Analogue and numerical modeling of the Stromboli hot avalanches

Abstract

Hot avalanches at Stromboli volcano were investigated by means analogue and numerical modeling. Analogue experiments were performed with the aim of understanding the effects of different trigger mechanisms of slope instability while the runout of Stromboli hot rock avalanches was modeled using two numerical codes DAN-W (2D version) and DAN-3D. The accumulation experiments demon- strate that the accretion of a portion of the slope alters the flank stability and triggers small landslides. Numerical models were able to reproduce the extension and the order of magnitude of the thickness of the hot avalanches reported in the literature. The best results of DAN-3D and DAN-W models on the 1930 hot avalanche were obtained using a Voellmy model with a frictional coefficient f = 0.19 and a turbulence parameter ξ = 1000 m/s.. The obtained results allowed to produce a hazard evaluation for the explosive-related, mass-wasting phenomena in the inhabited areas of Stromboli Island.

Full text

1493
Landslides and Engineered Slopes. Experience, Theory and Practice – Aversa et al. (Eds)
© 2016 Associazione Geotecnica Italiana, Rome, Italy, ISBN 978-1-138-02988-0
Analogue and numerical modeling of the Stromboli hot avalanches
S. Morelli, T. Salvatici, T. Nolesini, F. Di Traglia, C. Del Ventisette & N. Casagli
Dipartimento di Scienze della Terra, Università di Firenze, Firenze, Italy
A. Di Roberto, M. Bisson, M. Pompilio & A. Bertagnini
Istituto Nazionale di Geofisica e Vulcanologia, Sezione di Pisa, Pisa, Italy
ABSTRACT: Hot avalanches at Stromboli volcano were investigated by means analogue and numerical
modeling. Analogue experiments were performed with the aim of understanding the effects of different
trigger mechanisms of slope instability while the runout of Stromboli hot rock avalanches was modeled
using two numerical codes DAN-W (2D version) and DAN-3D. The accumulation experiments demon-
strate that the accretion of a portion of the slope alters the flank stability and triggers small landslides.
Numerical models were able to reproduce the extension and the order of magnitude of the thickness
of the hot avalanches reported in the literature. The best results of DAN-3D and DAN-W models on
the 1930 hot avalanche were obtained using a Voellmy model with a frictional coefficient f = 0.19 and a
turbulence parameter ξ = 1000 m/s.. The obtained results allowed to produce a hazard evaluation for the
explosive-related, mass-wasting phenomena in the inhabited areas of Stromboli Island.
duced the effect of spatter or lava loading on the
volcano flanks and were able to induce landslides.
With the purpose of testing the suitability of
landslide numerical models in simulating and assess-
ing the hazard related to hot avalanches, the results
of the back analysis of three events of avalanches
occurred at Stromboli volcano are also presented.
They were performed using 2D and 3D numerical
codes called DAN-W and DAN-3D respectively
(Hungr 1995, McDougall and Hungr, 2004; Hungr
and McDougall, 2009). In this work, the bidimen-
sional simulation was joined to a more complex
one (3D) in order to test if also the simplest math-
ematical approach maintains high levels of reliabil-
ity in case of strictly channeled events like those
investigated. As case studies, three flows occurred
on 1906, 1930 and 1944 were selected. Back analy-
sis was undertaken through the use of DAN-3D
and DAN-W codes, considering the 1930 event as
test case. In fact, for this event, detailed descrip-
tions are available from many authors. Rittmann
(1931) and Abbruzzese (1935) gathered many data
shortly after the eruption and in particular they
deduced the total runout distance, velocity, thick-
ness and distribution of deposits. More recently,
Di Roberto et al. (2014) provide additional data on
flow dynamics and distribution of deposits.
The simulations were able to reproduce the
extension and the order of magnitude of the
deposit thickness of events reported in the lit-
erature (Rittmann, 1931; Abbruzzese, 1935; Di
Roberto et al., 2014). The outcomes were also used
1 INTRODUCTION
Hot avalanches deposits, originated from the slid-
ing of the crater rim or the gravitational instabilities
of material accumulated during explosive erup-
tions, are widely identified on the flanks of several
volcanoes worldwide (Davies et al., 1978; Nairn
and Self, 1978; Hazlett et al., 1991; Arrighi et al.,
2001; Cole et al., 2005; Yamamoto et al. 2005; Beh-
ncke et al., 2008; Di Roberto et al. 2014; Di Traglia
et al., 2014). Such kind of events usually occurs
on edifices fed by mafic to intermediate magmas
and have small volumes (104–107 m3) but emplace
at very high temperatures and can travel far from
the source at very high speed. These features make
them potentially dangerous for communities that
live and concentrate their socio-economic activities
close to the volcanoes and for the high number of
tourists attending them each year for recreational
activities.
