Document Actions

IJEGE-11_BS-Avolio-et-alii

background image
Italian Journal of Engineering Geology and Environment - Book www.ijege.uniroma1.it © 2011 Casa Editrice Università La Sapienza
387
DOI: 10.4408/IJEGE.2011-03.B-044
DEBRIS FLOWS SIMULATION BY CELLULAR AUTOMATA:
A SHORT REVIEW OF THE SCIDDICA MODELS
M.V. AVOLIO
(*)
, F. BOZZANO
(**) (***)
, D. D’AMBROSIO
(*)
, S. DI GREGORIO
(*),
V. LU-
PIANO
(****)
, P. MAZZANTI
(***) (**)
, R. RONGO
(****)
& W. SPATARO
(*)
Department of Mathematics, University of Calabria, Italy.
(**)
Department of Earth Sciences, Sapienza Università di Roma, Italy.
(***)
NHAZCA S.r.l., spin-off Sapienza Università di Roma,
(****)
Department of Earth Sciences, University of Calabria,
K
ey
words
: debris flows, modelling & simulation, cellular
automata, susceptibility analysis
INTRODUCTION
Numerical modeling is a major challenge in the
prevention of risks related to the occurrence of land-
slides. Debris flows (DF) (i
veRson
, 1997; H
unGR
et
alii, 2001) are one of the most common and dan-
gerous types of landslides and may be classified as
complex systems of fluid-dynamical type. DF are
extremely rapid channeled landslides, composed of
slurry of rock, mud, organic matter and water, usually
originated by soil detachments in relation with intense
rainfalls, snow melt etc. DF may occur in subaerial,
subaqueous or mixed environment and their size can
vary from 10
2
to 10
9
m
3
(J
akob
, 2005). Nevertheless,
the final volume of a DF is significantly larger than
the detached volume (also 100 times larger) due to
the erosion and consequent entrainment of material
along the path (H
unGR
et alii, 2005). Furthermore, in
some cases the intense erosion can trigger secondary
landslides along the channel, thus furtherly increasing
the final volume of mobilized material. Hence, rheo-
logical properties (m
aJoR
& P
ieRson
, 1992) of initial
heterogeneous mixture of soil and water may change
during propagation by water loss or inclusion. Addi-
tional mechanisms, such as impulsive loss of matter
(water and finer grains) and energy dissipation at im-
pact, must be considered in the cases of coastal DF.
Differently, buoyancy effects, drag forces and peculiar
ABSTRACT
Cellular Automata models are a promising solu-
tion for the simulation of debris flows moving over
a 3D topography. In this paper, an extensive review
of SCIDDICA, a Cellular Automata model based on
the equivalent fluid approach, is presented. Moreover,
the main steps in the development of SCIDDICA are
described with a chronological criterion. The last ver-
sion of SCIDDICA (SS2) is suitable for the simulation
of completely subaerial, completely subaqueous and
combined subaerial-subaqueous debris flows. Main
features of a debris flow are accounted by the SS2 ver-
sion such as erosion and deposition and triggering of
secondary landslides along the path, presence of struc-
tures and buildings, run-up effects and, in the case of
coastal landslides, impulsive loss of matter (water and
finer grains) and energy dissipation at water impact.
Moreover, buoyancy effects, drag forces and peculiar
mechanisms like hydroplaning are also modeled for
submerged events. Several past debris-flows like the
1998-1999 Campanian debris flows (Italy), the 1997
debris flow at Lake Albano (Italy) and the 2008 sub-
marine debris flow at Bagnara Calabra (Italy) have
been simulated in the last years by SCIDDICA. A
short review of these case studies will be presented
and the main limitations encountered will be also
discussed, thus suggesting future improvements and
perspectives, such as susceptibility analysis and inter-
action with man made structures
background image
M.V. AVOLIO, F. BOZZANO, D. D’AMBROSIO, S. DI GREGORIO, V. LUPIANO, P. MAZZANTI, R. RONGO & w. SPATARO
388
5th International Conference on Debris-Flow Hazards Mitigation: Mechanics, Prediction and Assessment Padua, Italy - 14-17 June 2011
ly, new versions are generated step by step by intro-
ducing other “elementary processes” in order to model
more complex real cases.
A general description of the adopted approaches
in the most significant models will be given in this
paper, together with the results of their application to
real cases. However, the detailed description of physi-
cal laws controlling the models is not presented in this
review paper and we refer to specific references cited
in the text for this aspect.
Finally, we will discuss further perspectives of re-
search aimed at improving SCIDDICA and to extend
its application to DF analysis.
CELLULAR AUTOMATA: A SHORT OVER-
VIEW
CA are a paradigm of parallel computing for mod-
elling and simulating complex dynamical systems,
whose evolution depends mainly on the local inter-
actions of their constituent parts (e.g. Di Gregorio &
Serra, 1999). CA evolves in a discrete space-time con-
text; they are based on a regular division of the space
in cells (or, equivalently, a regular lattice, whose sites
correspond to the cell centres), each one embedding
an input/output computing unit: a finite automaton
(fa). S is the finite set of fa states, that the cell may
assume. The fa input of a cell c is given by the states
of m neighbouring cells, including the cell c. The
neighbourhood conditions are determined by a pat-
tern which is invariant in time and constant over the
cells. The fa have an identical state transition function,
which is simultaneously applied to each cell. At step
0, fa are in arbitrary states, describing the initial con-
ditions of the system; then, the CA evolves changing
the state of all cells simultaneously at discrete times
(CA step), according to the transition function
Hence, very complex behaviours can emerge by
relatively simple transition functions and few states.
Complex macroscopic phenomena like debris
flows need an extension of the original CA definition;
Macroscopic Cellular Automata were thus defined in
order to fit the modelling requirements of many mac-
roscopic phenomena, from a CA viewpoint, by con-
sidering:
a) the cell dimension, specified by the cell side,
and the CA step;
b) the state of the cell which accounts for several
mechanisms like hydroplaning play a significant role
in the submerged path.
