See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/282570833

Physical and numerical modelling of shallow landslides

Article  in  Landslides · October 2015

DOI: 10.1007/s10346-015-0642-x

CITATIONS READS

9 284

3 authors, including:

Luca Schilirò Andrea Terrone

Sapienza University of Rome CH2M Dubai
12 PUBLICATIONS   73 CITATIONS    5 PUBLICATIONS   15 CITATIONS   

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Luca Schilirò on 12 October 2015.

The user has requested enhancement of the downloaded file.

 Original Paper
Landslides L. Montrasio I L. Schilirò I A. Terrone
DOI 10.1007/s10346-015-0642-x
Received: 17 June 2015
Accepted: 23 September 2015 Physical and numerical modelling of shallow landslides
© Springer-Verlag Berlin Heidelberg 2015
Abstract Physical modelling is an extremely useful tool for the
study of the triggering process of shallow landslides. For this
reason, in this work, numerous laboratory tests have been per-
formed using a specific flume test apparatus. A wide range of
initial soil conditions (i.e. porosity and water content) has been
investigated to analyze the induced effect on failure time and
mode, even simulating the presence of preferential flow directions
within the soil. Different tests have been performed also reproduc-
ing, on a laboratory scale, the landslide event occurred on October
1, 2009, in the area where the testing material was sampled (i.e.
Giampilieri, north-eastern Sicily, Italy). Furthermore, the experi-
mental results have been employed to verify the capability of
shallow landslide instability prediction (SLIP), a simplified stabil-
ity model for the prediction of shallow landslide occurrence, to
reproduce the triggering process.
Keywords Laboratory flume . Triggering model . Shallow
landslides . Stability analysis . Physically based model
Introduction
Shallow landslides (e.g. soil slip-debris flow) are instability phe-
nomena which mostly affect superficial deposits characterized by
small thickness (generally lower than 2 m) as a consequence of
intense and/or prolonged rainfall events. Their particular hazard is
largely due to the nearly total absence of premonitory signs: for
this reason, the study of the triggering of these processes is an
important research topic that can provide useful information in
the framework of landslide risk management. Recently, several
physically based models for the prediction of shallow landslide
occurrence were proposed (e.g. Montgomery and Dietrich 1994;
Iverson 2000; Baum et al. 2008; Simoni et al. 2008). In this type of
models, the landslide occurrence is determined through a slope
stability analysis, resulting in the calculation of the safety factor. In
other words, these approaches attempt to extend over large areas
simplified stability models which use the physical and mechanical
parameters of the slope material. The transient hydrology compo-
nent is incorporated assuming a slope-parallel flow either in its
steady state as a function of slope and drainage area (steady-state
models) or by dynamically evaluating the entire process from
rainfall to the transient response of the groundwater (dynamic
models) (Soeters and Van Westen 1996). Alongside these develop-
ments, to better understand the mechanisms and conditions that
lead to shallow landslide triggering, in the last years, several
laboratory-scale landslide studies were also performed using ex-
perimental apparatuses, generally composed of a sloping flume
and a sprinkler system to simulate the rainfall input. For instance,
Eckersley (1990) and Iverson et al. (2000) conducted large-scale
experiments to quantify the importance of the initial placement
(initial void ratio) and drainage conditions, whereas Wang and
Sassa (2001, 2003) and Olivares and Damiano (2007) triggered a
number of shallow landslides in sandy soils using a small flume,
demonstrating how the failure mode and the pore water pressure
generation (during and immediately after slope failure) depend
not only on the initial density but also, greatly, on the grain size
and fine particle content, which can have a significant impact also
on the mobility of rainfall-induced landslides. Using the same
techniques, other authors focused on different aspects related to
shallow landslide initiation, like the effect of soil depth on failure
mode and sediment discharge (Acharya et al. 2009) or the pore
water pressure generation at the interface of layers of different
permeability (Lourenco et al. 2006).
In this paper, we analyze the triggering mechanisms of soil slips
through laboratory flume experiments. Specifically, the aim of this
study is to evaluate the effect of the initial soil conditions (i.e.
porosity and water content) and the presence of preferential flow
directions on the triggering conditions of the soil sampled near the
village of Giampilieri (few kilometres south of Messina, north-
eastern Sicily, Italy). This village and the surrounding areas were
affected on October 1, 2009, by an extreme rainfall event, which
caused the occurrence of hundreds of shallow landslides. For this
reason, the material has been tested also reproducing, on a labo-
ratory scale, the hyetograph recorded during the event by Santo
Stefano di Briga monitoring station, which is the rain gauge closest
to the study area. Furthermore, for each test, the development of
the failure process has been numerically simulated with shallow
landslide instability prediction (SLIP), a simple physically based
model that predicts the triggering of shallow landslides on the
basis of the soil characteristics and the rainfall amount
(Montrasio 2000; Montrasio and Valentino 2008). In particular,
the model has been applied to verify its capability in matching the
experimental results, introducing a new calibration parameter
related to the initial saturation conditions.
The paper is organized as follows: after a detailed description of
the experimental apparatus and testing modality (BExperimental
set-up^ section), the theoretical basis of SLIP model is outlined
(BThe SLIP model^ section). Then, the results of flume tests
(BFlume test results^ section) and numerical simulations
(BResults of SLIP numerical simulations^ section) are provided
and discussed (BDiscussion^ section). Finally, in BConclusions^
section, the main conclusions are summarized.
Materials and methods
Experimental set-up
The flume test apparatus (Fig. 1a, b) is composed of a plexiglass
flume, 136 cm long, 50 cm wide and 35 cm high, connected to a
threaded rod that allows to change the flume angle. To assure the
same friction between the soil particles and the base of the flume
as of that of particles inside the flume, a rough plastic panel was
applied to the surface of the flume base. A stiff permeable barrier
was fixed in front of the soil to contain it after the failure, whereas
a video camera was used to monitor failure initiation time and
location. The rainfall simulator consists of a spray nozzle placed
above the flume. By keeping the supplied water pressure constant
at 3.2 bar, the actual rainfall intensity has been measured (1.1 mm/
min), verifying the rainfall uniformity through checking tests,
Landslides
 Original Paper
Table 1 Opening-closing stages of the rainfall simulator during the tests to rep-
roduce the Santo Stefano di Briga 15-min hyetograph
Test time Rainfall Test time Rainfall
(mm:ss) status (mm:ss) status
00:00 Close 25:59 Open
00:50 Open 26:31 Close
01:40 Close 27:52 Open
02:51 Open 29:37 Close
04:39 Close 30:00 Open
05:53 Open 32:30 Close
06:36 Close 33:29 Open
Fig. 1 a Scheme of the experimental apparatus. b Top view of the soil slightly
before the test
08:20 Open 34:01 Close
09:09 Close 35:45 Open
carried out prior to the start of each experiment by using 13 bins 10:43 Open 36:45 Close
fixed at the base of the flume. The tested material was sampled on 11:47 Close 38:19 Open
12:37 Open 39:10 Close
14:52 Close 40:00 Open
15:05 Open 42:30 Close
17:25 Close 42:39 Open
17:43 Open 44:50 Close
19:47 Close 45:16 Open
20:47 Open 47:14 Close
21:42 Close 48:09 Open
23:25 Open 49:21 Close
24:05 Close 50:00 Close

