You are on page 1of 62

C

CEEN
NTTR RO
OE EU
URROO--M
MEEDDIITTE
ERRRRAANNEEOO
P E R I C A MB I A ME NT I C
PER I CAMBIAMENTI CLIMATICII LI M AT I C

ISC Impact on Soil and Coasts

Regional analysis of rainfall-induced


Technical Reports

landslides.
The case of Camaldoli hill, Naples:
test case nr.1 - October, 2004;
test case nr.2 - September, 2005

Luca Comegna, PhD


Centro euro-Mediterraneo per i Cambiamenti Climatici, CMCC

Centro Euro-Mediterraneo
per i Cambiamenti Climatici
www.cmcc.it March 2008 TR
Regional analysis of rainfall-induced landslides.
The case of Camaldoli hill, Naples:
test case nr.1 - October, 2004;
test case nr.2 - September, 2005

Summary
This report contains the results of some regional numerical analyses aimed to
simulate two rainfall-induced landslides events, recorded on October 13th, 2004 and
on September 17th, 2005, which involved the shallow pyroclastic deposits of the
south-eastern side of a hill, known as Camaldoli hill, set inside the western sector the
urban district of Naples.
The analyses have been performed thanks to the deterministic approach used by the
Transient Rainfall Infiltration and Grid-based Regional Slope-stability (TRIGRS)
program (Baum et al., 2002), a GIS-based code which allows to assess, over broad
areas, the timing and the location of shallow landslides caused by rainfall events.
Particular attention has been dedicated to the influence of the input parameters on the
results furnished by the code in response to the rainfall events.

Keywords: rainfall-induced landslide, regional slope stability analysis, numerical


simulation.

JEL Classification:

Address for correspondence:


Luca Comegna
PhD in Geotechnical Engineering
Seconda Universit degli studi di Napoli
Department of Civil Engineering
Via Roma, 29
81031 Aversa (CE)
E-mail: luca.comegna@unina2.it

2
CONTENTS

1. Foreword.4

2. Application of TRIGRS model6


2.1. Theoretical aspects7
2.2. Input data..10
2.2.1 Topographic slope.10
2.2.2 Soil thickness.15
2.2.3 Mechanical and hydraulic properties..15
2.2.4 Time-varying rainfall intensity..22
2.2.5 Initial hydraulic conditions24
2.3 TopoIndex Inizialization File24
2.4 TRIGRS Inizialization File26

3. Results of the analyses.31


3.1 Test case nr. 1..32
3.1.1 Hourly results.39
3.2 Test case nr. 2..48
3.2.1 Hourly results.52

4. Conclusions.60

5. References61

3
1. Foreword

Some numerical analyses have been performed in order to simulate the triggering of
rainfall-induced landslides that involved the shallow pyroclastic deposits of an area set
inside the western sector of the urban district of Naples, known as Camaldoli hill. The
chosen test cases are two and occurred on October 13th, 2004 and on September 17th,
2005. On October 13th, 2004, two slides, reported as A and B in Figure 1.1 were
triggered along the south-eastern side of the Camaldoli hill. They showed a rather
shallow depth (comprised between 0.5 m and 2.0 m), an average wideness equal to 10
m and an average length equal to 85 m. While the slide A started inside the pyroclastic
deposits set upslope the cliffs outcropping at the foot of the local Hermitage,
constituted by the Neapolitan Yellow Tuff, and then involved those set downslope, the
slide B originated inside the less steep pyroclastic deposits set downslope the tuff. The
landslide recorded on September 17th, 2005, was instead characterized by the
activation of three shallow slides, reported as C, D and E in Figure 1.2, that were
triggered upslope the tuff.
The analyses have been performed by the use of the Transient Rainfall Infiltration and
Grid-based Regional Slope-stability (TRIGRS) model (Baum et al., 2002). The code
couples an infinite-slope stability analysis with a one-dimensional analytical solution for
transient pore pressure response to steady state and transient rainfall infiltration,
allowing a regional slope stability evaluation in a Geographic Information System
framework. Due to its characteristics, TRIGRS can be used to investigate both the
timing and location of shallow landslides in response to rainfall over broad areas.
The input data needed by the code have been derived by the Technical Report
furnished by A.M.R.A S.c.a.r.l., which has permitted to combine meteorological data
with information about the geological and geomorphologic setting and about the
geotechnical properties of the soils.
The following paragraphs will first provide a sketch of the main theoretical aspects of
TRIGRS and an illustration of the input data necessary to run the program. Finally, the
results of the analyses coming from the application to the test cases will be described.

4
A

a)

b)
Figure 1.1. Landslides occurred on October 13th, 2004

5
D E
C

Figure 1.2. Landslides occurred on September 17th, 2005

2. Application of TRIGRS model

TRIGRS (Transient Rainfall Infiltration and Grid-based Regional Slope-stability model)


is a Fortran program able to compute transient pore pressures changes due to rainfall
infiltration and consequent changes in the factor of safety. It represents a regionally
based analytic method useful to individuate as the timing as the location of shallow
landslides in response to rainfall over wide regions: this is made possible through slope
stability calculations, performed by the recourse to infinite slope model, which are
coupled with simple hydrological models. Its worth noting that the infinite slope
hypothesis is strictly applicable only if the thickness of the landslide is small with
respect to its length.
Like other GIS-based models, the program operates on a map area that is subdivided
into a 3D-cells grid: each cell represents a homogenous soil column characterized by
own thickness, topographic parameters (elevation and slope), hydraulic properties
(permeability and diffusivity) and strength parameters (cohesion and friction angle).
The meteorological boundary conditions are assigned at the interface soil/atmosphere.
The program also contains an optional routing scheme that allows to the not infiltrated
water to flow toward downslope cells, respecting the mass balance between rainfall
input, infiltration, and runoff over the entire map area. To this aim, TRIGRS is furnished
with the utility software TopoIndex (Topographic Index), which prepares a group of

6
output data files for the runoff-routing calculations. These files contain a list of cell
numbers ranked in order from the highest to the lowest elevations, and a grid of cell
numbers indicating the neighboring cell on the steepest downslope path.
All the data must be saved to ASCII text files, that can be imported by GIS software for
display or further analysis. In order to run the analyses, TRIGRS and TopoIndex need
the compilation of appropriate initialization files, containing a list of all the input file
names and constants.