The aforementioned features cannot be neglected
in the numerical simulation of a DF. However, many
problems arise due to the extreme complexity of these
events, which lead to difficulties in estimating kine-
matic geotechnical soil parameters for real phenomena
(a
nCey
, 2007). These features may be approximately
described in terms of fluid-dynamics by PDE (partial
differential equations), and several models have been
developed in the last years by using this approach
(H
unGR
, 1995; d
enlinGeR
& i
veRson
, 2001; m
C
d
ou
-
Gall
& H
unGR
, 2004; P
iRulli
& m
aRCo
, 2010).
A different methodology to approximately de-
scribe and model the main features of a debris flow
is represented by MCA (Macroscopic Cellular Au-
tomata). MCA are an extension of classical Cellular
Automata (CA), developed for overcoming some of
the limitation affecting conventional CA frames such
as the modeling of large scale complex phenomena.
Due to its particulate nature and local dynamics, MCA
are very powerful in dealing with complex bounda-
ries, incorporating of microscopic interactions, and
parallelization of the algorithm.
CA models for DF are based on the principle of
the equivalent fluid, formalized by H
unGR
(1995), stat-
ing that: “the flowing mass behaves like a fluid, whose
rheological features cannot be measured through labo-
ratory or in situ testing, but can only be obtained by
the back-analysis of real past events”.
MCA were proposed for the first time in 1982
to model the dynamics of macroscopic spatially ex-
tended systems, and firstly applied to the simulation
of basaltic lava flows (C
RisCi
et alii, 1982). Since
then, MCA were adopted for the simulation of diverse
natural phenomena: pyroclastic flows (a
volio
et alii,
2006a), snow avalanches (b
aRPi
et alii, 2007; a
vo
-
lio
et al., 2010b) and, in particular, flow type land-
slides (s
eGRe
& d
e
a
nGeli
1995; C
leRiCi
& P
eReGo
,
2000; a
volio
et alii, 2010a). Among existing Cellu-
lar Automata models, we will focus on SCIDDICA,
a family of deterministic MCA models, specifically
developed for simulating debris flows. This model has
been developed according to an incremental strategy,
permitted by the underlying CA properties, that allow
to build a model by the composition of “elementary
processes”. This permits to consider first models of
the family for less complex case studies. Subsequent-
background image
DEBRIS FLOWS SIMULATION BY CELLULAR AUTOMATA: A SHORT REVIEW OF THE SCIDDICA MODELS
Italian Journal of Engineering Geology and Environment - Book www.ijege.uniroma1.it © 2011 Casa Editrice Università La Sapienza
389
identifies the co-ordinates of cells, covering the
finite region, where the phenomenon evolves (N is the
set of natural numbers);
X is the neighbourhood relation (the cell and its ad-
jacent cells), SQ5 and HEX respectively for square and
hexagonal tessellation;
S is the finite set of states of the fa (the list of all
sub-states used in the different versions of SCIDDICA
is given in Tab. 1);
P is the finite set of parameters (e.g. pap and pt are
respectively the cell apothem and the time correspond-
ing to a step of the CA). Specific parameters will be
presented in the following sections;
τ is the CA transition function.
In the following, variables concerning sub-states
are indicated by their subscripts. When sub-states need
the specification of the neighborhood cell, an index at
subscript is used. Furthermore, nQ and ∆Q represent,
respectively, the new value and variation of the sub-
state S
Q
.
BASIC ASSUMPTIONS AND INITIAL CON-
DITIONS
Let us consider a ‘‘column’’ of mass m (with con-
stant density ρ), with base B (the area) and thickness
TH at altitude A (cf. example in fig. 1 for hexagonal
cells). If g is gravity acceleration, the sub-state total
energy, E, is defined as E=ρ-g-B-TH-(TH/2+A+kH),
while the sub-state run-up, R, is defined as R=TH+kH.
Velocity v is related to kH by KH=v
2
/2g (d’a
mbRosio
et alii, 2003).
Initial conditions are specified by initial values of
sub-states:
-
S
TH
is zero everywhere except for the detach-
ment area, where the thickness of landslide mass is
characteristics that are assumed to be relevant to the
evolution of the system (each characteristic corre-
sponds to a sub-state). Moreover, thanks to the sub-
state “altitude of the cell” the model can be considered
fully 3D.
c) the state transition function τ which must ac-
count for all the processes (physical, chemical, etc.)
that are assumed to be relevant to the system evolu-
tion.
Outflows from the central cell to the other n
neighbouring cells must be determined according to
the purely local CA approach in order to minimise the
differences in “height” in the CA neighbourhood. The
general problem may be stated for a generic quantity
q according to the following definitions:
q
d
= distributable quantity in the central cell;
q
0
= not distributable quantity in the central cell;
q
i
= quantity in the cell i 1≤i ≤n;
f
0
is the part of qd remaining in the central cell;
f
i
= flow from the central cell towards the cell i
1≤i ≤n;
q
i’
= q
i
+ f
i
0≤i ≤n with the bound q
d
= ∑
i
f
i
0≤i ≤n
The Algorithm for the Minimisation of Differences
(AMD) was determined (d
i
G
ReGoRio
& s
eRRa
, 1999)
in order to compute fi such that the summation of differ-
ences among all cells in the neighbourhood ∑
{(i,j)|0≤i<j≤n}
(|q
i
’- q
j
’|) is minimised. A surface flow evolves towards
minimum imbalance conditions in “height” at a lo-
cal level as a “pure” gravitational flow tending to the
hydrostatic equilibrium (Avolio et al., 2000). In this
framework, run-up effects of flow may be managed
by introducing a component of the kinetic head in the
“height” of the central cell (e.g., d’a
mbRosio
et alii,
2003b) while momentum may be managed by an op-
portune alteration of the “height” in cells along the mo-
tion direction (d’a
mbRosio
et alii, 2006). As a general
rule, the dynamics in CA is introduced by relatively
simple local laws, observing the conservation laws of
physics in an approximation context.