Analogue experiments were conducted with the
aim to understand the triggering mechanism and
the evolution of landslides along the Sciara del
Fuoco, a horse-sharpened depression on the north-
west flank of the Stromboli volcano. To simulate
the brittle behavior of the volcanic material along
the Sciara del Fuoco slope we used analogue mate-
rials reported in Nolesini et al. (2013): i) quartz
Fontainebleau’s sand; ii) uniform sand; iii) Sciara
del Fuoco volcaniclastic material; iv) silty-sand.
The analogue models reveal that the accumulation
of material in the summit part of the slope repro-

1494
to assess whether and how the inhabited areas of
Stromboli (Stromboli and Ginostra villages) can
be struck if one of these events would repeat in
the future with similar dynamics of the historical
episodes.
2 STUDY AREA: THE STROMBOLI
VOLCANO
Stromboli is a volcanic island of the Aeolian archi-
pelago (Tyrrhenian Sea southern Italy) (Fig. 1).
The island is the subaerial part of volcanic edifice,
characterized by a rather regular conical shape,
rising up to 924 m above sea level (a.s.l.) from a
base that lies between 2300 m and 1300 m of water
depth. The volcanic activity of Stromboli has been
continuous since the 8th century AD and mainly
consists in low energy intermittent explosions
(Strombolian activity) occasionally interrupted by
effusive events and by violent explosions regularly
called paroxysms (Barberi et al., 1993; Rosi et al.,
2013). The active vents are located in the crater ter-
race, at about 750 m in the upper part of the Sciara
del Fuoco (SdF), a horseshoe-shaped depression
that occupies the NW flank of the volcano (Fig. 1).
At Stromboli volcano, the formation of mass flows
of hot pyroclasts have been observed and reported
several times. These occur directly as a result of the
explosive and effusive volcanic activity and usually
spread within the SdF thus not representing a seri-
ous menace for the population of Stromboli (Bar-
beri et al., 1993; Rosi et al., 2013). However, at least
in 1930, 1944 and possibly in 1906, the hot ava-
lanches occurred outside the SdF and in the first
two cases they reached the coastline. In particular,
the 1930 event reached the village of Stromboli on
the NE part of the island causing extensive dam-
ages and four fatalities (Rittmann, 1931).
Few data are reported on the hot avalanches
occurred on 15 July 1906 and 20 August 1944 and
the main information can be obtained by the writ-
ing of Riccò (1907) and Ponte (1948) respectively.
On the other hand, the best-described event is
undoubtedly that occurred during the 1930 par-
oxysm thanks to some distinguished post-event
studies supported by more recent investigations
on deposits. This hot avalanche was triggered by
the sliding of an approximately 1 m-thick deposit,
corresponding to an estimated volume of at least
75,000 m3, and consisting of meter-sized spatter,
decimeter-sized bombs, lapilli, and ash (Rittmann,
1931; Abbruzzese, 1935). This deposit accumu-
lated over an area of more than 60,000 m2 on the
steep cliff side of Chiappe Lisce about one hour
before the landslide event (Rittmann, 1931) dur-
ing an extremely violent paroxysm (Bertagnini
et al., 2011). The description of the 1930 deposits,
including grain-size and nature of main compo-
nents is reported by Di Roberto et al. (2014).
3 MATERIALS AND METHODS
3.1 Analogue modeling
Analogue experiments were performed with the
aim of understanding the effects of the triggering
mechanism of slope instability and constraining
the geometry of the induced deformations. Experi-
ments were conducted to consider the accumula-
tion of material on the slope. For this study 20
different models were completed using different
materials (Nolesini et al., 2013). Much of the Sci-
ara del Fuoco is inaccessible due to continuous
ejection of products from the frequent explosions
and frequent falling rocks. The samples taken in
the marginal position of Sciara del Fuoco are con-
sidered representative of the deep position.