2.1 Theoretical aspects

The infiltration model implemented in TRIGRS is based on the extended formulation,


applied by Savage et al. (2003) to the original Iversons solution (2000) of the Richards
equation. In fact, while Iversons solution is valid for an infinitely deep impervious basal
boundary and the surface condition of constant flux for a specified time and zero flux
thereafter, Savage et al. (2003) generalize it the for the case of a time-varying
sequence of surface fluxes, characterized by variable intensities and durations, and a
layer of finite thickness.
Assuming that:
- the hillslope is saturated or tension-saturated;
- the infiltration is vertical (one-dimensional);
- is the constant slope angle;
- KZ is the vertical saturated permeability of the soil;
- is the pressure head;
- is the volumetric water content;
d
- C0 = is the specific moisture capacity of the soil;
d
- D0 = KZ/C0 is the hydraulic diffusivity of the soil

the Richards equation can be written as

2
= D1 [2.1]
t Z 2

where:

- t is the time;
z
- Z= is the vertical coordinate (Fig. 2.1);
cos
- D1 = D0 cos2 .


zcos

xsen
xsen

Figure 2.1.Coordinates system used in TRIGRS (Baum et al., 2002)

The model computes the pore pressure response in the Z direction resulting from a
time-varying rainfall input on the ground surface. In fact, once assigned a time-varying
specified flux boundary condition at the ground surface and an impermeable basal
boundary at a finite depth dLZ, the solution of equation [2.1], implemented in TRIGRS
is given by

( Z , t ) = [ Z d ] +
[2.2]
N
I nZ 1
(2m 1)d LZ (d LZ Z )
(2m 1)d LZ + (d LZ Z )
+ 2 H (t t n ) D1 (t t n ) 2 ierfc + ierfc
m =1 1
1

n =1 KZ
2[D1 (t t n )]2 2[D1 (t t n )]2

N
I nZ 1
(2m 1)d LZ (d LZ Z )
(2m 1)d LZ + (d LZ Z )
2 H (t t n +1 ) D1 (t t n +1 ) 2 ierfc + ierfc
m =1 1
1

n =1 KZ
2[D1 (t t n +1 )]2 2[D1 (t t n +1 )]2

where:
- d is the depth of the steady-state water table measured in the Z direction;
- = cos;
I
- = cos - Z is the long-term (steady-state) surface flux in the Z direction;
K Z LT
- KZ is the saturated permeability of the soil in the Z direction;
- InZ is the surface flux in correspondence of the nth time interval tn;
- N is the total number of time intervals;
- H (t - tn) is the Heavyside step function, equal to 0 if t tn and equal to 1 if t > tn.;
- ierfc( ) =
1
( )
exp 2 erfc( ) is the first integral of the complementary error

function (Carslaw e Jaeger, 1959).

TRIGRS also imposes to the [2.2] the physical limit

8
(Z; t) Z [2.3]

that doesnt allow that the calculated pressure head exceeds the value which would
result with the water table set at the ground surface and subject to the long-term
hydraulic gradient.
Its worth noting that if the impermeable basal boundary is set at an infinite depth (dLZ
) and the surface flux is applied for one single time interval, the equation [2.2]
coincides with the solution proposed by Iverson for the equation [2.1].

Once the pressure head (Z; t) is known, slope stability of each cell is calculated by
TRIGRS using an infinite-slope model, which is strictly applicable only if the thickness
of the landslide is small with respect to its length and width. The factor of safety FS(Z;t)
is so defined as the ratio of the resisting shear stress lim and driving shear stress
acting on the considered plane. Its so calculated by the code at a depth Z

FS ( Z ; t ) = =
[ ]
lim ( Z ; t ) c'+ Z cos ( Z ; t ) w tan ' tan ' c' ( Z ; t ) w tan '
2
= + [2.4]
Z sen cos tan Z sen cos

where:
- c and (effective cohesion and friction angle) are the strength parameters of the
soil for effective stresses;
- is the unit weight of soil;
- w is the unit weight of water.