SCIDDICA MODEL: GENERAL FRA-
MEWORK
The SCIDDICA model can be defined as follows:
SCIDDICA = (R, X, S, P, τ)
where:
Tab. 1 - SCIDDICA sub-states.
background image
M.V. AVOLIO, F. BOZZANO, D. D’AMBROSIO, S. DI GREGORIO, V. LUPIANO, P. MAZZANTI, R. RONGO & w. SPATARO
390
5th International Conference on Debris-Flow Hazards Mitigation: Mechanics, Prediction and Assessment Padua, Italy - 14-17 June 2011
erage on the thickness for run-up values associated to
remaining debris in the central cell plus inflows. The
elementary process “energy dissipation” is specified by
a run-up reduction of a constant value (parameter rl):
nR
0
=R
0
-rl if rl>R
0
else nR
0
=0.
- Despite experiments carried out on real cases
were satisfactory, the above described versions were
not suitable for the simulation of debris flows which
are characterized by strong soil erosion and entrain-
ment along the landslide path. Hence, the (S
x
) family
versions were developed (d’a
mbRosio
et alii, 2002,
2003a; 2003b) and were firstly tested on the 1998-1999
Campanian debris flows event (i
ovine
et alii, 2003a;
d’a
mbRosio
et alii, 2007).
Version S
3-hex
(d’a
mbRosio
et alii, 2003b), was
characterized by the following main features: X is
HEX (for obtaining six flow directions that are the
maximum number permitted in a regular plane tes-
sellation), S=S
A
×S
TH
×S
O
6
×S
kH
×S
D
, AMD is specified
by q
0
=A
0
+kH
0
+adh, q
d
=T
H0
-adh, q
i
=A
i
+TH
i
1≤i≤6
and nTH
0
=TH
0
+∑
1≤i≤6
(I
i
-O
i
) is the new thickness. nE
0
is computed similarly, considering the balance of out-
flows (remaining energy) and inflows (acquired ener-
gy). The energy dissipation is the same as O version.
The erosion occurs when E>mt (where mt is a thresh-
old parameter); the eroded quantity of the soil cover is
-∆D=(E−mt)−er if (E−mt)−er<D else -∆D=D (where
er is the erosion parameter).
It must be pointed out that the parameter er in its
present form is an empirical parameter and is not re-
lated to any traditional governing equations for erosion
(e.g., e
GasHiRa
2007). A more rigorous treatment of this
parameter is one of the planned future improvements
of the model.
Version S
4
(d’a
mbRosio
et alii, 2006), holds the
general setting up of S3-hex, but improves the man-
agement of inertial effects by introducing indicators of
momentum S
px
and S
py
.
Velocity “indicators” v are deduced for cell inflows
and remaining debris in the cell, considering their as-
sociated kinetic head, by formula kH=v
2
/2g, and their
momentum “indicators” are deduced considering their
mass. Two components PX and PY (respectively along
axes x and y) are computed by the sum of all the contri-
butions. A “height” alteration, based on PX and PY, in
the neighbouring cells was introduced in AMD for S
4
in
order to opportunely account for inertial effects along
the motion directions.
specified;
-
S
A
and S
D
are the morphological height and
the initial depth of detrital cover, respectively. At this
regard must be pointed out that in the case of the de-
tachment area, the thickness of the landslide mass is
subtracted from the value of both the morphology and
detrital cover;
-
S
R
coincide with STH debris thickness;
-
E=ρ-g-B-TH-(TH/2+A) (potential energy
related to debris mass);
all the values related to the remaining sub-states
are zero.
EVOLUTION OF SCIDDICA OVER THE TIME
As discussed above, SCIDDICA has been devel-
oped according to an incremental strategy so that the
latest versions include all the improvement of previous
ones. In the following, a brief review of the evolution
of the model in time is presented.
- The first version (T) of SCIDDICA was a simple
MCA model with purely gravitational flows and it was
validated on the 1992 Tessina (Italy) earth flow (a
vo
-
lio
et alii, 2000), characterised by a velocity up to few
meters per day.
In this version X is SQ5, S=S
A
×S
TH
×S
O
4,
AMD
is specified by q
0
=A
0
+adh, q
d
=TH
0
-adh, q
i
=A
i
+TH
i
,
1≤i≤4 (where adh is the adherence parameter, i.e. the
thickness of the mud quantity, that becomes adherent to
soil). The balance of inflows and outflows in the cells is
defined as nTH
0
=TH
0
+
1≤i≤4
(I
i
-O
i
).
- The subsequent version (O) was modified in order
to simulate extremely rapid landslides characterized by
a run up effect, and was tested on the 1984 Mt. Ontake
landslide occurred in Japan (d
i
G
ReGoRio
et alii, 1999).
In this version X is SQ5, S=S
A
×S
TH
×S
O
4
×S
R
, AMD
is specified by q
d
=TH
0
−adh, q
0
=A
0
+R
0
-TH
0
+adh,
q
i
=A
i
+R
i
, 1≤i≤4 and nTH
0
=TH
0
+∑
1≤i≤4
(I
i
-O
i
). The re-
sulting new run-up is computed by the weighted av-
Fig. 1 - 3D visualization of some sub-states for a hexago-
nal cell together with an ideal section of a debris
flow along a slope
background image
DEBRIS FLOWS SIMULATION BY CELLULAR AUTOMATA: A SHORT REVIEW OF THE SCIDDICA MODELS
Italian Journal of Engineering Geology and Environment - Book www.ijege.uniroma1.it © 2011 Casa Editrice Università La Sapienza
391
of energy dissipation by turbulence is expressed by
-∆kH= tdkH, where td is the turbulence dissipation
parameter. The resulting new mass, barycentre and
energy (related to velocity) are computed by the com-
position of all the inflows from the neighbors and the
residual quantities inside the cell. Air-water interface
is managed only for external flows from air to water.