Data on grain size and the structure of volcani-
clastic materials of Stromboli volcano are reported
in Apuani et al. (2005), Rotonda et al. (2009)
and Nolesini et al. (2013). To simulate the brittle
behavior of Sciara del Fuoco material, four differ-
ent analogue materials reported in Nolesini et al.
(2013) were used: i) quartz Fontainebleau’s sand;
ii) uniform sand; iii) Sciara del Fuoco volcaniclas-
tic material; iv) silty-sand.
Nolesini et al. (2013) observed that the Strom-
boli Sciara del Fuoco material is quite different
Figure 1. Study area: Stromboli volcano with the main
toponyms (by QuickBird Satellite Sensor). The three red
rectangular areas cover the volcano sector involved in the
event of 1930 (zone A), 1944 (zone B) and 1906 (zone
C). These zones are used for numerical modeling with
DAN-3D. In yellow the three events source areas.

1495
from the other types of analyzed granular mate-
rials, due to the large heterogeneity and, specifi-
cally, to the presence of fine material, resulting
in a high internal friction angle of the Stromboli
material. The brittle behavior of rocks can be
expressed by the Mohr-Coulomb criterion of fail-
ure (τ=µσ(1−λ)+c; where τ and σ are the shear
and normal stress on the sliding surface, µ is the
internal friction coefficient, λ is the Hubbert—
Rubey coefficient of fluid pressure and c is the
cohesion). Since cohesion has the dimensions of
a stress, it should share a similar scaling ratio. In
the same way the internal friction coefficient must
have similar values both in models and in nature.
The models were suitably scaled such that 1 cm in
the model represents 100 m in nature, involving a
geometrical length ratio
l*=lmod=lnat=10−4.
Considering the length ratio, the gravity ratio
(g* = 1; the models where performed in the natu-
ral gravity field) and the density ratio ρ* ≈ 0.5, the
stress σ* acting on the model is 5 × 10−5 Pa. The
Plexiglas tank is 25 × 30 × 50 cm, and these dimen-
sions limit the boundary effects due to confine-
ment. The modeling material was sieved on a slope
with an inclination that varied between 45° and
25°. During the deposition of the material, a color-
ful reference level was used as a marker, and a final
grid 5 × 5 cm was put on the last sand layer to better
observe the deformations. The tank was then situ-
ated in a horizontal position, and the experiment
was started. All of the models were developed in
the Earth Science Department Laboratory of the
University of Firenze.
3.2 Numerical modeling
Considering the sliding processes associated to
this kind of event, the motion of Stromboli hot
avalanches was modeled using DAN-W (for two-
dimensional analyses) and DAN-3D (for three-di-
mensional analyses) numerical models (McDougall
& Hungr, 2004; Hungr & McDougall, 2009), that
are commonly used for the simulation of vol-
canic debris avalanches (Morelli et al., 2010; Sosio
et al., 2012b) and other landslides in which grain
collisions are dominant factors (McDougall &
Hungr, 2004; Sosio et al., 2008; Hungr & McDou-
gall, 2009; Sosio et al., 2012a). These numerical
codes both assume a simplified approach for the
simulation of mass flow motion that is consid-
ered an “equivalent-fluid” (McDougall & Hungr,
2004; Hungr & McDougall, 2009). DAN-W and
DAN-3D use the Lagrangian numerical method
to solve the depth-averaged St. Venant equations
(Monaghan, 1989; 1992; Benz, 1990). The momen-
tum equations evaluate an internal frictional rheol-
ogy, governed by an internal friction angle and by a
basal rheology, chosen by the modeler according to
one of the eight rheological kernel provided by the
numerical code (Hungr, 1995). In DAN-3D and in
DAN-W it is possible to use eight different rheo-
logical functions: frictional, plastic, Newtonian,
turbulent, Voellmy, Bingham, Coulomb frictional,
and power law.
Simulations were performed on three areas
roughly corresponding to those affected by the
hot avalanches of 1930 (zone A), 1944 (zone B),
and the area possibly involved in the 1906 event
(zone C). The zone A was used as calibration area,
since the data about the deposits distribution (areal
spread, thickness and travel distance), the velocity
of the flow and the temporal duration are available
from previous studies (Rittmann, 1931; Abbruzz-
ese, 1935; Di Roberto et al., 2014).
In performing these analyses, all the materials
involved in 1906 and 1944 hot avalanches were
assumed similar to those of 1930 (Bertagnini et al.,
2011). Subsequently, the parameters obtained for
1930 event (zone A) were used to calibrate the
model by back analysis and to find the input rheo-
logical parameters used later in the zone B and C.