Fredlund (1979) has furnished the following formulation for the shear strength envelope
of unsaturated soils

lim = [c'+(u a u w ) tg '] + ( u a ) tg ' [2.5]

where tlim is the shear strength; c and are the shear strength parameters for
saturated conditions; (ua-uw) is the matric suction, is a coefficient function of various
factors (confining total stress, suction, degree of saturation, void ratio, etc.) and - ua
is the net normal stress. The expression of tlim reported in equation [2.4] would
coincide with the one furnished by Fredlund for unsaturated soils if we assumed that
the coefficient was equal to 1. Because of the absence of any experimental data that
could confirm this hypothesis, we have decided to dont use the factor of safety
automatically calculated by TRIGRS through the [2.4], but to calculate it separately by
the use of the following expression

lim ( Z ; t ) c(Z ; t ) + [ (Z ) u a ] tan '


FS ( Z ; t ) = = [2.6]
Z sen cos

where:

9
- c(Z;t) is the intercept of cohesion, function of the matric suction ua-uw(Z;t) according
to the equation proposed by Vanapalli et al. (1996)

c = c'+(u a u w ) k tg ' [2.7]

where c and are the strength parameters of the soil for effective stresses,
wr
= w is the relative volumetric water content and k is a fitting parameter;
ws wr

- matric suction ua-uw(Z;t) = -(Z;t)gw;


- ua = 0 is the relative air pressure;
- s(Z) = gZcos2a is the total normal stress, acting at the depth Z in the direction
normal to the slope.

Due to the dependence of the intercept of cohesion by time, the factor of safety is a
time-function.

2.2. Input data

The input data to assign to each cell are those required by every GIS-based landslide
model:

a) topographic slope;
b) soil thickness;
c) mechanical and hydraulic properties;
d) time-varying rainfall intensity;
e) initial hydraulic conditions.

Its worth noting that in most cases its not so easy to obtain detailed information for
wide areas and frequently it is even impossible to acquire the complete input data,
especially those regarding soil properties and thickness of soil.

2.2.1 Topographic slope

Topographic slope has been derived from a Digital Elevation Model (DEM), created
using the available 1:5.000 scaled topography map of the Camaldoli hill realized by the
Neapolitan Provincial administration. The area of the studied zone is about 1.60 Km2.
The grid function contained in ArcView GIS has been used in order to create the
elevation grid file dem.asc (Fig. 2.2). The dem.asc file presents 191 columns and 206
rows and each cell is 10 m wide. The obtained elevations range from 125 m and 457
m.

10
Slope gradients for each cell of the DEM have been derived from dem.asc and saved
in the slope.asc grid file (Fig. 2.3). The calculated slope angles range from 0 to
68.084: the steepest cells coincide with the outcropping cliffs.
In order to implement runoff-routing calculations, TRIGRS needs the knowledge of the
direction of the steepest confining downslope cells in correspondence of each cell. This
operation can be obtained through ArcView Gis deriving from the dem.asc grid file the
directions.asc grid file (Fig. 2.4), which assigns to each cell an integer number
representative of the flow direction toward one of its eight neighbors. According to the
ArcView convention, the values for each direction from the central cell are:

32 64 128
16 1
8 4 2

For example, if the direction of the steepest downslope cell was to the right of the
current processing cell, its flow direction would be coded as 1. This method of deriving
flow direction from a DEM is presented in Jenson and Domingue (1988).

11
Figure 2.2. Display of the dem.asc grid file

12
Figure 2.3. Display of the slope.asc grid file

13
Figure 2.4. Display of the directions.asc grid file

14
2.2.2 Soil thickness

In order to assign to each cell the depth of bedrock (which coincides with the
impervious boundary depth), a simple linear relation between topographic slope and
soil thickness has been established (Fig. 2.5). This relation has been derived imposing
the value 25 m to the maximum measured soil thickness (regarding the pyroclastic
deposits set upslope the Neapolitan Yellow Tuff) and, at the same time, associating a
thickness equal to zero to the maximum slope angle derived in slope.asc file (max =
68.084). These data are contained in the zmax.asc grid file (Fig. 2.6).

30.00
y = -0.37x + 25.00
25.00
soil thickness [m]

20.00

15.00

10.00

5.00

0.00
0 10 20 30 40 50 60 70 80
slope angle []

Figure 2.5. Linear soil thickness model

2.2.3 Mechanical and hydraulic properties

Mechanical and hydraulic properties have been assigned to each cell of the grid,
through the creation of other four grid files, k.asc, diffus.asc, c.asc and phi.asc, which
respectively contain information about permeability, hydraulic diffusivity, cohesion and
friction angle. To this aim, the available geological map, furnished by A.M.R.A. and
reported in Figure 2.7, has been simplified. In particular, three classes have been
considered (Fig. 2.8):
1) loose pyroclastic soil, representative of every Intracaldera Phlegraean Deposits that
were produced by the different volcanic activities;
2) artificial deposit;
3) lithified deposit, representative of Neapolitan Yellow Tuff, Breccia Museo, Piperno
and Torre Franco Tuff.

The studied landslide events involved only loose pyroclastic soils. Their assigned
properties have been derived from the available laboratory results. In particular,
permeability kw has been assigned as a function of matric suction ua-uw (Fig. 2.9a),
using the formulation proposed by Mualem (1976)

15
k ws (u a u w ) 1
kw = [2.8]
k ws [ (u a u w )]( 2+5 / 2 ) (u a u w ) > 1

where kws is the saturated permeability, a is the reciprocal of the air-entry suction (ua-
uw)e and l is a dimensionless parameter representing the slope of the characteristic
curve. Table 2.1 contains the parameters that have given the best fitting of
experimental data.

Table 2.1. Hydraulic properties of Intracaldera Phlegraean pyroclastic Deposits


kws ws wr (ua-uw)e
[m/s] [kPa-1] kPa
5.4010-5 0.45 0.00 0.4019 0.1053 9.5

Also the hydraulic diffusivity D (Fig. 2.9b) has been treated as a function of suction,
combining the equation [2.8] with the correlation representative of the characteristic
curve, proposed by Brooks and Corey (1964)

ws (u a u w ) 1
w = [2.9]
wr + ws [ (u a u w )] (u a u w ) > 1

where w is the volumetric water content (ratio of water volume to the total soil volume),
ws is the saturated volumetric water content and wr is the residual volumetric water
content.