An external flow from an air cell (altitude higher than
water level) to water cell (altitude lower than water
level) implies always a loss of matter (water inside de-
bris and components lighter than water) proportional
to debris mass, specified by a parameter ml.
SCIDDICA-SS2 was tested on the 1997 subaeri-
al-subaqueous debris flow at Lake Albano (Rome, Ita-
ly) and on a recent completely submarine debris-flow,
in the nearshore of Bagnara Calabra (Italy).
- A further improvement of SCIDDICA (SS-
2blocks) was devised in order to simulate landslides
characterized by the presence of large blocks inside
the main landslide debris. Blocks are idealised as cyl-
inders; they move along the line of maximum slope
when there is no fluid matter in all the cells or in part
of the cells occupied by it; the shift of blocks is com-
puted by classic motion equations (see at a
volio
et
alii, 2009 for further details).
- Latest developments of the model regard the
management of the interaction between the debris
flow and different types of man-made structures (i.e.
houses, cables, pipelines, and defenses structures)
both in subaerial and in submerged environment
SUSCEPTIBILITY ANALYSIS
Susceptibility analysis of debris flow impact on
human settlements and structures is the basic require-
ment for risk assessment and reduction strategies.
Simple susceptibility analyses have been performed
by SCIDDICA S3−hex in recent years (Iovine et al.
2002; 2003a; 2003b; 2005; 2007).
Basic information required for such analysis are:
(i) detailed topographic data; (ii) map of the erod-
ible soil cover overlying the bedrock; (iii) location
of landslide sources; (iv) areal and volumetric size of
landslide sources; (v) back analysis simulation of real
landslides occurred in the same area necessary for the
calibration of model parameters; (vi) validation of the
model by the comparison of real and simulated events.
In Fig. 2 an example of a simple susceptibility
zonation is shown, where different grey-tones are as-
However, in spite of the relevant improvements
described above, the SCIDDICA models were still af-
fected by a relevant limitation: the flow velocity could
be only deduced “a posteriori” by averaging in space
(i.e. considering clusters of cells and computing the
resulting velocity for all cluster flows) or in time (e.g.
considering the average velocity of the advancing flow
front in a sequence of CA steps) since the flow moves
from a cell to another one in a CA step (which corre-
sponds to a constant time). Hence, in the CA context of
discrete space/time, “velocity” is constant.
- This limitation has been resolved in the version
(SS2) (a
volio
et alii, 2008, 2009; m
azzanti
et alii,
2009) by applying the same approach previously de-
veloped for the lava flow model SCIARA (a
volio
et
alii, 2006b). This approach is based on the introduc-
tion of barycentre co-ordinates for debris flows, thus
obtaining “explicit” velocity. Furthermore, this ver-
sion was significantly improved, in order to simulate
combined subaerial-subaqueous debris flows, by in-
troducing different parameters and different transition
functions for the modelling of both the subaerial and
the submerged path at the same time. Effects like loss
of energy and loss of material at the air-to-water tran-
sition were added instead.
In this version X is HEX, while S=S
A
×S
TH
×S
X
×
S
Y
×S
KH
×SE
6
×S
XE
6
×S
YE
6
×S
XI
6
×S
YI
6
×S
OE
×S
OI
, AMD and
erosion are the same as S3-hex. Regarding co-ordi-
nates X and Y; its
“form” (as the form of outflows) is ideally a “cyl-
inder” centred at the barycentre and tangent to the
next edge of the hexagonal cell. Outflows move on
such ideal path, whose shift are computing according
to the simple kinetic formula which depends whether
the flow is subaerial or subaqueous. In fact, the shift
formula for subaqueous debris considers also a water
resistance parameter, using modified Stokes equations
(a
volio
et alii, 2008) with a form factor parameter
which is proportional to mass and a parameter g’<g,
accounting for buoyancy. The motion involves three
possibilities: (1) only internal flow (the entire debris
remain inside the cell); (2) only external flow (all the
debris moves to the adjacent cells); (3) the flow is di-
vided between the central and the adjacent cell (a part
of the debris remains inside the cell and a part of it
moves towards adjacent cells) . The kinetic head vari-
ation is computed according to the new position of in-
ternal and external flows; furthermore, a contribution
background image
M.V. AVOLIO, F. BOZZANO, D. D’AMBROSIO, S. DI GREGORIO, V. LUPIANO, P. MAZZANTI, R. RONGO & w. SPATARO
392
5th International Conference on Debris-Flow Hazards Mitigation: Mechanics, Prediction and Assessment Padua, Italy - 14-17 June 2011
signed to zones affected by one/more simulated phe-
nomena. Specifically, areas affected by only one of the
simulated landslides are considered less susceptible
(light gray) than zones affected by more landslides
(dark gray). More sophisticated and powerful ap-
proaches for susceptibility analysis have been devel-
oped by our research group and tested on lava flows
(C
RisCi
et alii, 2010). A similar approach could be ap-
plied (with some modifications) to the susceptibility
analysis of debris flows.
DEBRIS FLOW SIMULATION BY SCID-
DICA
Simulations by SCIDDICA of four debris flow
events occurred in Italy in the recent years are present-
ed below. Specifically, simulations of two completely
subaerial debris flows (Chiappe di Sarno-Curti, 1998
and San Martino Valle Caudina, 1999), one submarine
debris flow (Bagnara Calabra, 2008) and one mixed
subaerial-submerged debris flow (Lake Albano, 1997)
are shown. A short description of simulation results
are also presented together with a quantitative evalu-
ation of simulations according to a fitness function
based on the areal comparison of real and simulated
event (D’Ambrosio & Spataro, 2007). Specifically,
the considered evaluation function is
where R is the set of cells involved in the real
event and S the set of cells involved in the simulated
event. A value between 0 (total failure) and 1 (perfect
simulation) is obtained, with values greater than 0.7
considered as satisfactory for landslide simulations
CHIAPPE DI SARNO–CURTI
On May 5–6 1998, about 150 debris flows were
triggered by exceptional rainfalls in Campania (South-
ern Italy), mostly on the slopes of Pizzo d’Alvano
massif. Hundreds of small debris slides originated in
the volcanoclastic mantle overlying carbonate bed-
rock and propagated downslope as an extremely rap-
id, highly erosive debris flow, dramatically increasing
their volume (z
anCHetta
et alii, 2004). One of these
events took place in the Chiappe di Sarno slope. The
size of the detachment area was approximately 100
m
2
, and the volume of the involved material was less
than 100 m
3
. the sliding mass rapidly transformed
into a fast-flowing mixture of mud, debris, and water,
running down the slope along a smooth pre-existing
channel, for about 375 m, and eroding the available
detrital cover. The debris flowed for a distance of
about 325 m, triggering some minor debris slides on
both flanks of the channel. After that, influenced by
the pre-existing morphology, it made a left turn, en-
larged, and subdivided into two distinct flows.