The main characteristics of the three considered
events are listed in Table 1.
Three input files are necessary to run the simu-
lation: i) path topography; ii) source of landslide
depth and iii) erosion depth. The first file is the
landscape which includes the total sliding surface
(obtained in a post-event scenery) and it is rep-
resented by a Digital Elevation Model (DEM) at
very high resolution (50 cm cell size). This DEM
was obtained elaborating the 3D data acquired
during the airborne laser scanning survey car-
ried out in 2012 by the BLOM company (www.
blomasa.com) using a Leica ADS80 device (verti-
cal accuracy ± 10/20 cm and horizontal accuracy
± 25 cm).
Table 1. Main characteristics of the three studied
events: zone A, B, and C.

1496
The source depth file corresponds to the thick-
ness of the sliding mass before the collapse. This
file, stored as grid, was defined considering the
area where the accumulation of spatter and fall-
out deposits was observed during the 1930 par-
oxysm (Rittmann, 1931; Bertagnini et al., 2011;
Di Roberto et al., 2014) or argued for the 1906
and 1944 eruptions (e.g. considering where topog-
raphies exceeds 29°; for more details see Apuani
et al., 2005). From the source data file, the soft-
ware computes the area covered by the critical
mass and the volume of slide material contained
in each cell of the global reference grid. The initial
value of 1 m for the thickness of the sliding mate-
rial was established according to the descriptions
of Rittmann (1931) about the material accumu-
lated on the summit of the volcano and involved
in the collapse.
The erosion depth file distinguishes the entrain-
ment area and it is stored as grid. For our simula-
tion the erosion rate and the entrainment ratio are
both zero and the volume of the landslide does not
change during the runout. This is clearly suggested
by the 1930 hot avalanche deposits observed on the
field by Di Roberto et al. (2014).
For the elaborations constrains imposed by
DAN-3D all grids were resampled to spatial
resolution of 5 meters before inserting them in
the calculation code. The preparation of grids
used as input data for DAN-3D was done by
means of ESRI ArcGIS(TM) and Golden Software
SURFER(TM).
The duration of a simulation depends on three
main parameters: a) the size of the global refer-
ence grid, b) the length of the time step and c)
the number of the used particles. DAN-3D allows
the user to change the smoothing length constant
and the number of particles (McDougall, 2006).
For our simulations, 2000 particles and smooth-
ing length constant B = 4 were used. Since this
modeling software does not implement a routine
for the automatic stop, the simulation was sus-
pended in Zone A and B when the flow reached
the coastline. In the case in which the flow never
reached the sea level (Zone C) the simulation was
manually blocked 180 seconds after that the flow
stopped. Two input files are necessary for the two-
dimensional processing: path and top. The path
file is represented by the topographic profile of the
slope along which the mass moves, and the top file
defines the thickness of the source area along the
selected longitudinal section profile (initial slid-
ing mass geometry). The width of the landslide
can be also specified only for the pseudo-three-
dimensional visualization. The hot avalanche ini-
tial mass was split in 50 equally-spaced boundary
blocks with constant volume and shape factor 1
which indicates a rectangular cross-section of the
channel in which the material moved. In DAN-W
the simulation automatically stops when the flow
arrives at the end of the established path profile
or when the mass stops. The topographic and
geometrical input data (path and top files) were
directly extrapolated by the results obtained with
DAN-3D. In fact, the 2D analysis was performed
along profiles representing the lines connecting the
source regions with the most distal extension of the
deposits previously modeled with the 3D tool. For
the zone A two separate models were performed
along the two principal line of flow described by
Abruzzese (1935): one in Vallonazzo valley and the
other in San Bartolo valley. Through the results
of DAN-3D it was possible to calculate the vol-
umes channeled in the two valleys: 26000 m3 and
16000 m3 respectively. While for the zones B and
C only one line profile was chosen for each case
with a calculated slide volume of 29000 m3 and
34000 m3 respectively.
4 RESULTS
4.1 Analogue models
In total, 20 different experiments were conducted.
All models were constructed in a series of progres-
sive steps, starting from a simplified model and
moving toward a more realistic representation.
The rheological properties of the sand were tested
at different slope angles over the range of 35° to
40°, and the best angle to approximate the Sciara
del Fuoco condition, according to the data, is 35°
(Fig. 2). An increase in the thickness of the depos-
ited material always leads to frequent landslides.