The considered shear strength envelope

lim = c + ( u a ) tg ' [2.10]

is characterized by the friction angle = 35 and by an intercept of cohesion c (also


known as apparent cohesion), which is a function of the matric suction according to the
relation

c = (u a u w ) 2 tg ' [2.11]

obtained substituting c = 0 and k=2 in the equation [2.7].

Due to the absence of experimental data regarding the artificial deposits, we have
assigned to them the same properties of the pyroclastic soils.
Being the lithified deposits strongly less permeable and more resistant than pyroclastic
soils, we have assigned to them permeability kw = 10-12 m/s, cohesion c = 1.50 MPa
and friction angle = 24. In particular, the strength parameters have been derived
from the results of some laboratory tests performed on Neapolitan Yellow Tuff and
illustrated by Croce and Pellegrino (1967).
All the adopted properties are shown in Table 2.2.

16
Figure 2.6. Display of the zmax.asc grid file

17
N

anthropic deposits
mine material

Intracaldera Phlegraean pyroclastic Deposits ( < 10 ky B.P.)


Intracaldera Phlegraean pyroclastic Deposits ( < 12 ky B.P.)
lithified Neapolitan Yellow Tuff (12 ky B.P.)
Intracaldera Phlegraean pyroclastic Deposits ( > 12 ky B.P.)
Breccia Museo (39 ky B.P.)

Piperno (39 ky B.P.)

Torre Franco tuff ( > 39 ky B.P.)

Figure 2.7. Geological map of the Camaldoli hill (modified from Di Crescenzo et al., 2007)

18
Figure 2.8. Display of the property classes

19
1.00E-03

1.00E-04

1.00E-05

1.00E-06
kw [m/s]

1.00E-07
1.00E-08

1.00E-09

1.00E-10

1.00E-11
0.1 1 10 100 1000

ua-uw [kPa]
a)

1.00E+00

1.00E-01

1.00E-02
D [m 2/s]

1.00E-03

1.00E-04

1.00E-05

1.00E-06

1.00E-07
0.1 1 10 100 1000

ua-uw [kPa]
b)
ua-uw kw D
kPa m/s m2/s
0 5.40E-05 5.88E-03
9.5 5.40E-05 5.88E-03
9.5 5.40E-05 4.87E-04
10 4.63E-05 4.32E-04
20 5.77E-06 1.15E-04
30 1.71E-06 5.51E-05
40 7.19E-07 3.32E-05
50 3.68E-07 2.25E-05
60 2.13E-07 1.65E-05
70 1.34E-07 1.27E-05
80 8.95E-08 1.01E-05
90 6.28E-08 8.32E-06
100 4.58E-08 6.97E-06

Figure 2.9. Permeability (a) and diffusivity (b) functions assigned to loose pyroclastic soils

20
Table Table Soil
2.2.3.1. Soilproperties
properties
hydraulic properties strength properties
unit weight, g kws ws wr (ua-uw)e permeability, K intercept of cohesion, c friction angle, '
3 -1
[kN/m ] [m/s] [kPa ] [kPa] [m/s] [kPa] []

21
k ws if (ua uw) 1
-5
loose pyroclastic soil 13 5.4010 0.45 0.00 0.4019 0.1053 9.5 35
( 2+ 5 / 2 ) (ua uw ) > 1
k ws [ (u a u w )] if

-12
lithified deposit 25 - - - - - - 10 1500 24
2.2.4 Time-varying rainfall intensity

The characteristics of the rain events which triggered the landslides events, have been
derived from the hourly rainfall measurements made by a pluviometer installed by the
Civil Protection of the Campania Region.
Taking into account the available hourly rainfall, we have imposed as boundary
condition a rainfall history constituted by seven periods, which are characterized by an
hourly duration and the following different intensities (Figs. 2.10 and 2.11):

Test case nr. 1: October, 2004


1st period of the precipitation intensity I1 = 0.6 mm/h (1.6710-7 m/s), duration t1 =
1h;
2nd period of the precipitation intensity I2 = 6.8 mm/h (1.8910-6 m/s), duration t2 =
1h;
3rd period of the precipitation intensity I3 = 19.2 mm/h (5.3310-6 m/s), duration t3 =
1h;
4th period of the precipitation intensity I4 = 4.4 mm/h (1.2210-6 m/s), duration t4 =
1h;
5th period of the precipitation intensity I5 = 10.8 mm/h (3.0010-6 m/s), duration t5 =
1h;
6th period of the precipitation intensity I6 = 4.0 mm/h (1.1110-6 m/s), duration t6 =
1h;
7th period of the precipitation intensity I7 = 1.8 mm/h (5.0010-7 m/s), duration t7 =
1h.

45
13/10/2004
40

35

30
I [mm/h]

25

20

15

10

0
0 1 2 3 4 5 6 7
time [h]

22
Figure 2.10. Test case nr.1 (13/10/2004): assigned rainfall history

Test case nr. 2: September, 2005


1st period of the precipitation intensity I1 = 0.4 mm/h (1.1110-7 m/s), duration t1 =
1h;
2nd period of the precipitation intensity I2 = 8.0 mm/h (2.2210-6 m/s), duration t2 =
1h;
3rd period of the precipitation intensity I3 = 44.2 mm/h (1.2310-5 m/s), duration t3 =
1h;
4th period of the precipitation intensity I4 = 4.0 mm/h (1.1110-6 m/s), duration t4 =
1h;
5th period of the precipitation intensity I5 = 2.0 mm/h (5.5610-7 m/s), duration t5 =
1h;
6th period of the precipitation intensity I6 = 0.2 mm/h (5.5610-8 m/s), duration t6 =
1h;
7th period of the precipitation intensity I7 = 0.4 mm/h (1.1110-7 m/s), duration t7 =
1h.