Simulation of this event was performed by SCID-
DICA S3-hex. A graphic comparison between the map
of the best simulation, and the real case is shown in
Fig. 2 - Debris-flow susceptibility map of the southern
slope of the Pizzo d’Alvano massif: ten soil-slip
sources have been considered (case a, b, c, d cor-
respond to the may 1998 debris flows) along the
southern slope of Pizzo d’Alvano. Area “poten-
tially” affected by one (1), two (2), three-four (3),
and five-six (4) simulated landslides; 5) wall; 6)
secondary soil slips; 7) border of the area consid-
ered in simulations
Fig. 3 - The 1998 Curti DF: comparison between the real
case and the best simulation. (1) area affected by
real landslide, (2) simulated landslide, (3) both
cases; and (4) border of the area considered for
comparison
background image
DEBRIS FLOWS SIMULATION BY CELLULAR AUTOMATA: A SHORT REVIEW OF THE SCIDDICA MODELS
Italian Journal of Engineering Geology and Environment - Book www.ijege.uniroma1.it © 2011 Casa Editrice Università La Sapienza
393
The SCIDDICA SS2 model was validated
against the 1997 Albano lake (Rome, Italy) event
(Figure 5) which is a rare case of combined subaer-
ial-subaqueous debris-flow. This landslide occurred
in the eastern slope of the Albano lake on the 7th of
November 1997 after an intense rainfall event (128
mm in 24 hours), and it began as a soil slide, mobi-
lizing about 300 m
3
of eluvial material. The mobi-
lized mass was channelled within a steeply dipping
impluvium (about 40°) and thus evolved as a debris
flow which entrained a large amount of debris mate-
rial along the bottom of the channel and reached an
estimated volume of some thousands of m3 at the
coastline. A few amount of material was deposited at
the coastline while a greater quantity entered in wa-
ter generating a little tsunami wave. Simulation re-
sults are quite satisfactory since the achieved fitness
function ev value was 0.85 (a
volio
et alii, 2008).
Furthermore, results show a good agreement also in
terms of erosion and deposits on both subaerial and
subaqueous parts and reasonable values of the land-
slide velocity (up to 15 m/s).
BAGNARA CALABRA
A completely submarine landslide was detected
in the nearshore of Bagnara Calabra (Italy) by com-
paring detailed bathymetries coming from two sonar
multibeam surveys carried out in November 2007
and in September 2008. Landslide detachment area
was located between 10 m and 20 m b.s.l., about 100
m far from the coastline. Initial landslide volume
Fig. 3. As one can see, the overall area affected by
the DF and depth of regolith erosion along the path
are in good accordance with surveyed evidences; the
branching of the flow at the base of the slope is fairly
well simulated. The obtained ev fitness value was
0.76-0.78 (d’a
mbRosio
et alii, 2003a-b)
SAN MARTINO VALLE CAUDINA
On December 15th–16th, 1999, heavy rainfalls
severely stroke the Campania region (Italy), trig-
gering numerous debris flows on the slopes of the
Monti di Avella massif (v
ittoRi
et alii, 2000). Sev-
eral soil slips originated on the slopes of the Vallone
Castello (Cervinara) and Mt. Pizzone (San Martino
V.C.), within the weathered volcaniclastic mantle.
These landslides turned into fast flowing mixtures
of matrix and large blocks, downslope eroding the
soil cover and increasing their original volume. At
the base of the slopes, debris flows impacted on the
cited urban areas, causing one casualty and severe
destruction. Simulation of this event was performed
by SCIDDICA S4. A graphic comparison among real
landslides and “best” simulations is shown in Fig.4
(i
ovine
et alii, 2003b). The achieved ev fitness val-
ues were 0.77, 0.86 and 0.64 for the case A, B and
C respectively (Fig.4). These results demonstrate the
efficacy of the SCIDDICA models to simulate also
cases of open channel debris flows (mudflows).
ALBANO LAkE
Fig. 4 - The 1999 San Martino Valle Caudina DF: com-
parison between the real case and the best simula-
tion. (1) area affected by real landslide, (2) simu-
lated landslide, (3) both cases; and (4) border of
the area considered for comparison.
Fig 5 - The 1997 Albano lake subaerial-subaqueous
debris flow as simulated by the SCIDDICA SS2
model. key: (1) real event, (2) simulated event, (3)
intersection between real and simulated event, (4)
water level
background image
M.V. AVOLIO, F. BOZZANO, D. D’AMBROSIO, S. DI GREGORIO, V. LUPIANO, P. MAZZANTI, R. RONGO & w. SPATARO
394
5th International Conference on Debris-Flow Hazards Mitigation: Mechanics, Prediction and Assessment Padua, Italy - 14-17 June 2011
TION WITH OTHER MODELS
Simulations carried out by SCIDDICA have been
recently compared with simulations by other debris
flow models. In particular, the 1997 Lake Albano
debris flow has been simulated by SCIDDICA SS2
as well as by DAN-w (H
unGR
, 1995) and DAN3D
(m
C
d
ouGall
& H
unGR
, 2004). Results achieved by
DAN3D and SCIDDICA were surprisingly similar in
terms of areal debris distribution, velocity and time
of propagation (m
azzanti
et alii, 2009) while results
obtained by DAN-w show a significant discrepancy.