The material inserted in the model breaks the slope
stability, generates a complex system of landslides
and increases the volume of material involved in
the sliding processes. The succession of phenom-
ena is derived from instability in the lower part of
the slope where landslide material progressively
accumulates.
This lower part achieves a state of equilibrium
each time until the arrival of new material from
the upper portion, which determines a remobiliza-
tion of the previous material in a larger landslide
(Fig. 2). The addition of material to the slope, cre-
ates a greater tension at a single point, changes the
slope equilibrium and is the origin of the material
sliding down the slope.Accumulation corresponds
to the continuous deposition of material (spatter
or lava) and to an increase in the thickness of the
slope.
The analogue models reveal that the internal
friction angles alone are not able to generate the
slope instability. The accumulation experiments
demonstrate that the accretion of a portion of

1497
the slope alters the flank instability and trig-
gers small landslides. Breaking the equilibrium
that exists in the slope forms a successive series
of landslides that propagate upwards along the
slope.
4.2 Numerical models
In the back analysis with DAN-3D relative to the
1930 hot avalanche the main rheological param-
eters within each rheological function were varied
using a trial-and-error procedure. For example,
we started varying the basal friction angle for the
Frictional rheology and the frictional coefficient
and turbulence parameter in the Voellmy rheology
(1955). This was executed in order to obtain values
of runout, velocity and duration of the flow match-
ing as much as possible those deriving from the lit-
erature and from the more recent studies. The best
results for the simulation of the 1930 hot avalanche
(Figs. 3a, b) were obtained using a Voellmy rheo-
logical model with a frictional coefficient f = 0.19
and a turbulence parameter ξ = 1000 m/s. The flow
length is the main parameter used during the back
analysis to constrain the model Historical accounts
and field data revealed that during the 1930 event
incandescent flows moved into the S. Bartolo and
Piscità (Vallonazzo) valleys. The first one stopped
few meters from the church of S. Bartolo, while the
second one had a major impulse and arrived at the
coastline (Abbruzzese, 1935; Rittmann, 1931; Di
Roberto et al., 2014).
The simulation was able to reproduce the behav-
ior of these two main flows with a first event
along the Vallonazzo valley reaching the sea level
at the urbanized area of Piscità after 135 seconds
and a second one in the San Bartolo valley stop-
ping close to the San Bartolo church, after about
120 seconds. The model successfully simulated also
the 1930 event deposits’ thicknesses, matching the
order of magnitude of those measured by Di Rob-
erto et al. (2014) during their field survey. In the
San Bartolo valley, at about 780 m a.s.l. the meas-
ured thickness is about 4.5 m, versus a simulated
thickness of 1.6 m. In the NE rim of Vallonazzo
valley at about 300 m a.s.l, the measured and simu-
lated thicknesses are instead fully comparable and
attain 30–40 cm, whereas close to the outlet of
Vallonazzo valley, the measured thickness is about
100 cm and the simulated thickness is 60 cm. The
areas impacted by the avalanche and covered by the
deposit are, for both impluviums as a whole, 3.6 ×
105 m2 and 0.8 × 105 m2, respectively. The outcomes
of DAN-W code along the considered sections of
this area show a good correspondence with the 3D
simulation (Fig. 3).
The rheological parameters obtained with this
procedure for 1930 hot avalanche were then used
in the simulation of the events occurred in zones
B (1944) and C (1906). The simulation results in
the zone B show that the flow moves in “Le Sch-
icciole” valley and reaches the sea in 88 seconds.
In this case, the simulated thickness of 60 cm well
matches the field measurement (Di Roberto et al.,
2014). The area impacted by the avalanche and
covered by the deposit in zone B are 3.1 × 105 m2
and 0.8 × 105 m2, respectively. The 2D simulation
also obtained similar results. The simulation in
the zone C shows that the flow never reaches the
sea level. After an initial spreading into a flat area
above the village of Ginostra, the flow moves (com-
pactly canalized and with a significant change of
direction) inside the village and stops after about
180 seconds. In this case, no records that describe
the flow dynamics, the runout or the thickness of
the deposits exist. Thus, any direct comparison
between simulation results and field evidences is
not possible. In zone C, the areas impacted by the
avalanche and covered by the deposit are, for both
impluviums as a whole, 1.5 × 105 m2 and 0.3 × 105
m2, respectively. Also in this case, DAN-W results
show a very good correlation with the DAN-3D
simulations.