the slopes near Giampilieri, a small village located approximately

20 km SW of Messina (north-eastern Sicily, Italy). The choice of
testing this specific material can be explained considering that this

Fig. 3 a Detail of a drain used during the tests. b Disposition scheme of drains
Fig. 2 a Fifteen-minute hyetograph recorded at Santo Stefano di Briga between within the soil (left: top view; right: side view). Red and black colours indicate
4:00 and 9:00 pm of October 1, 2009. b Laboratory hyetograph used in the flume the part oriented towards the surface and the flume base, respectively. c
tests Disposition of the drains during the placing of soil into the flume

Landslides
Fig. 4 a Grain size distribution curve for the tested material. b Triaxial test
apparatus for large specimens

Fig. 5 a Conceptual sketch of the hydrological model in SLIP. b Shear strength

region is frequently affected by rainfall-induced shallow landslides versus degree of saturation for two soil samples. The symbols indicate the
(Schilirò et al. 2015). In particular, on the evening of October 1, experimental results from Fredlund et al. 1996, and the lines represent the cΨ*
2009, numerous soil slips were triggered by a severe rainstorm, fitting equation. c From left to right: evolution of saturation process through the
causing 37 casualties and serious damage to property and infra- soil macropores (from Montrasio and Valentino 2008)
structure. Considering the features of the event, it was used as a
reference for the execution of the flume tests. For this reason, the
flume angle was kept constant at 38°, in agreement with the With the aim of investigating a wide range of initial soil condi-
average slope observed within the 2009 landslide source areas, tions, in the first test series, four different porosity values (30, 35,
whereas the soil thickness (13 cm) was equal to approximately
one sixth of the real average soil thickness measured on site
(80 cm). In order to evaluate the most appropriate soil geometry,
different preliminary tests have been performed. In the first tests,
the soil volume was directly adjacent to the permeable barrier:
however, this solution has been discarded due to the excessive
resistance given by the barrier itself, resulting in an unrealistic soil
stability condition even in response to extremely prolonged rain-
fall inputs. For this reason, in order to achieve the least possible
disturbance of the slope conditions, a soil wedge has been placed
between the soil volume and the permeable barrier. The inclina-
tion of the wedge has been progressively decreased to 30°, in such
a way as to ensure that failure occurred within the soil volume and
not into the soil wedge. This geometry was considered suitable for
the experimental purposes, and then, it has been kept constant in
each test. Once the rainfall system was calibrated and the geom-
etry defined, 29 flume tests have been performed varying the
initial soil conditions (porosity and water content), simulating
the presence of preferential flow directions of infiltrating water
and using two different rainfall inputs. During each test, the
failure mode was observed and the triggering time was measured.

Table 2 Physical and mechanical properties of the tested material

Granulometric characteristics
Gravel (%) 58.1
Sand (%) 30.6
Silt (%) 9.5
Clay (%) 1.8
Physical and mechanical properties
Specific gravity (–) 2.725
Internal friction angle (°) 36.3 Fig. 6 a Temporal evolution of failure in soil with n 40 % and w 8 % (test n. 15).
Effective cohesion (kPa) 0 b Localization of failure in soils with n 35 % and different initial water content
(left: w 4 %—test n. 18; right: w 8 %—test n. 20)

Landslides
 Original Paper
Table 3 Initial soil conditions and failure time (Tf) for each of 17 tests performed with constant rainfall and absence of preferential flow directions (CN)
Test n (%) w (%) Sr (%) Tf (min) Test n (%) w (%) Sr (%) Tf (min)
1 30 8 50.8 – 10 37 6 27.8 13
2 30 10 63.6 – 11 37 8 37.1 11
3 30 12 76.3 – 12 37 10 46.4 7
4 35 4 20.2 29 13 40 4 16.3 12
5 35 6 30.4 22 14 40 6 24.5 9
6 35 8 40.5 13 15 40 8 32.7 6
7 35 10 50.6 7.5 16 40 10 40.9 4
8 35 12 60.7 3 17 40 12 49.0 -
9 37 4 18.6 17.5