This is once more a confirmation of the limitations af-
fecting 2D models, particularly for the simulation of
channeled events like debris flows.
CONCLUSION REMARKS AND OUTLO-
OKS
A state of the art description of Cellular Automata
models for the simulation of debris flows has been
presented together with some examples of applica-
tion to real events occurred in Italy in recent years.
Thanks to the significant improvements made in the
latest years, and briefly summarized in this paper in a
chronological way (Tab.2), Cellular Automata models
can now be considered as a powerful and reliable tool
for the simulation of debris flows and hyper-concen-
trated flows. In particular, the introduction of explicit
velocity (in the SCIDDICA version SS2) has brought
these models to the same level of the most accepted
codes as DAN3D
Main criticisms moved to such models argue that
they are not fully physically based. However, with the
introduction of explicit velocity, the management of
momentum (even if yet in a rough way) and the intro-
duction of turbulence forces, which can be found in
the latest versions of SCIDDICA, the main require-
ments of the “equivalent fluid” approach (H
unGR
,
1995) are satisfied.
was also estimated at about 16.000 m
3
with a maxi-
mum thickness of 9 m. Erosion up to 4 m has been
recorded along the pathway between 20 m and 60
m b.s.l.. Final deposit is partly distributed between
60 m and 90 m b.s.l. and partly below 100 m with a
maximum thickness of 5 m.
A fitness value of 0.85 was achieved for this
landslide (a
volio
et alii, 2009). Furthermore, de-
posit and erosion locations in the simulation agree
very satisfactorily with the real event; moreover,
deposit thickness and erosion depth values do not
differ substantially. The detachment area was com-
pletely emptied after about one minute and the flow
propagated until its final position in few minutes.
Landslide velocity was up to 6 m/s in the upper part
of the slope, immediately after the mass release, and
dropping below 4 m/s in the following stages. Such
values of velocity are considered reasonable for the
type and volume of landslide and the slope gradi-
ent (up to 12°). Fig. 6 shows an example of possible
output layers than can been obtained by SCIDDICA
SS2 and specifically: final deposit thickness, erosion
thickness and maximum velocity.
COMPARISON OF SCIDDICA SIMULA-
Fig 6 - simulation of Bagnara Calabra subaqueous
landslide: a) deposit thickness, b) erosion, c)
maximum occurred debris thickness; d) maxi-
mum velocity; contour lines are referred to: 1
perimeter of the simulated event; 2 perimeter
of real event; 3 probable real event perimeter
in missing data area
Tab.2 - features of the SCIDDICA versions
background image
DEBRIS FLOWS SIMULATION BY CELLULAR AUTOMATA: A SHORT REVIEW OF THE SCIDDICA MODELS
Italian Journal of Engineering Geology and Environment - Book www.ijege.uniroma1.it © 2011 Casa Editrice Università La Sapienza
395
and reliable approaches are under development (i.e.
C
RisCi
et alii, 2010; C
ePeda
et alii, 2010).
On the other hand, it must be pointed out that
some improvements of the SCIDDICA model are still
possible. First of all, a new solution for the manage-
ment of the momentum is needed, as the importance
of non isotropic internal earth pressure (s
avaGe
&
H
utteR
, 1989) for landslide simulation has been dem-
onstrated (P
iRulli
et alii, 2007). This is one of the
main tasks for the next future, even if the approxima-
tion of hydrostatic internal earth pressures does not
lead to significant mistakes in the case of debris flows
with a high percentage of water (C
ePeda
, 2007; m
az
-
zanti
et alii, 2009).
A most advanced computation of erosion process
is also under development with specific reference to
the erosion parameters.
A further improvement required to the SCIDDI-
CA model is the introduction of explicitly interactions
between solid and fluid-phases on the debris as the rel-
evance of inner fluid pressures in the propagation of a
debris flow has been demonstrated (i
veRson
, 1997).
A new objective is an efficient management of the
interaction of debris flow with man made structures.
First preliminary results have been achieved for test-
ing the resistance of submarine cables, even if signifi-
cant improvements are still necessary.
However, in spite of the aforementioned approxi-
mations, the last versions of the SCIDDICA are quite
stable, well-performing and suitable for the simulation
of different types of debris flows in different environ-
ments. In other words, SCIDDICA can be considered
a valid and efficient tool for the susceptibility analysis
of debris flow run-out at a local or global scale.
Rather, a general critic could be moved to the ef-
ficacy of the equivalent fluid approach which assumes
a complex mixtures of water and heterogeneous debris
(which often changes its features during the propaga-
tion) to behave as an equivalent and homogeneous
fluid (usually integrated in depth) controlled by only 2
or 3 parameters. It is evident that this is a very strong
approximation; however, at this time, models based
on the equivalent fluid approach are probably the only
ones able to give reasonable results (in a reasonable
time of computation).
As a matter of fact, the most advanced models
based on the equivalent fluid approach like SCID-
DICA SS (i.e. m
C
d
ouGall
& H
unGR
, 2004; Pirulli et
al., 2010) are able to simulate the main characteristic
of debris flows, such as:
(i)
propagation of debris in channeled slopes
over a 3D topography;
(ii) erosion and entrainment of material along
the path;
(iii) interaction with structures;
(iv) immersion in water (in the case of costal DF).
Nevertheless, it must be pointed out that users
and researchers which intend to use such an approach
must be completely conscious of these limitations.