Figure 2. Initial model set-up. Each series is built
into a plexiglass box on an inclined plane. The incli-
nation of the plane change between 25° and 45°. The
material, mobilized by the first landslide movement,
remains in the lower part of the model and creates a
potentially unstable area which generates the second
landslide movement. Afterward, a series of progres-
sive landslides were triggered and were all localized
approximately 10 cm up the slope (from Nolesini et al.,
2013).

1498
5 DISCUSSION
Hot avalanches are able to travel long distances
with destructive consequences. So, the evaluation
of their mobility is important for the assessment
of the areas at risk in future events. The morpho-
metric parameters that best describe their mobility
are: the height of vertical drop (H), the runout dis-
tance (L), the volume (V) and the covered area (A).
Hayashi and Self (1992), Calder et al. (1999) and
Saucedo et al. (2005) used different relationships
between these four above mentioned parameters to
explain the mobility of hot avalanches. In particu-
lar, Hayashi and Self (1992) asserted that different
material properties rather than different emplace-
ment mechanisms appear to be the best explanation
in the Log (V) versus Log (H/L) plot (Fig. 4a). If
the historical data of hot avalanches and volcanic
avalanches are compared, the plot shows an inverse
correlation with coincident regression lines for the
two types of deposits, implying that the material
properties are similar for both phenomena.
The main difference is represented by the posi-
tion in the graph. In fact, by plotting the data of
hot avalanches found by Calder et al. (1999) and
Saucedo et al. (2005), it is easily noticeable that
they usually have smaller volumes and higher H/L
ratios than volcanic avalanches. This is in good
agreement with the general results obtained from
the work of Corominas (1996). The H/L ratios
resulting from the analysis of the hot avalanches
of Stromboli are 0.42, 0.57 and 0.39 for A, B and
C zones respectively (Fig. 4b).
Further analysis of the mobility of Strom-
boli hot avalanches was performed considering
the parameter A/V2/3 used by Calder et al. (1999)
Figure 3. Results of DAN-3D simulation of a) flow thickness and b) deposit thickness for the zone A (1930-like
event); comparison between the DAN-W and DAN-3D for the c) Vallonazzo and d) San Bartolo valleys.

1499
to discriminate between the mobility of hot ava-
lanches at Soufriere Hills Volcano (Montserrat).
The Stromboli events show that the A/V2/3 factor
varies between 60 (zone A) and 117 (zone B), being
comparable to what measured by Calder et al.
(1999) for small volume hot avalanches.
6 CONCLUSIVE REMARKS
Hot avalanches at Stromboli volcano were ana-
lyzed by means of analogue models for the trig-
gering mechanism evaluation. Then DAN-W and
DAN-3D numerical codes were used to investigate
the behavior evaluation of the collapsed material
during the runout using as test case the 1906, 1944
and 1930 events. The main outcomes can be sum-
marized as follow:
− DAN-3D was able to reproduce the extension
and the order of magnitude of the thickness of
two hot avalanches reported in the literature;
− the best modeling results on the 1930 hot ava-
lanche were obtained using a Voellmy model
with frictional coefficient of f = 0.19 and a tur-
bulence parameter ξ = 1000 m/s;
− the outcomes of DAN-W code along the consid-
ered sections show a very good correspondence
with the 3D simulation;
− the suitability of analogue modeling to under-
stand the triggering mechanism and the evolu-
tion of landslide along the Sciara del Fuoco
depression has been verified; moreover the best
analogue modeling results are obtained with
35° inclination slope and reveal that the inter-
nal friction angles alone are not able to gen-
erate the slope instability. The accumulation
of material on the slope (lava, spatter) trigger
little landslides, developing in retrogressive
landslide;
− H/L index values resulting from the analysis of
the Stromboli hot avalanches are respectively
0.42, 0.57 and 0.39 for the three zones, and are in
good agreement with the inverse trend in reduc-
tion of H/L with increase of landslides volumes
(V) and H/L values less of 0.6 for landslides vol-
umes below 10–4 km3 (Corominas 1996);
− the estimated A/V2/3 factor for Stromboli hot
avalanches varies between 60 and 117, being
comparable with that of small volume events
generated by column collapse (Calder et al.,
1999).