45

40 17-18/09/2005
35

30
I [mm/h]

25

20

15

10

0
0 1 2 3 4 5 6 7
time [h]

Figure 2.11. Test case nr.2 (17/09/2005): assigned rainfall history

2.2.5 Initial hydraulic conditions

23
Model results are very sensitive to the assigned initial depth of the water table, which
has been assigned to each cell through the depthwt.asc grid file.
For an infinite slope model with an inclination equal to , crossed by a water flow
parallel to the ground surface and with a water table set at the depth d, the steady-state
pressure head is given by

= (Z d ) cos 2 [2.12]

Being assumed a tension-saturated initial condition by the code, negative pressure


heads initially develop according to a linear law at depths shallower than the water
table. In particular, the initial suction value at the ground surface (ua-uw)GS,0 is
determined by the assigned d, through the relation

(u a u w )GS .,0 = w d cos 2 [2.13]

We have decided to assign the initial depth of the water table to each cell as a function
of the hypothesized suction at ground surface (ua-uw)GS,0. So doing, d is univocally
related to the (ua-uw)GS,0

(u a u w )GS .,0
d= [2.14]
w cos 2

Because different hydraulic and mechanical properties of the pyroclastic soils


correspond to any assigned suction (see Tab. 2.2), this choice may strongly influence
the response.

2.3 TopoIndex Inizialization File

After having prepared dem.asc and directions.asc files, it has been possible to
compilate the inizialization text file of the utility software TopoIndex (Topographic
Index) for the runoff-routing calculations, named topoindex.ini (Fig. 2.12). This file,
which is the same for the analyses related to both the examined test cases, is
characterized by a sequence of comment lines, which systematically precede the
corresponding command lines.
The first command line contain the name assigned to the project.
The second command line reports the number of rows and columns of every grid files
and the conventional numbering scheme used in the directions.asc file (the convention
adopted by ESRI coincides with that reported in par. 2.2.1).

Name of project (up to 255 characters)


Camaldoli TopoIndex analysis
Rows, Columns, flow-direction numbering scheme (ESRI=1, TopoIndex=2)
206, 191, 1
Exponent, Number of iterations
-1, 21
Name of elevation grid file

24
dem.asc
Name of direction grid
directions.asc
Save listing of D8 downslope receptor cells? Enter T (.true.) or F (.false.)
T
Save grid of D8 downslope receptor cells? Enter T (.true.) or F (.false.)
T
Save cell index number grid ? Enter T (.true.) or F (.false.)
T
Save list of cell number and corresponding index number? Enter T (.true.) or F (.false.)
T
Save flow-direction grid remapped from ESRI to TopoIndex? Enter T (.true.) or F (.false.)
T
Name of folder to store output?

ID code for output files? (8 characters or less)


cam
Figure 2.12. The TopoIndex Inizialization File, topoindex.ini

The exponent contained in the third command line indicates the typology of pattern
chosen for the runoff-routing calculations. In our case, using -1 we impose the not
infiltrated water to flow only toward one or two neighboring downslope cells that
straddle the steepest downslope path. The flow is so transferred to the downslope cells
according to weighting factors proportional to the respective slopes.

Because TopoIndex uses an iterative process to determine the order in which runoff
calculations should proceed, the maximum number of iterations needed to complete
the indexing process varies with the size of the grid. According to the advice furnished
by the manual, the maximum number of iterations reported in the third command line,
set equal to 21, has been derived by the formula

rows + columns
iterations 1 +
20

The successive two lines contain the names of the input grid files, where are reported
the elevation (dem.asc) and the direction of flow (directions.asc).
In the final group of lines, the user must specify which optional output files will be saved
and an identifying code for them (cam in our case). The five optional files regard:
- a list of downslope receptor cells in flow direction (TIdsneiList_cam.txt);
- a grid of downslope receptor cells (TIdscelGrid_cam.asc);
- a grid of cell index numbers that specify computation order for runoff routing and
corresponding computation order number (TIcelindxGrid_cam.asc);
- a list of cell numbers and corresponding computation order number
(TIcelindxList_cam.txt);
- a flow-direction grid in terms of the TopoIndex numbering scheme
(TIflodirGrid_cam.asc).

Moreover, TopoIndex will automatically provide other three output files:


- a log file, which contains information about program operation and repeats the data
from the inizialization file (TopoIndexLog.txt);
- a cell by cell list of downslope cells for which non-zero weighting factors have been
computed (TIdscelList_cam.txt);

25
- a cell by cell list of weighting factors for downslope receptor cells
(TIwfactorList_cam.txt).
Once the trigrs.ini text file has been compiled, TopoIndex program has been launched,
providing all the output files.

2.4 TRIGRS Inizialization File

After having launched TopoIndex program, the inizialization text file of TRIGRS,
trigrs.ini, has been prepared (Figs. 2.13 and 2.14). This file has a structure very similar
to the topoindex.ini text file, being characterized by alternating comment and command
lines. In particular the command lines can be divided into three group files:
- numeric data: the first 6 lines report the project title and some numeric data;
- input files: the next 14 lines contain the names of the input files;
- output files: the final lines specify which output files will be saved.
In the follow, referring to the test case nr. 1, the various command lines will be
described.