Furthermore, these models have been tested by
simulating several real events, and are now able to
give good results also in terms of forecasting analy-
ses. In particular, forecasting instruments for DF im-
pact are a fundamental tool for the society in order to
produce hazards maps which must become the basic
data for the landscape management. The first step in
this direction has been already carried out with the
SCIDDICA models by using a simple approach for
susceptibility analysis. However, more sophisticated
REFERENCES
a
nCey
C. (2007) - Plasticity and geophysical flows: A review.- J. Non-Newtonian Fluid Mech., 142: 4–35
a
volio
m.v., d
i
G
ReGoRio
s., m
antovani
f., P
asuto
a., R
onGo
R., s
ilvano
s. & s
PataRo
w. (2000) - Simulation of the 1992
Tessina landslide by a cellular automata model and future hazard scenarios. Intern. Journal of Applied Earth Observation
and Geoinformation, 2 (1): 41-50.
a
volio
m.v. (2004) - Esplicitazione della velocità per la modellizzazione e simulazione di flussi di superficie macroscopici con
automi cellulari ed applicazioni alle colate di lava di tipo etneo. Ph. D. Thesis (in Italian). Dept. of Mathematics, University
of Calabria.
a
volio
m.v., C
RisCi
G.m., d
i
G
ReGoRio
s., R
onGo
R., s
PataRo
w. & d’a
mbRosio
d. (2006a) - Pyroclastic flows modelling using
cellular automata. Computers and Geosciences, 32: 897–911.
a
volio
m.v., C
RisCi
G.m., d
i
G
ReGoRio
s., R
onGo
R.,. s
PataRo
w. & t
Runfio
G.a. (2006b) - SCIARA 2: an improved Cellular
Automata model for Lava Flows and Applications to the 2002 Etnean crisis. Computers & Geosciences, 32: 897–911.
background image
M.V. AVOLIO, F. BOZZANO, D. D’AMBROSIO, S. DI GREGORIO, V. LUPIANO, P. MAZZANTI, R. RONGO & w. SPATARO
396
5th International Conference on Debris-Flow Hazards Mitigation: Mechanics, Prediction and Assessment Padua, Italy - 14-17 June 2011
Avolio M.V., Lupiano V., Mazzanti P. & Di GregorioS. (2008) - Modelling combined subaerial-subaqueous flow-like landslides
by Cellular Automata. Umeo H., Chopard B. & Bandini S. (eds): ACRI 2008, LNCS 5191: 329-336.
a
volio
m.v., l
uPiano
v., m
azzanti
P. & d
i
G
ReGoRio
s. (2009) - An advanced Cellular Model for Flow-type Landslide with
Simulations of Subaerial and Subaqueous cases. – In: V.Wohlgemuth, B. Page, K. Voigt (eds) “EnviroInfo 2009, Environ-
mental Informatics and industrial Environmental Protection: Concepts, Methods and Tools”, 1: 131-140, Germany, ISBN
978-3-8322-8397-1.
a
volio
m.v., d
i
G
ReGoRio
s., l
uPiano
v., m
azzanti
P. & s
PataRo
w. (2010a) - Application context of the SCIDDICA model
family for simulations of flow-like landslides. WORLDCOMP2010 Proceedings: 40-46, CSREA Press, ISBN 1-60132-137.
a
volio
m.v., e
RReRa
a., l
uPiano
v., m
azzanti
P. & d
i
G
ReGoRio
s. (2010b) - Development and calibration of a preliminary
Cellular Automata Model for Snow Avalanches. S. Bandini et al. (eds.): ACRI 2010, LNCS 6350: 83–94.
b
aRPi
f., b
oRRi
-b
Runetto
m. & d
elli
v
eneRi
l. (2007) - Cellular-Automata Model for Dense-Snow Avalanches. Journal of Cold
Regions Engineering, 21(4): 121-140.
C
ePeda
J. (2007) - The 2005 Tate’s Cairn debris flow: back-analysis, forward predictions and a sensitivity analysis. In: Ho K., Li
V. (eds) 2007 International Forum on Landslide Disaster Management, 813–833.
C
ePeda
J., C
Hávez
J.a. & m
aRtínez
C.C. (2010) - Procedure for the selection of runout model parameters from landslide back-
analyses: application to the Metropolitan Area of San Salvador, El Salvador. Landslides, 7: 105–116.
C
leRiCi
a. & P
eReGo
s. (2000) - Simulation of the Parma River blockage by the Corniglio landslide (Northern Italy). Geomor-
phology, 33: 1–23.
C
RisCi
G.m., d
i
G
ReGoRio
s. & R
anieRi
G.a. (1982) - A cellular space model of basaltic lava flow. Proceedings AMSE Confer-
ence Modelling & Simulation, (Paris, France, Jul.1-3), 65–67.
C
RisCi
G.m., a
volio
m.v., b
eHnCke
b., d'a
mbRosio
d., d
i
G
ReGoRio
s., l
uPiano
v., n
eRi
m., R
onGo
R. & s
PataRo
w. (2010)
- Predicting the impact of lava flows at Mount Etna, Italy. J. Geophys. Res., 115, B04203, doi:10.1029/2009JB006431.
d’a
mbRosio
d., d
i
G
ReGoRio
s., i
ovine
G., l
uPiano
v., m
eRenda
l., R
onGo
R. & s
PataRo
w. (2002) - Simulating the Curti-
Sarno Debris Flow through Cellular Automata: the model SCIDDICA (release S2). Physics and Chemistry of the Earth,
27:
1577-1585.
d'a
mbRosio
d., d
i
G
ReGoRio
s., i
ovine
G., l
uPiano
v., R
onGo
R. & s
PataRo
w. (2003a) - First simulations of the Sarno debris
flows through cellular automata modelling. Geomorphology, 54 (1-2): 91-117.
d'a
mbRosio
d., d
i
G
ReGoRio
s. & i
ovine
G. (2003b) - Simulating debris flows through a hexagonal cellular automata model:
SCIDDICA S3-hex. Natural Hazards and Earth System Sciences, 3: 545-559.
d'a
mbRosio
d., s
PataRo
w. & i
ovine
G. (2006) – Parallel genetic algorithms for optimising cellular automata models of natural
complex phenomena: an application to debris-flows. Computers & Geosciences, 32: 861–875.