37 and 40 %) and five initial water contents (4, 6, 8, 10 and 12 %)
were used, whereas in the subsequent tests, only the initial condi-
tions considered representative of the reference event (i.e. porosity
of 35 % and water content not exceeding 10 %) were imposed.
Before placing the soil into the flume, the established water con-
tent was obtained by wetting a specific quantity of oven-dried soil
with the quantity of water necessary to reach the desired water
content value. After the soil was set into the flume, the water
content was checked by sampling the soil in different points. The
initial porosities were attained by placing the soil in 2-cm
compacted layers parallel to the flume base. Considering that the
test geometry was fixed, each layer occupied a known volume, and
then, for each test, the exact weight of soil (having a specific water
content) required to fill that volume was calculated. Obviously, six
2-cm layers and one 1-cm layer at the top of the soil were necessary
to obtain the final geometry.
With regard to the artificial rainfall, tests have been per-
formed using a constant rainfall input or conveniently
opening-closing the rainfall simulator to reproduce the Santo
Stefano di Briga 15-min hyetograph related to the main phase
of the October 1 event, i.e. between 4:00 pm and 9:00 pm
(Fig. 2a, b). In this way, it is possible to reconstruct as
accurately as possible the triggering event, characterized by
short duration but extremely high-intensity peaks (e.g.
21.3 mm of cumulated rain between 7:00 pm and 7:15 pm,
corresponding to an intensity of approximately 1.4 mm/min),
for a total of 194 mm of rain falling in just 5 h. To keep the
same rainfall intensity recorded on site and considering that
the test soil thickness is approximately one sixth of the real
average soil thickness, a simulation time step of 2.5 min (15/6)
has been defined, during which the spray nozzle has been
activated for the time necessary to reach one sixth of the real
rain amount falling in the corresponding 15 min. However,
only in the case of the highest rainfall peak (7:00–7:15 pm), it
was necessary to increase the rainfall intensity up to 1.4 mm/
min to achieve the exact rainfall amount in 2.5 min. Opening
and closing stages of the rainfall simulator (Table 1) were
defined in such a way as to centre the rainfall peak within
the time step.
The presence of preferential flow directions of infiltrating water
has been simulated to account for the natural macropore structure
of soils (Flury et al. 1994). To simulate these preferential flow
channels, six small Bdrains^ have been made, approximately
30 cm long and 1 cm wide, composed of gravel enveloped in a
plastic net and placed into the soil according to a fixed scheme
(Fig. 3).

Fig. 7 Failure time versus initial water content (left) and degree of saturation Fig. 8 a Final configuration of soil with n 30 % and w 12 % (test n. 3): no failure
(right) for different initial porosity values (35, 37 and 40 %) occurs. b Shallow flows in soil with n 37 % and w 4 % (test n. 9)

Landslides
Table 4 Initial soil conditions, cumulated rainfall time (Tc), total failure time (Tf) and equivalent time for each of four tests performed with BGiampilieri scaled hyet-
ograph^ rainfall input and absence of preferential flow directions (GN)
Test n (%) w (%) Sr (%) Tc (min) Tf (min) Equivalent real time
18 35 4 20.2 26 44 8:15–8:30 pm
19 35 6 30.4 19 36 7:30–7:45 pm
20 35 8 40.5 12 22 6:15–6:30 pm
21 35 10 50.6 5 13 5:15–5:30 pm

The physical and mechanical properties of the tested material
were obtained by standard laboratory tests (Fig. 4). The soil is
characterized by a high content of gravel and sand, with minimum
amounts of silt and clay, whereas the resulting friction angle is
equal to 36.3°, assuming a zero value of the effective cohesion
(Table 2).
The SLIP model
SLIP is a mathematical model that calculates the safety factor by
applying the limit equilibrium method to an equivalent infinite
slope that is composed of two homogeneous soil portions: a
partially saturated portion and a fully saturated portion, which
represents the saturated zones around the preferential flow chan-
nels. Homogenization is used to obtain, with respect to the original
conditions, the loss of shear strength of an equivalent soil that is
stable in the presence of both saturated and partially saturated
zones. The model evaluates the destabilizing and stabilizing forces
(including the partial saturation contribution to the soil shear
strength) acting at the base of each soil slice (Fig. 5a), according
to Mohr-Coulomb strength criterion:
0 0
cotβtanφ ½Γ þ mðnw −1Þ þ C Ω
FS ¼
Γ þ m nw
with
Γ ¼ Gs ð1−nÞ þ n Sr
nw ¼ n ð1− Sr Þ
where c’ is the soil cohesion for effective stress, L is the length of
soil slice, and cΨ is the apparent cohesion due to the partial
saturation of soil. This last parameter is related to the matric
suction, which in turn, depends on the degree of saturation of
the soil (Montrasio and Valentino 2008). In the model, cΨ is
expressed as a function of the degree of saturation on the basis
of experimental results on different kinds of soil at low confining
pressures (Fredlund et al. 1996). The experimental points (shown
in Fig. 5b) can be fitted by a curve having a peak, whose value
depends on the type of soil and corresponds to a degree of
saturation approximately equal to 0.7. This function is given by
the following:
h i
cΨ * ¼ A Sr ð1−Sr Þλ ð6Þ
where A (kPa) is a parameter depending on the kind of soil and
linked to the peak shear stress at failure and λ is a numerical
parameter for data fitting that allows to shift the peak of the curve
along the abscissa axis. Indicative values of A and λ for the most
common kinds of soil on the Italian territory have been obtained
ð1Þ through a best fitting procedure between the cΨ* function and the
experimental results deriving from a number of suction-controlled
triaxial tests, as described in previous publications (Mari 2000;
Montrasio and Valentino 2003, 2007). Considering that the soil
slice is composed of an unsaturated (cΨ*≠0) and a saturated
(cΨ*=0) layer, these two layers can be homogenized in a single
ð2Þ stratum having an apparent cohesion as a function of the thickness
of the saturated stratum. From experimental results on stratified
soils of different characteristic values of cΨ* and different thickness
ð3Þ (from Silva 2000; Montrasio and Valentino 2007), the following
equation, which expresses the apparent cohesion homogenized in
the whole soil slice, has been obtained:

Ω¼
2
ð4Þ cΨ ¼ cΨ * ð1−mÞα ð7Þ
sin2β H γ w

where β is the slope angle, φ’ is the soil friction angle for Table 5 Initial soil conditions and failure time (Tf) for each of four tests performed
effective stress, m is the saturate fraction of the soil layer with with constant rainfall and presence of preferential flow directions (CD)
respect to its thickness (m=Hsat/H), Gs is the specific gravity
of soil solids, n is the porosity, Sr is the degree of saturation, Test n (%) w (%) Sr (%) Tf (min)
H is the thickness of the potentially unstable layer, γw is the 22 35 4 20.2 28
unit weight of groundwater and C’ is a parameter that repre-
23 35 6 30.4 20
sents the total cohesion given by the following:
h0 i 24 35 8 40.5 12
0
C ¼ c þ cΨ L ð5Þ 25 35 10 50.6 10.5

Landslides
 Original Paper
where ξ is a coefficient that quantifies the portion of rainfall
amount that does not infiltrate into the soil due to runoff process-
es and h is the rainfall depth. In order to simulate the decrease of
the m parameter in time due to the evapotranspiration, down-flow
and percolation, the model uses a simple negative exponential
function of time:

mðdecreaseÞ ¼ m0 expð−K T t Þ ð9Þ

where m0 is the initial value of the m parameter, t is the time

interval, and KT is a global drainage capability (its dimension is the
inverse of time) which is due not only to the intrinsic soil perme-
ability, but also to the presence of preferential down-flow ways, as
previously mentioned. Combining Eqs. 8 and 9, the final formu-
lation of m in time is given by following:

Fig. 9 Post-failure images of tests with preferential flow channels and constant
rainfall. The dashed lines indicate the fractures, while the numbers refer to the
location of the drains, according to the scheme reported in Fig. 3b: a w=10 % (test
n. 25) and b w=6 % (test n. 23)
where α is a parameter that gives a non-linear trend to the curve
representing the function. The α parameter assumes the value of
3.4 according to experimental tests (Montrasio and Valentino
2008). As regard the infiltration process during a rainfall event,
SLIP hypothesizes a water flow through the soil macropores that
generates a progressively saturation of non-adjacent volumes of
soil (dual-porosity effect) (Zhang and Zhang 2009) while, if the
raining process persists, the saturated portions of soil extend and
become continuous (Fig. 5c). The sliding process starts when a
relatively wide continuous stratum of saturated soil has formed
(Montrasio et al. 2011). However, from a physical point of view, the
model must take into account for different forms of down-flow
(such as the seepage through the substratum or through other
preferential ways) and other phenomena like evapotranspiration
that cause a decrease of water amount into the soil in absence of
rainfall. The modelling of this complex mechanism is summarized
into the abovementioned m parameter, whose increase after a
rainfall event is given by the following:
ξh
mðincreaseÞ ¼
nH ð1−Sr Þ
ω
ξ X
mðt Þ ¼ hi exp½−K T ðt i −t 0 Þ ð10Þ
nH ð1−Sr Þ i¼1
where hi and ti are, respectively, the rainfall depth and time at the
ith time interval. Therefore, m is computed at each time step and
directly correlates the rainfall depth (h) with the safety factor. A
more detail explanation of the assumptions, fundamental concepts
and equation derivation of the SLIP model can be found in
previous publications (e.g. Montrasio 2000; Montrasio and
Valentino 2008; Montrasio et al. 2011, 2012).
Results
Flume test results
In each of performed tests, the triggering mechanisms induced by
rainfall have been analyzed in detail. The first observation is that
the failure mode is generally extremely rapid, with no evidences of
incipient instability (Fig. 6a), and involves a soil thickness between
7 and 10 cm; thus, considering that the total soil thickness is 13 cm,
the failure surface always develops within the soil profile and not
at the contact between soil and flume base. Furthermore, it is
worth noting that, if the initial water content is low (e.g. 4 %),
failure typically occurs in the upper part of the flume and vice
versa (Fig. 6b).
As regard the first set of tests (constant rainfall and absence of
preferential flow directions), a wide range of initial soil conditions
has been investigated (Table 3) and different observations can be
ð8Þ
made: (1) As the initial water content (w) or porosity (n) increases,

Table 6 Initial soil conditions, cumulated rainfall time (Tc), total failure time (Tf) and equivalent time for each of four tests performed with BGiampilieri scaled hyet-
ograph^ rainfall input and presence of preferential flow directions (GD)
Test n (%) w (%) Sr (%) Tc (min) Tf (min) Equivalent real time
26 35 4 20.2 27 46 8:30–8:45 pm
27 35 6 30.4 18 31 7:00–7:15 pm
28 35 8 40.5 13 23 6:30–6:45 pm
29 35 10 50.6 12 19 5:45–6:00 pm