REFERENCES
Abbruzzese, D. 1935. Sulla catastrofica esplosione dello
Stromboli dell’11 settembre 1930. Gioenia Proc. Soc.
Nat. Sci., 1, 1–13.
Apuani, T., Corazzato, C., Cancelli, A. & Tibaldi A. 2005.
Physical and mechanical properties of rock masses at
Stromboli: a dataset for volcano instability evaluation.
Bull Eng. Geol. Environ., 64(4), 419–431, doi: 10.1007/
s10064–005–0007–0.
Arrighi, S., Principe, C., & Rosi, M. 2001. Violent strom-
bolian and subplinian eruptions at Vesuvius during
post-1931 activity. Bull. Volcanol., 63(2–3), 126–150,
doi: 10.1007ls004450100130.
Barberi, F., Rosi, M. & Sodi, A. 1993. Volcanic hazard
assessment at Stromboli based on review of historical
data. Acta. Vulcanol., 3, 173–187.
Behncke, B., Calvari, S., Giammanco, S., Neri, M. & Pink-
erton, H. 2008. Pyroclastic density currents resulting
from the interaction of basaltic magma with hydro-
thermally altered rock: an example from the 2006sum-
mit eruptions of Mount Etna, Italy. Bull. Volcanol.,
70(10), 1249–1268, doi: 10.1007/s00445-008-0200-7.
Benz, W. 1990. Smooth particle hydrodinamics: a review.
In: Buchler, J. R. (ed.). The Numerical Modelling of
Nonlinear Stellar Pulsation. Kluwer Accademic, Dor-
drecht, 269–288.
Bertagnini, A., Di Roberto, A. & Pompilio, M. 2011.
Paroxysmal activity at Stromboli: lessons from the
past. Bull. Volcanol., 73(9), 1229–1243, doi: 10.1007/
s00445-011-0470-3.
Calder, E.S., Cole, P.D., Dade, W., Druitt, T.H., Hoblitt,
R.P., Huppert, H.E., Ritchie, L., Sparks, R.S.J. &
Young, S.R. 1999. Mobility of pyroclastic flows and
surges at the Soufriere Hills Volcano, Montserrat.
Geophys. Res. Letts., 26(5), 537–540.
Cole, P.D., Fernandez, E., Duarte, E., & Duncan, A.M.
2005. Explosive activity and generation mecha-
nisms of pyroclastic flows at Arenal volcano, Costa
Rica between 1987 and 2001. Bull. Volcanol., 67(8),
695–716.
Figure 4. a) Log(V) vs Log(H/L), b) H/L ratio of the
Stromboli hot avalanches (called here PDCs) compared
with world-wide events (data from Hayashi and Self,
1992; Calder et al., 1999; Suacedo et al., 2005).

1500
Corominas, J. 1996. The angle of reach as a mobility
index for small and large landslide. Can. Geotech. J.,
33, 260–271.
Davies, K.D., Quearry, M.W. & Bonis, S.B. 1978.
Glowing avalanches from the 1974 erup-
tion of the volcano Fuego, Guatemala. Geol.
Soc. Am. Bull., 89, 369–384. doi:10.1130/0016-
7606(1978)89 < 369:GAFTEO > 2.0.CO;2.
Di Roberto, A., Bertagnini, A., Pompilio, M. & Bisson,
M. 2014. Pyroclastic density currents at Stromboli vol-
cano (Aeolian Islands, Italy): a case study of the 1930
eruption. Bull. Volcanol., 76(6), 1–14, doi: 10.1007/
s00445-014-0827-5.
Di Traglia, F., Intrieri, E., Nolesini, T., et al. 2014. The
ground-based InSAR monitoring system at Stromboli
volcano: linking changes in displacement rate and
intensity of persistent volcanic activity. Bull. Volca-
nol., 76, 1–18, doi: 10.1007/s00445-013-0786-2.
Hayashi, J.N. & Self, S. 1992. A Comparison of Pyroclas-
tic Flow and Debris Avalanche Mobility. J. Geophys
Res., 97B6, 9063-9071.
Hazlett, R.W., Buesch, D., Anderson, J.L. Elan, R. &
Scandone, R. 1991. Geology, failure conditions, and
implications of seismogenic avalanches of the 1944
eruption at Vesuvius, Italy. J. Volcanol. Geoth. Res.,
47(3–4), 249–264, doi: 10.1016/0377-0273(91)90004-J.