Numeric data
These lines contain data necessary to determine array sizes, rainfall history, initial
conditions and physical properties. The user can assign constant values over the entire
grid to the various parameters (except slope), specifying positive values for them. On
the other hand, for spatially variable parameters, the user has to set a negative value
for them: so doing, the user prompts the program to read the corresponding input grid
file.
The 1st command line reports the title of the project.
The 2nd command line contains some data, useful for runoff analysis, which are
furnished by the TopoIndex output file TopoIndexLog.txt:
- imax, number of grid cells excluding no-data cells;
- row, total number of rows;
- col, number of columns;
- nwf, length of arrays to store flow routing locations and weighting factors.
Do to the fact that TRIGRS computes for each cell the pore pressure at equally spaced
depths between a minimum and a maximum depth assigned by the user, we have
specified in the 3rd command line:
- nzs, number of vertical increments assigned to each cell for infiltration computing;
- mmax, number of terms in series solution for impermeable base at finite depth (see
equation [2.2]);
- nper, number of periods of different rainfall intensities during the considered rain
event;
- zmin [m], shallowest depth at which TRIGRS has to calculate pore pressures (due
to computational problems, it must be greater than zero, so we have imposed zmin
= 1mm);
- uww [kN/m3], unit weight of water;
- t [s], elapsed time since the beginning of rain event.
For example, assigning t = 25200 seconds, we impose to TRIGRS to compute the pore
pressures at the end of the rain event (which lasts 7 hours).
The 4th line contains other 8 parameters:

26
- cc [kN/m2], cohesion;
- cphi [], angle of internal friction;
- czmax [m], maximum depth at which TRIGRS has to calculate pore pressures;
- cuws [kN/m3], total unit weight of soil;
- crizero [m/s], steady pre-rain event infiltration rate;
- cdep [m], initial steady depth of water table;
- cdif [m2/s], hydraulic diffusivity of soil;
- cks [m/s], permeability of soil.
Setting a positive a value, we have assigned a constant value only to crizero. Assigning
a negative value to the other parameters, we have imposed to these a spatially variable
value, that TRIGRS will be able to read through the corresponding input grid files.
The 5th line lists rainfall intensities (cri(1), cri(2), etc.), expressed in [m/s],
corresponding to the number of periods of distinct rainfall intensities nper.
Finally, the 6th line contains times (capt(1), capt(2), etc.), expressed in [s], which define
the rainfall periods listed in their chronological order.

Input files
The successive group of command lines, lists the names of the input files (already
prepared) in the order expected by TRIGRS.
The 7th command line contains the slope.asc grid file.
From the 8th line to the 14th line, the grid files regarding the spatially variable
parameters are reported:
- c.asc, cohesion grid file;
- phi.asc, angle of internal friction grid file;
- zmax.asc, maximum depth grid file;
- uws.asc, total unit weight grid file;
- depthwt.asc, initial steady depth of water table grid file;
- diffuse.asc, hydraulic diffusivity grid file;
- k.asc, permeability grid file.
Having assigned a spatially constant value to the pre-rain event initial infiltration, the
15th line is blank.
The lines from the 16th to the 19th refer to the TopoIndex output files necessary for
runoff calculations:
- TIdscelGrid_cam.asc, downslope receptor cells grid file;
- TIcelindxList_cam.txt, list of cell numbers and corresponding computation order
number;
- TIdscelList_cam.txt, cell by cell list of downslope cells for which non-zero weighting
factors have been computed;
- TIwfactorList_cam.txt, cell by cell list of weighting factors for downslope receptor
cells.
Finally, the 20th line regard the rainfall intensity grid files assigned to each period of the
rain event (ri1.asc, ri2.asc,.. ri7.asc).

Output files
The final group of lines regards the optional output files, which can be furnished by
TRIGRS. We have decided to save only the following files:

27
- TRrunoffPer1cam.asc, TRrunoffPer2cam.asc, etc., grid of runoff computed at the
end of each period of the rain event (23rd line);
- TRinfilratPer1cam.asc, TRinfilratPer2cam.asc, etc., grid of infiltration rate
computed at the end of each period of the rain event (29th line);
- TRlist_z_p_fs_cam.txt, detailed cell by cell listing of pressure head and factor of
safety at each depth increment (30th line).
As for TopoIndex, TRIGRS also furnishes a TrigrsLog.txt file, which gives information
about program operation, rewriting the data from the inizialization file.
As explained in the par. 2.1, we have decided to dont use the factor of safety
automatically calculated by TRIGRS through the [3.4], but to calculate it separately by
the use of the equation [3.5], in order to take into account the non-linearity between the
intercept of cohesion c and the matric suction ua-uw (eq. [2.11]). In particular, we have
combined the results for pressure head coming from the output file
TRlist_z_p_fs_cam.txt with the equation [2.7], which has permitted to calculate the
value of the intercept of cohesion at any desired depth and so to calculate also the
factor of safety by the use of the equation [2.6]. These operations have induced the
creation of another grid file, named fs.asc, which furnishes the factor of safety
calculated cell by cell at different depths. Being Camaldoli hill always characterized by
shallow landslides, with a maximum depth equal to 2 m, the factor of safety of each cell
has been calculated at the following depths z:

- z = zmax, if zmax 2m;


- z = 2m, if zmax > 2m.