Landslides
the failure time decreases on equal porosity or water content, Table 7 Input parameters for SLIP model
respectively (Fig. 7); (2) in tests with n=30 % no failure occurred, H (soil depth) 13 (cm)
but only a gradual erosion of the superficial layers of the toe
(Fig. 8a), even with the wettest initial soil conditions (i.e. h (rain depth) Variable (cm)
w=12 %). On the other hand, tests with n=40 % and w=12 % have β (slope) 38 (°)
not been performed due to the detachment of the soil from the n (porosity) Variable (%)
uphill barrier during the flume tilting; and (3) in tests with n=37–
40 % and w=4 %, after about 10 min, a superficial fracture in the Gs (specific gravity) 2.725 (–)
upper part of the flume appeared. However, this fracture did not φ′ (internal friction angle) 36.3 (°)
induce the detachment of soil. Instability occurred by means of c′ (effective cohesion) 0 (kPa)
progressive shallow flows (Fig. 8b) caused by the advance of the
wetting front. Srreal (saturation ratio) Variable (%)
Unlike the Bconstant rainfall^ input, in the second type of tests, Srm (saturation ratio) Variable (%)
opening and closing stages of rainfall system have been conve- ξ (runoff parameter) Variable (%)
niently alternated to reproduce the Santo Stefano di Briga 15-min
hyetograph; thus, both the cumulative opening time of rainfall A (apparent cohesion parameter) 50 (kPa)
system (Tc) and the total test duration (Tf), which hypothetically λ (apparent cohesion parameter) 0.4 (–)
corresponds to an equivalent real time, have been measured for α (SLIP parameter) 3.4 (–)
each test (Table 4).
In the following tests (Table 5), the presence of preferential flow KT (global drainage capability) 1.5×10–5 (m/s)
directions has been simulated introducing six small drains into the
soil (Fig. 3) and maintaining a constant rainfall input. It is funda-
mental noting that in all of four tests, the mobilized soil volume is
However, also in these tests, the mobilized soil volumes are greater
greater than that of the preceding tests. Furthermore, it has been
than those observed in the corresponding tests without infiltration
observed that with w=8–10 %, a unique fracture develops in
channels. Finally, it is important to note that test reproducibility
correspondence of the drains located in the upper part of the
has been assessed by repeating the most complex tests (i.e.
flume (Fig. 9a), whereas with lower water contents, failure occurs
Giampilieri scaled hyetograph rainfall and presence of preferential
with a double fracture system, where the second failure surface is
flow directions) and measuring the difference between failure
located in correspondence of the drains placed in the lower part of
times: in all of four tests, this difference is lower than 1 min.
the flume (Fig. 9b). The same configuration with preferential flow
directions has been tested using the BGiampilieri scaled
Results of SLIP numerical simulations
hyetograph^ rainfall (Table 6). Analyzing the failure mode, it
The input parameters of SLIP model (Table 7) have been found by
results a multi-fractured pattern with the drier (w=4–6 %) initial
using data provided from preceding works (Montrasio and
soil conditions (Fig. 10a), while with the wetter ones (w=8–10 %),
Valentino 2008; Montrasio et al. 2014) and by calibration based
the instability is localized in the lower part of the flume (Fig. 10b).
on the results of flume tests. More specifically, α has been main-
tained equal to 3.4 (Montrasio and Valentino 2008), whereas A, λ,
KT, ξ and Srm have been adjusted through a procedure of back
analysis, carrying out different numerical simulations to repro-
duce the performed flume tests. If A, λ and KT values substantially
agree with those referred to this soil type (Montrasio and
Valentino 2008; Montrasio et al. 2011), ξ and Srm have been mod-
ified for each test according to the varying initial soil conditions
(i.e. porosity and presence of preferential flow directions), keeping
all the other parameters constant. In fact, these two parameters
vary with soil density, considering that ξ is the parameter that
quantifies the runoff component whereas Srm is a new calibration

Fig. 10 Post-failure images of tests with preferential flow channels and

BGiampilieri scaled hyetograph^ rainfall: a w=4 % (test n. 26) and b w=8 %
(test n. 28) Fig. 11 Hydraulic scheme of SLIP model

Landslides
 Original Paper
Table 8 Initial soil conditions, calibrated input parameters (Srm, m0 and ξ), real failure time (Tr), predicted failure time (Tp) and time difference (ΔTrp) for flume tests
performed with 35 % porosity and absence of preferential flow channels
Test Type n w Srreal Srm m0 ξ Tr Tp ΔTrp
(%) (%) (%) (%) (–) (%) (min) (min) (min)
4 CN 35 4 20.2 15.0 0.06 20 29 28 1
5 CN 35 6 30.4 15.0 0.18 20 22 21 1
6 CN 35 8 40.5 15.0 0.30 20 13 14 1
7 CN 35 10 50.6 15.0 0.42 20 7.5 8 0.5
18 GN 35 4 20.2 15.0 0.06 20 44 47 3
19 GN 35 6 30.4 15.0 0.18 20 36 38 2
20 GN 35 8 40.5 15.0 0.30 20 22 24 2
21 GN 35 10 50.6 15.0 0.42 20 13 14 1
Mean 1.4
CN constant rain, GN BGiampilieri scaled hyetograph^ rain

parameter, introduced and tested in a recent application at the
field scale (Montrasio et al. 2014), which determines the equivalent
thickness of the initial saturated layer (m0), according to a water
balance between the real amount of water and the fictitious model
(Fig. 11) expressed by the following:
n H Sr real ¼ nð1−m0 Þ H Sr m þ n m0 H
ðSr real −Sr m Þ
m0 ¼
ð1−Sr m Þ
where Srreal is the real initial saturation ratio of the soil for each
flume test, depending on the initial water content, porosity (n) and
the specific gravity of soil solids and Srm is a fictitious initial
degree of saturation of soil that allows the separation in two layers:
the first (whose thickness is equal to (1−m0) H) assumes the form
of a partially saturated layer (Sr=Srm) that contributes to shear
resistance through apparent cohesion; the second (m0 H) assumes
the form of a saturated layer at the bottom of the soil. In this
respect, it is important to note that, for each test, the soil has been
left into the flume for approximately 4 h before the test started, in
such a way as to allow the pore water to reach the lower portion of
the soil. This assumption of the model was made in order to
compare tests with same initial porosity, keeping constant all
the other input parameters (with the only exception of m0,
ð11Þ function of the initial varying water content). For each test,
the simulated failure time has been compared with the real
ð12Þ one, calibrating ξ and Srm with the aim to minimize the time
difference. Tables 8 and 9 report the calibration results for
flume tests with porosity equal to 35 %: as can be noted, the
mean variation between real and predicted failure time is
comprised between 1.4 and 2.2 min in absence and presence
of preferential flow directions, respectively. Furthermore, it is
important to emphasize that these results have been obtained
also varying the rainfall input (constant or Giampilieri scaled
hyetograph rainfall). The results obtained by reproducing the
flume tests with porosity equal to 37 % (Table 10) and 40 %
(Table 11) are substantially similar: in this case, the mean
difference is lower than 2.5 min.