Hungr, O. & McDougall, S. 2009. Two numerical
models for landslide dynamic analysis. Comput-
ers & Geosciences, 35(5), 978–992, doi: 10.1016/j.
cageo.2007.12.003.
Hungr, O. 1995. A model for the runout analysis of rapid
flow slides, debris flows, and avalanches. Can. Geo-
tech. J., 32, 610–623.
Mc Dougall, S. 2006. A new continuum dynamic model for
the analysis of extreme rapid landslide motion across
complex 3D terrain. Unpublished Ph.D Dissertation,
Department of Earth and Ocean Sciences, University
of British Columbia, 253.
McDougall, S. & Hungr, O. 2004. A model for the analy-
sis of rapid landslide motion across three-dimensional
terrain. Canad. Geotech. J., 41(6), 1084–1097, doi:
10.1139/T04–052.
Monaghan, J.J. 1989. On the problem of penetration in
particle methods. J. Comput. Phys. 82, 1–15.
Monaghan, J.J. 1992. Smoothed particle hydrodynamics.
Ann. Rev. Astron. Astrophys. 30, 543–574.
Morelli, S., Garduño Monroi, V.H., Gigli, G., Falorni,
G., Arreygue Rocha, E. & Casagli, N. 2010. The Tan-
citaro debris avalanche: characterization, propagation
and modeling. J. Volcanol. Geoth. Res., 193, 93–105,
doi: 10.1016/j.jvolgeores.2010.03.008.
Nairn, I.A. & Self, S. 1978. Esplosive eruptions and pyro-
clastic avalanches from Ngauruhoe in February 1975.
J. Volcanol. and Geoth. Res., 3(1–2), 39–60.
Nolesini, T., Di Traglia, F., Del Ventisette, C., Moretti, S.
& Casagli, N. 2013. Deformations and slope instability
on Stromboli volcano: Integration of GBInSAR data
and analog modeling. Geomorphology, 180, 242–254.
Ponte, G. 1948. Attività straordinaria dello Stromboli.
Annals of Geophysics, 1(2), 200–202.
Riccò, A. 1907. Sull’attività dello Stromboli dal 1891 in
poi. Boll Soc Sismol Ital, 12, 183–205.
Rittmann, A. 1931. Der Ausbruch des Stromboli am 11.
September 1930. Z Vulkanologie 14, 47–77.
Rosi, M., M. Pistolesi, A. Bertagnini, P. Landi, Pom-
pilio, M. & Di Roberto, A. 2013. Stromboli Volcano,
Aeolian Islands (Italy): present eruptive activity and
hazard. Geological Society, London, Memoirs, 37(1),
473–490.
Rotonda, T., Tommasi, P., & Boldini, D. 2009. Geome-
chanical characterization of the volcaniclastic material
involved in the 2002 landslides at Stromboli. Journal of
geotechnical and geoenvironmental engineering, 136(2),
389–401.
Saucedo, R., Macıas, J.L., Sheridan, M.F., Bursik, M.I &
Komorowski, J.C. 2005. Modeling of pyroclastic flows
of Colima Volcano, Mexico: implications for hazard
assessment. J. Volcanol.Geoth. Res., 139, 103–115.
Sosio, R., Crosta, G.B. & Hungr, O. 2012b. Numerical
modeling of debris avalanche propagation from col-
lapse of volcanic edifices. Landslides 9, 315–334, doi:
10.1007/s10346-011-0302-8.
Sosio, R., Crosta, G.B., & Hungr, O. 2008. Complete
dynamic modelling calibration for the Thurwieser rock
avalanche (Italian Central Alps). Engineering Geology
100, 11–26, doi: 10.1016/j.enggeo.2008.02.012.
Sosio, R., Crosta, G.B., Chen, J.H. & Hungr, O. 2012a.
Modelling rock avalanche propagation onto glaciers.
Quaternary Science Reviews, 47, 23–40.
Voellmy, A. 1955. Über die Zerstörunskraft von Lawinen.
Schweizerische Bauzeitung, 73, 212–285. (in German)
Yamamoto, T., Takada, A., Ishizuka, Y. Miyaji, N. &
Tajima, Y. 2005. Basaltic pyroclastic flows of Fuji vol-
cano, Japan: characteristics of the deposits and their
origin. Bull. Volcanol., 67(7), 622–633, doi: 10.1007/
s00445-004-0398-y.