Name of project (up to 255 characters)


Camaldoli_1 TRIGRS analysis
imax, row, col, nwf

28
15842, 206, 191, 26807
nzs, mmax, nper, zmin, uww, t
5, 20, 7, 0.001, 10, 25200
cc, cphi, czmax, cuws, crizero, cdep, cdif, cks
-5, -30, -3.5, -13, 1.0e-9, -20, -5.0e-4, -5.0e-5
cri(1), cri(2), ..., cri(nper)
1.67e-07, 1.89e-06, 5.33e-06, 1.22e-06, 3.00e-06, 1.11e-06, 5.00e-07
capt(1), capt(2), ..., capt(n), capt(n+1)
0, 3600, 7200, 10800, 14400, 18000, 21600, 25200
File name of slope angle grid (slofil)
slope.asc
File name of cohesion grid (cfil)
c.asc
File name of Phi-angle grid (phifil)
phi.asc
File name of depth grid (zfil)
zmax.asc
File name of total unit weight of soil grid (uwsfil)
uws.asc
File name of initial depth of water table grid (depfil)
depthwt.asc
File name of diffusivity grid (diffil)
diffus.asc
File name of hydraulic conductivity grid (ksfil)
k.asc
File name of initial infiltration rate grid (rizerofil)

File name of grid of D8 runoff receptor cell numbers (nxtfil)


TIdscelGrid_cam.asc
File name of list of defining runoff computation order (ndxfil)
TIcelindxList_cam.txt
File name of list of all runoff receptor cells (dscfil)
TIdscelList_cam.txt
File name of list of runoff weighting factors (wffil)
TIwfactorList_cam.txt
List of file name(s) of rainfall intensity for each period, (rifil())
ri1.asc
ri2.asc
ri3.asc
ri4.asc
ri5.asc
ri6.asc
ri7.asc
Folder where output grid files will be stored (folder)

Identification code to be added to names of output files (suffix)


t=7h
Save grid files of runoff? Enter T (.true.) or F (.false.)
T
Save grid of factor of safety at maximum depth, zmax? Enter T (.true.) or F (.false.)
F
Save grid of pore pressure at maximum depth, zmax? Enter T (.true.) or F (.false.)
F
Save grid of minimum factor of safety? Enter Enter T (.true.) or F (.false.)
F
Save grid of depth of minimum factor of safety? Enter Enter T (.true.) or F (.false.)
F
Save grid of pore pressure at depth of minimum factor of safety? Enter Enter T (.true.) or F
(.false.)
F
Save grid files of actual infiltration rate? Enter T (.true.) or F (.false.)
T
Save listing of pressure head and factor of safety ("flag")? (Enter -2 detailed, -1 normal, 0
none)
-2

Figure 2.13. The TRIGRS Inizialization File, trigrs.ini (test case nr.1)

29
Name of project (up to 255 characters)
Camaldoli_2 TRIGRS analysis
imax, row, col, nwf
15842, 206, 191, 26807
nzs, mmax, nper, zmin, uww, t
5, 20, 7, 0.001, 10, 25200
cc, cphi, czmax, cuws, crizero, cdep, cdif, cks
-5, -30, -3.5, -13, 1.0e-9, -20, -5.0e-4, -5.0e-5
cri(1), cri(2), ..., cri(nper)
1.11e-07, 2.22e-06, 1.23e-05, 1.11e-06, 5.56e-07, 5.56e-08, 1.11e-07
capt(1), capt(2), ..., capt(n), capt(n+1)
0, 3600, 7200, 10800, 14400, 18000, 21600, 25200
File name of slope angle grid (slofil)
slope.asc
File name of cohesion grid (cfil)
c.asc
File name of Phi-angle grid (phifil)
phi.asc
File name of depth grid (zfil)
zmax.asc
File name of total unit weight of soil grid (uwsfil)
uws.asc
File name of initial depth of water table grid (depfil)
depthwt.asc
File name of diffusivity grid (diffil)
diffus.asc
File name of hydraulic conductivity grid (ksfil)
k.asc
File name of initial infiltration rate grid (rizerofil)

File name of grid of D8 runoff receptor cell numbers (nxtfil)


TIdscelGrid_cam.asc
File name of list of defining runoff computation order (ndxfil)
TIcelindxList_cam.txt
File name of list of all runoff receptor cells (dscfil)
TIdscelList_cam.txt
File name of list of runoff weighting factors (wffil)
TIwfactorList_cam.txt
List of file name(s) of rainfall intensity for each period, (rifil())
ri1.asc
ri2.asc
ri3.asc
ri4.asc
ri5.asc
ri6.asc
ri7.asc
Folder where output grid files will be stored (folder)

Identification code to be added to names of output files (suffix)


t=7h
Save grid files of runoff? Enter T (.true.) or F (.false.)
T
Save grid of factor of safety at maximum depth, zmax? Enter T (.true.) or F (.false.)
F
Save grid of pore pressure at maximum depth, zmax? Enter T (.true.) or F (.false.)
F
Save grid of minimum factor of safety? Enter Enter T (.true.) or F (.false.)
F
Save grid of depth of minimum factor of safety? Enter Enter T (.true.) or F (.false.)
F
Save grid of pore pressure at depth of minimum factor of safety? Enter Enter T (.true.) or F
(.false.)
F
Save grid files of actual infiltration rate? Enter T (.true.) or F (.false.)
T
Save listing of pressure head and factor of safety ("flag")? (Enter -2 detailed, -1 normal, 0
none)
-2

Figure 2.14. The TRIGRS Inizialization File, trigrs.ini (test case nr.2)

30
3. Results of the analyses

The domain of the analyses is limited to the south-eastern side of the hill, where
landslides occurred on October 13th, 2004 and on September 17th, 2005.
First of all, we have analyzed the influence of the assigned initial suction at ground
surface (ua-uw)GS,0 on the minimum factor of safety FS0,min calculated at the end of the
inizialization phase (i.e. before the application of the rainfall history).
Assigning (ua-uw)GS,0 = 15 kPa, 831 cells present a factor of safety FS0 1 and FS0,min
is equal to 0.96.
On the other hand, for suctions equal or higher than 16 kPa, initial factor of safety FS0
is higher than 1.00 in correspondence of every cell, increasing with suctions according
to the trend reported in Fig. 3.1. Due to these observations, the analyses will not take
into account suctions at ground surface lower than 20 kPa as initial condition.