Table 9 Initial soil conditions, calibrated input parameters (Srm, m0 and ξ), real failure time (Tr), predicted failure time (Tp) and time difference (ΔTrp) for flume tests
performed with 35 % porosity and presence of preferential flow channels
Test Type n w Srreal Srm m0 ξ Tr Tp ΔTrp
(%) (%) (%) (%) (–) (%) (min) (min) (min)
22 CD 35 4 20.2 15.0 0.06 10 28 25 3
23 CD 35 6 30.4 15.0 0.18 10 20 19 1
24 CD 35 8 40.5 15.0 0.30 10 12 12 0
25 CD 35 10 50.6 15.0 0.42 10 10.5 6 4.5
26 GD 35 4 20.2 15.0 0.06 10 46 44 2
27 GD 35 6 30.4 15.0 0.18 10 31 32 1
28 GD 35 8 40.5 15.0 0.30 10 23 22 1
29 GD 35 10 50.6 15.0 0.42 10 19 14 5
Mean 2.2
CD constant rain, GD BGiampilieri scaled hyetograph^ rain

Landslides
Table 10 Initial soil conditions, calibrated input parameters (Srm, m0 and ξ), real failure time (Tr), predicted failure time (Tp) and time difference (ΔTrp) for flume tests
performed with 37 % porosity and absence of preferential flow channels
Test Type n w Srreal Srm m0 ξ Tr Tp ΔTrp
(%) (%) (%) (%) (–) (%) (min) (min) (min)
9 CN 37 4 18.6 10.0 0.09 10 17.5 22 4.5
10 CN 37 6 27.8 10.0 0.20 10 13 16 3
11 CN 37 8 37.1 10.0 0.30 10 11 11 0
12 CN 37 10 46.4 10.0 0.40 10 7 5 2
Mean 2.4
CN constant rain

Discussion
The results of flume tests show the influence of initial soil condi-
tions (porosity and water content) on times and modalities of
slope failure. With regard to the porosity, it is worth noting that
the widest range of failure times is observed for n=35 % (Fig. 7);
thus, this specific porosity level is particularly sensitive to varia-
tion of the initial water content. On the contrary, in tests with
n=37–40 %, there is a lower difference between failure times to
changing the initial water content. However, for these porosity
values, processes which cannot be properly classified as soil slips
occurred both with the drier (i.e. w=4 %) and wetter (w=12 %)
initial soil conditions. In the first case, the occurrence of progres-
sive shallow flows (Fig. 8b) can be explained with the scarce
coherence of the material that favours the formation of a superfi-
cial zone wetter than the underlying one, causing instability only
in the upper part of the soil profile. In the second case, the
detachment of the soil during the flume tilting was caused by the
excessive looseness and wetness of the material. Finally, n=30 %
can be considered a too low porosity value for the investigated
phenomenon, given that the excessive soil consolidation results in
a very high runoff (and, consequently, in an extremely low infil-
tration rate), which does not induce any mass movement (even
with the wettest initial soil conditions).
Therefore, on the basis of these first results, in the following tests,
we decided to focus our attention on the n=35 % case, in order to
evaluate the effect of the water content variation. In this respect, it is
worth noting that, in tests with Giampilieri scaled hyetograph rainfall
input and absence of preferential flow directions (GN), an initial
water content varying between 4 and 8 % induces instability in a real-
time corresponding interval compatible with the real landslide event,
whose main phase occurred between 6:00 pm and 9:00 pm (Table 4).
On the other hand, a water content equal to 10 % does not seem to be
as representative because failure occurs too early (13 min, equivalent
to 5:15 pm–5:30 pm). In the absence of field data, this result repre-
sents at least a reference point to hypothesize the possible soil
conditions preceding the 2009 event, even though a laboratory
model for the study of rainfall-induced landslides cannot reproduce
in detail the real site conditions, due to unavoidable scale and
laboratory effects.
If we compare the cumulative opening times of rainfall system
(Tc) with the failure times obtained in the constant rainfall tests
(Fig. 12), it can be noted that trends are very similar, but failure
times are slightly lower in the case of Bhyetograph^ rainfall tests.
This feature can be explained considering that during the closing
stages, water continues to infiltrate, reducing the runoff
component.
With regard to the effect of the macroporosity, the development
of fracture systems in correspondence of the drains (Fig. 9) can be
explained with two different hypotheses: (a) the preferential flow
through the gravel inclusions induces the build-up of local water
pressures acting on the permeable soil surrounding the drain,
thereby causing the triggering of the instability and (b) the pres-
ence of the drain forms a local weakness zone within the soil
volume due to the lower compaction of soil. Although the soil
below the drain was well compacted before the drain was placed,
at the current state of work, we cannot exclude an eventual
different compaction of the soil. However, it is important to stress
that tests with preferential flow elements represent only a

Table 11 Initial soil conditions, calibrated input parameters (Srm, m0 and ξ), real failure time (Tr), predicted failure time (Tp) and time difference (ΔTrp) for flume tests
performed with 40 % porosity and absence of preferential flow channels
Test Type n w Srreal Srm m0 ξ Tr Tp ΔTrp
(%) (%) (%) (%) (–) (%) (min) (min) (min)
13 CN 40 4 16.3 5.0 0.12 0 12 16 4
14 CN 40 6 24.5 5.0 0.21 0 9 11 2
15 CN 40 8 32.7 5.0 0.29 0 6 6 0
15 CN 40 10 40.9 5.0 0.38 0 4 2 2
Mean 2
CN constant rain

