering, respect to damage to wooden houses due to breaking of levee.
pp. 681688. Bull. DPRI, Kyoto Univ. 37-2, 5990.
Kitou, K., Hirano, M., Hashimoto, H., 1993. Characteristics of gran- Takahashi, T., Tsujimoto, H., 1997. Mechanics of granular flow in
ular flow in an inclined open channel. Proc. Hydraulic Eng., an inclined chute. J. Hydraulic, Coastal, Environ. Engng, JSCE
JSCE 37, 617622 (in Japanese with English abstract). 565/II-39, 5771 (in Japanese with English abstract).
Kusakabe, M., Sato, H., Nakada, S., Kitamura, T., 1999. Water Takahashi, T., Satofuka, Y., Kondo, M., Takeuchi, R., 1995. Fluid
contents and hydrogen isotopic ratios of rock and minerals dynamics of the pyroclastic flow. Proc. Int. Sabo Symp., Tokyo,
from the eruption of Unzen volcano Japan. J. Volcanol. 18.
Geotherm. Res. 89, 231242. Taniguchi, H., Aramaki, S., Suzuki-Kamata, K., Umakoshi, K.,
Lane, E.W., Kalinske, A.A., 1941. Engineering calculations of Nakada, S., 1996. Temperature of pyroclastic flows and volca-
suspended sediment. Trans. AGU 22, 603607. nic disasters. J. HTSJ 35, 1120 (in Japanese).
Lowe, D.R., 1976. Grain flow and grain flow deposits. J. Sediment. Ui, T., Matsuo, N., Sumita, M., Fujinawa, A., 1999. Generation of
Petrol. 46, 188199. block and ash flows during the 19901995 eruption of Unzen
Matsumura, K., Takahama, J., Shima, D., Michiue, M., Miyamoto, Volcano, Japan. J. Volcanol. Geotherm. Res. 89, 123137.
K., Hokuei, M., 1997. Experimental study of dry granular flow. Valentine, G.A., 1987. Stratified flow in pyroclastic surges. Bull.
Proc. JSECE Conf., 180181 (in Japanese). Volcanol. 49, 616630.
Miyabuchi, Y., 1999. Deposits associated with the 19901995 erup- Wadge, G., Jackson, P., Bower, S.M., Woods, A.W., Calder, E.,
tion of Unzen volcano, Japan. J. Volcanol. Geotherm. Res. 89, 1998. Computer simulations of pyroclastic flows from dome
139158. collapse. Geol. Res. Lett. 25 (19), 36773680.
Nakada, S., Fujii, T., 1993. Preliminary report on the activity at Wilson, C.J.N., 1980. The role of fluidization in the emplacement of
Unzen volcano (Japan), November 1990November 1992: pyroclastic flows: an experimental approach. J. Volcanol.
Dacite lava domes and pyroclastic flows. J. Volcanol. Geotherm. Res. 8, 231249.
Geotherm. Res. 54, 319333. Yamamoto, T., Takarada, S., Suto, S., 1993. Pyroclastic flows from
Sato, H., Fujii, T., Nakada, S., 1992. Crumbling of dacite dome lava eruption of Unzen volcano, Japan. Bull. Volcanol. 55, 166175.

Potrebbero piacerti anche