d’a
mbRosio
d. & s
PataRo
w. (2007) - Parallel evolutionary modelling of geological processes. Parallel Computing, 33(3):
186-212.
d
enlinGeR
R.P. & i
veRson
R.m. (2001) - Flow of variably fluidized granular masses across three-dimensional terrain: 2. Nu-
merical predictions and experimental tests. J. Geophys. Res., 106(B1): 553–566.
d
i
G
ReGoRio
s., R
onGo
R., s
iCiliano
C., s
oRRiso
-v
alvo
m. & s
PataRo
w. (1999) - Mount Ontake landslide simulation by the
cellular automata model SCIDDICA-3. Physics and Chemistry of the Earth (A), 24(2): 97-100.
d
i
G
ReGoRio
s. & s
eRRa
R. (1999) - An empirical method for modelling and simulating some complex macroscopic phenomena
by cellular automata. Future Generation Computer Systems, 16(2/3): 259-271.
e
GasHiRa
s. (2007) - Review of Research Related to Sediment Disaster Mitigation. Journal of Disaster Research, 2(1): 11-18.
H
unGR
o. (1995) - A model for the runout analysis of rapid flow slides, debris flows, and avalanches. Can. Geotech. J., 32:
610–623.
H
unGR
o., e
vans
s.G., b
ovis
m.J. & H
utCHinson
, J.n. (2001) - A review of the classification of landslides in the flow type. En-
vironmental and Engineering Geoscience, 7(3): 221–228.
H
unGR
o., m
C
d
ouGall
s. & b
ovis
b. (2005) - Entrainment of material by debris flows. In Debris Flow Hazards and Related
Phenomena, Jakob M., Hungr O. (eds). Springer – Praxis: Heidelberg, Germany, 135–158.
i
ovine
G., d
i
G
ReGoRio
s., d’a
mbRosio
d. & l
uPiano
v. (2002) - Debris-flows and Cellular Automata: an example of simula-
tion from the 1998 disaster of Sarno (Italy). In: Proceedings of the International Symposium “Geomorphology: from expert
background image
DEBRIS FLOWS SIMULATION BY CELLULAR AUTOMATA: A SHORT REVIEW OF THE SCIDDICA MODELS
Italian Journal of Engineering Geology and Environment - Book www.ijege.uniroma1.it © 2011 Casa Editrice Università La Sapienza
397
opinion to modelling”, Delahaye D., Levoy F. & Maquaire O. (eds), Strasbourg, 26–27 April, 55–64.
I
ovine
G., d
i
G
ReGoRio
s. & l
uPiano
v. (2003a) - Assessing debris-flow susceptibility through cellular automata modelling: an
example from the May 1998 disaster at Pizzo D'Alvano (Campania, Southern Italy). In: Debris-flow hazards mitigation:
mechanics, prediction and assessment. Rickenmann D. & Chen C.L. (eds): 623-634.
i
ovine
G., d
i
G
ReGoRio
s. & l
uPiano
v. (2003b) - Debris Flows susceptibility assessment through Cellular Automata modeling:
an example from the 15-16 December 1999 disaster at Cervinara and San Martino Valle Caudina (Campania, southern
Italy).
NHESS 3: 457-468.
i
ovine
G., d’a
mbRosio
d. & d
i
G
ReGoRio
s. (2005) - Applying genetic algorithms for calibrating a hexagonal cellular automata
model for the simulation of debris flows characterised by strong inertial effects. Geomorphology, 66: 287-303
i
ovine
G., d
i
G
ReGoRio
s., s
HeRidan
m. & m
iyamoto
H. (2007) - Modelling, computer-assisted simulations, and mapping of
dangerous phenomena for hazard assessment. Environmental Modelling & Software 22(10): 1389-1391.
i
veRson
R.m. (1997) - The Physics of Debris Flow. Reviews of Geophysics, 35(3): 245-296.
J
akob
m.t.(2005) - A size classification for debris flows. Engineering Geology, 79: 151–161.
m
aJoR
J.J. & P
ieRson
t.C. (1992) - Debris flow rheology: experimental analysis of fine-grained slurries. Wat. Resour. Res.,
28(3): 841–857.
m
azzanti
P., b
ozzano
f., a
volio
m.v., l
uPiano
v. & d
i
G
ReGoRio
s. (2009) - 3D numerical modelling of submerged and coastal
landslides propagation. In: Mosher D.C., Shipp C., Moscardelli L., Chaytor J., Baxter C., Lee H. & Urgeles R. (eds):
Submarine Mass Movements and Their Consequences IV. Advances in Natural and Technological Hazards Research, 28.,
Springer Netherlands, 127-138.
m
C
d
ouGall
s. & H
unGR
o. (2004) - A model for the analysis of rapid landslide runout motion across three-dimensional terrain.
Canadian Geotechnical Journal, 41: 1084- 1097.
P
iRulli
m., b
Risteau
m.o., m
anGeney
a. & s
Cavia
C. (2007) - The effect of the earth pressure coefficients on the runout of
granular material. Environ Model Software, 22: 1437–1454.
P
iRulli
m. & m
aRCo
f. (2010) - Description and numerical modelling of the October 2000 Nora debris flow, Northwestern Ital-
ian Alps. Can. Geotech. J., 47: 135–146.
s
avaGe
s.b. & H
utteR
k. (1989) - The motion of a finite mass of granular material down a rough incline. J. Fluid Mech. 199:
177–215.
s
eGRe
e.& d
eanGeli
C. (1995) - Cellular automaton for realistic modelling of landslides. Nonlinear Processes in Geophysics
2(1): 1–15.
z
anCHetta
G., s
ulPizio
R., P
aResCHi
m.t., l
eoni
f.m. & s
antaCRoCe
R. (2004) - Characteristics of May 5-6, 1998 volcaniclastic
debris fows in the Sarno area (Campania, southern Italy): relationships to structural damage and hazard zonation. J. Volc.
Geoth. Res. 133: 377-393.
Statistics