Landslides
 Original Paper
Fig. 12 Failure time versus initial water content/saturation ratio for constant (blue
line) and BGiampilieri scaled hyetograph^ (red line) flume tests in absence of
preferential infiltration channels
preliminary analysis of the effect of the macroporosity on the
triggering mechanisms of soil slips that will be fully investigated
in future experimental tests. Nevertheless, several general consid-
erations about the soil behaviour at failure can still be inferred
from such tests. Firstly, the results indicate that the preferential
flow directions influence the failure mode rather than the trigger-
ing time. In fact, the resulting failure times are substantially
similar to those obtained in absence of preferential flow directions
both with the constant and Giampilieri scaled hyetograph rainfall
input, excluding the w=10 % case, where the instability is delayed
by about 3 and 7 min, respectively (Figs. 13 and 14). On the
contrary, it has been observed that the different infiltration process
causes the progressive saturation of different portions of soil and
Fig. 13 Failure time versus initial water content/saturation ratio for constant
rainfall input flume tests in absence (blue line) and presence (red line) of
preferential infiltration channels
Landslides
so the development of multi-fracture systems during the triggering
phase, in particular with the drier initial soil conditions (Figs. 9b
and 10a). Furthermore, in all tests, the mobilized volumes are
greater than those observed in the corresponding tests without
infiltration channels.
As regard the application of the SLIP model, it can be said that the
simulation results substantially agree with the outcome of the flume
tests, considering that the mean variation between real and predicted
failure time is in the order of a few minutes, even for the longest tests.
These results confirm the good predictive capability of the model, in
particular with different initial conditions. In this respect, it is worth
noting that both ξ and Srm accordingly change to changing the initial
soil conditions. For instance, ξ is lower in the case of presence of
preferential flow directions, given that a soil with higher macroporosity
is affected by a lower surface runoff, whereas Srm decreases as the
porosity increases, considering that a greater porosity corresponds to
a lower degree of saturation, on equal initial water content.
Furthermore, it is important to note that, in the case of n=35 % and
presence of preferential flow directions (Table 9), the major differences
between real and predicted failure time occur for the wettest initial
condition (w=10 %). As previously mentioned, the same differences
were observed, for the same initial water content value, comparing the
results of tests with/without preferential flow directions (Figs. 13 and
14). This means that the macroporosity begins to influence the trigger-
ing time only above a certain soil moisture level, modifying the infil-
tration process in such a way as the model tends to slightly
underestimate the failure time. On the contrary, in tests with n=37–
40 %, the major discrepancy between numerical simulations and
experimental tests occurs for the driest initial condition (w=4 %).
However, in this case, the time difference is probably related to the
particular density state that, as described above, induces the failure only
in the superficial soil layers. For this reason, the numerical model,
which considers the entire soil profile involved in the instability pro-
cess, predicts a triggering time slightly higher than the real one. In this
respect, it is important to stress that SLIP model is a numerical tool for
Fig. 14 Failure time versus initial water content/saturation ratio for BGiampilieri
scaled hyetograph^ rainfall input flume tests in absence (blue line) and presence
(red line) of preferential infiltration channels
global stability which does not analyze the behaviour of the represen-
tative volume element. The model bypasses the transient analysis of the
infiltration process assuming the formation, as a consequence of a
rainfall event, of the final condition of a perched water table, then
calculating the safety factor. Although no measurements of pore water
pressure have been recorded during the tests, in previous works (i.e.
Montrasio and Valentino 2007, 2008), experimental measurements of
suction and pore pressure have been recorded during flume tests under
different rainy conditions, confirming some hypotheses of the simpli-
fied stability model. Specifically, in all tests, shallow soil layers reach a
zero value of suction well before the deeper ones and the first slip
occurs only after the matric suction has dropped to zero everywhere.
This condition corresponds to a state of complete saturation of the soil
layer that is in contact with the impermeable bed (Montrasio and
Valentino 2007). Finally, with regard to the water content, in a recent
paper, it has been demonstrated how the mean equivalent water
content predicted by the model over a 6-month period is consistent
with the field measurements (Montrasio et al. 2014).
Conclusions
In this study, numerous flume tests have been performed to analyze the
triggering process of soil slips and to verify the reliability of SLIP, a
simplified stability model that is able to predict the occurrence of these
phenomena. The results of flume tests describe the effect of the initial
soil conditions on the triggering process, both in terms of triggering
time and failure mode. In particular, it has been emphasized how the
increase of the initial water content or porosity anticipates the trigger-
ing time (in particular, in the case of n=35 %), whereas the
macroporosity influences almost exclusively the failure mode,
favouring the mobilization of major soil volumes and the development
of multi-fracture systems during the triggering phase. In addition, an
estimate of the hypothetical initial conditions preceding the October 1,
2009, event has been made through the Giampilieri scaled hyetograph
rainfall tests, which suggest an initial water content varying between 4
and 8 %, corresponding to a degree of saturation of 20.2–40.5 %. With
regard to the numerical simulations, it results that the SLIP model,
further improved through the addition of a new calibration parameter
(Srm), is able to predict with a good approximation the triggering time
of all the performed tests, specifically accounting for the different initial
soil conditions.
Therefore, if we consider that the experimental tests, excluding
the laboratory and scale effects, reproduce as accurately as possi-
ble the site conditions, on the basis of the obtained results, we
suggest that the model can be a useful tool for shallow landslide
susceptibility assessment even at the regional scale.
References
Acharya G, Cochrane TA, Davies T, Bowman E (2009) The influence of shallow landslides
on sediment supply: a flume-based investigation using sandy soil. Eng Geol 109:161–
169. doi:10.1016/j.enggeo.2009.06.008
Baum RL, Savage WZ, Godt JW (2008) TRIGRS—a Fortran program for transient rainfall
infiltration and grid-based regional slope-stability analysis, version 2.0. U.S. Geological
Survey Open-File Report 2008-1159: 75 pp.
Eckersley D (1990) Instrumented laboratory flowslides. Geotechnique 40:489–502.
doi:10.1680/geot.1990.40.3.489
Flury M, Flühler H, Jury WA, Leuenberger J (1994) Susceptibility of soils to preferential
flow of water: a field study. Water Resour Res 30:1945–1954. doi:10.1029/
94WR00871
View publication stats
Fredlund DG, Xing A, Fredlund MD, Barbour SL (1996) The relationship of the unsatu-
rated soil shear strength to the soil-water characteristic curve. Can Geotech J 33:440–
448. doi:10.1139/t96-065
Iverson RM (2000) Landslide triggering by rain infiltration. Water Resour Res 36:1897–
1910. doi:10.1029/2000WR900090
Iverson RM, Reid ME, Iverson NR, LaHusen RG, Logan M, Mann JE, Brien DL (2000) Acute
sensitivity of landslide rates to initial soil porosity. Science 290:513–516. doi:10.1126/
science.290.5491.513
Lourenço SDN, Sassa K, Fukuoka H (2006) Failure process and hydrologic response of a
two layer physical model: implications for rainfall-induced landslides. Geomorphology
73:115–130. doi:10.1016/j.geomorph.2005.06.004
Mari M (2000) Un metodo per la valutazione del grado di stabilità di pendii a rischio di
scivolamenti superficiali. Degree (MSc) Thesis, University of Parma
Montgomery DR, Dietrich WE (1994) A physically based model for the topographic
control on shallow landsliding. Water Resour Res 30:1153–1171. doi:10.1029/
93WR02979
Montrasio L (2000) Stability analysis of soil slip. In: Brebbia CA (ed) Proceedings of
International Conference BRisk 2000^. Wit Press, Southampton, pp 357–366
Montrasio L, Valentino R (2003) Experimental analysis on factors triggering soil slip. In:
Picarelli L (ed) Proceedings of international conference on fast slope movements:
prediction and prevention for risk mitigation. Patron, Bologna (Italy): 371–378
Montrasio L, Valentino R (2007) Experimental analysis and modelling of shallow
landslides. Landslides 4:291–296. doi:10.1007/s10346-007-0082-3
Montrasio L, Valentino R (2008) A model for triggering mechanisms of shallow land-
slides. Nat Hazards Earth Syst Sci 8:1149–1159. doi:10.5194/nhess-8-1149-2008
Montrasio L, Valentino R, Losi GL (2011) Towards a real-time susceptibility assessment of
rainfall-induced shallow landslides on a regional scale. Nat Hazards Earth Syst Sci
11:1927–1947. doi:10.5194/nhess-11-1927-2011
Montrasio L, Valentino R, Losi GL (2012) Shallow landslides triggered by rainfalls:
modeling of some case histories in the Reggiano Apennine (Emilia Romagna Region,
Northern Italy). Nat Hazards 60:1231–1254. doi:10.1007/s11069-011-9906-5
Montrasio L, Valentino R, Terrone A (2014) Application of the SLIP model. Procedia Earth
Planetary Sci 9:206–213. doi:10.1016/j.proeps.2014.06.023
Olivares L, Damiano E (2007) Post-failure mechanics of landslides: laboratory investiga-
tion of flowslides in pyroclastic soils. J Geotech Geoenviron 133:51–62. doi:10.1061/
(ASCE)1090-0241(2007)133:1(51)
Schilirò L, Esposito C, Scarascia Mugnozza G (2015) Evaluation of shallow landslide
triggering scenarios through a physically-based approach: an example of application
in the southern Messina area (north-eastern Sicily, Italy). Nat Hazards Earth Syst Sci
15:2091–2109. doi:10.5194/nhess-15-2091-2015
Silva D (2000) Analisi sperimentale del comportamento di terreni stratificati in pendio.
Degree (MSc) Thesis, University of Parma
Simoni S, Zanotti F, Bertoldi G, Rigon R (2008) Modelling the probability of occurrence of
shallow landslides and channelized debris flows using GEOtop-FS. Hydrol Process
22:532–545. doi:10.1002/hyp.6886
Soeters R, Van Westen CJ (1996) Slope instability recognition, analysis and zonation. In:
Turner AK, Schuster RL (eds) Landslides investigation and mitigation. Transportation
Research, US National Research Council, Washington, DC Board Special Report 247:
129-177
Wang G, Sassa K (2001) Factors affecting rainfall-induced flowslides in laboratory flume
tests. Geotechnique 51:587–599. doi:10.1680/geot.2001.51.7.587
Wang G, Sassa K (2003) Pore-pressure generation and movement of rainfall-induced
landslides: effects of grain size and fine-particle content. Eng Geol 69:109–125.
doi:10.1016/S0013-7952(02)00268-5
Zhang H, Zhang WW (2009) Rainfall infiltration in cracked soil and its effect on slope stability
analysis. In: Buzzi O, Fityus S, Sheng D (eds) Unsaturated soils: Proceedings of the 4th Asia
Pacific Conference on Unsaturated Soils. CRC Press, Newcastle, pp 313–318
L. Montrasio : A. Terrone
Department of Civil Environmental Engineering and Architecture,
University of Parma,
Parco Area delle Scienze, 181/A, 43100, Parma, Italy
L. Schilirò ())
Department of Earth Sciences,
BSapienza^ University of Rome,
Piazzale Aldo Moro 5, 00185, Rome, Italy
e-mail: luca.schiliro@uniroma1.it
Landslides