1.4

1.3

1.2
FS0,min

1.1

1.0

0.9
0 10 20 30 40 50 60 70 80 90 100
(ua-uw)GS,0 [kPa]

(ua-uw)0 FS0,min
[kPa]
15 0.96
16 1.01
20 1.05
30 1.12
40 1.17
50 1.21
60 1.24
70 1.27
80 1.29
90 1.31
100 1.33

Figure 3.1. Calculated initial minimum factor of safety FS0,min versus the initial suction imposed
at ground surface (ua-uw)GS,0

31
3.1 Test case nr. 1

In order to investigate the sensitivity of the model to the initial conditions and to the
hydraulic parameters, we have launched a number of simulations, characterized by the
same rainfall history (Fig. 2.10), but by different initial suctions (ua-uw)GS,0 imposed at
the ground surface and by different permeability kw and diffusivity D assigned to each
cell representative of the pyroclastic deposits. In particular, (ua-uw)GS,0 has been varied
between 20 and 100 kPa, while, according to the permeability and diffusivity functions
reported in Fig. 2.9, kw and D have been varied between the saturated values
(respectively 5.4010-5 m/s and 5.8810-3 m2/s) and the values coming from the
permeability and the diffusivity functions corresponding to a suction equal to 100 kPa
(respectively 4.5810-8 m/s and 6.9710-6 m2/s).
The results of these analyses are reported in Fig. 3.2, which shows the calculated
number N of cells that at the end of the rainfall history present a final factor of safety FS
lower than 1.0 (Fig. 3.2a), than 1.1 (Fig. 3.2b) and than 1.2 (Fig. 3.2c). As expected, we
have noted that:
- N increases with kw in correspondence of the same (ua-uw)GS,0,
- N decreases with (ua-uw)GS,0 in correspondence of the same kw,;
In particular, the reduction of N with suction is more abrupt for the permeabilities lower
than the saturated value.
Besides, the red line drawn in Figure 3.3 represents the envelope of the minimum
threshold values of permeability that should be assigned to the pyroclastic soil in order
to have at least one calculated cell characterized by a final factor of safety FS 1.0
(i.e. subjected to failure): these values clearly grow with (ua-uw)GS,0.
Considering this response given by the model, its evident that the hydraulic
parameters are able to highly influence the results. As already explained, the code
assumes an initial tension-saturated condition, which induces the suctions to decrease
with depth according to a decreasing linear law. Due to this aspect, the permeability is
not constant with depth (as required by the code), but grows with it. Insofar, the
permeability assigned to the cells should represent a sort of equivalent permeability of
a layer characterized by different permeabilities. Because the layer is crossed in
series by the infiltrating flow, we are able to easily assume that the equivalent
permeability is very next to the minimum value, that is the permeability at the ground
surface (which has the maximum suction). Again, its worth noting that the general
growth of the equivalent permeability caused to the system by the infiltration process
will be necessarily neglected (the code requires a permeability constant not only with
depth, but also with time).

Following these observations, we have decided to univocally link the permeability of


each cell to the initial suction imposed at the ground surface, through the experimental
permeability curve (Fig. 2.9.). In particular, if we overlap this curve to the calculated red
threshold line reported in Fig. 3.3, we note that the examined test case will not cause
any failure if we assigned an initial suction higher than 50 kPa, which becomes the
maximum attributable value in order to simulate the failure event. In the follow, we will
compare the different results corresponding to three examined cases (Tab.3.1),
obtained at the end of the assigned rainfall history: particular attention will be given to
the cells A and B, where landslides triggered on October 13th, 2004 (Fig. 3.4).

32
FS<=1
kw [m/s]
10000
5.40E-05
5.77E-06
1.71E-06
1000 7.19E-07
3.68E-07
2.13E-07
100 1.34E-07
N

10

1
20 30 40 50 60 70 80 90 100
(ua-uw )GS,0 [kPa]
a)
FS<=1.1

10000 5.40E-05
5.77E-06
1.71E-06
1000
7.19E-07
3.68E-07
100 2.13E-07
N

1.34E-07

10

1
20 30 40 50 60 70 80 90 100
(ua-uw )GS,0 [kPa]
b)
FS<=1.2

10000 5.40E-05
5.77E-06
1.71E-06
1000
7.19E-07
3.68E-07
100 2.13E-07
N

1.34E-07

10

1
20 30 40 50 60 70 80 90 100
(ua-uw )GS,0 [kPa]
c)
Figure 3.2. Calculated number of cells N with FS 1 (a), FS 1.1 (b) and FS 1.2 (c) at the end of
the precipitation versus the initial suction imposed at ground surface (ua-uw)GS,0 (for different
assigned permeability)

33
1.00E-04
Threshold (N=1)
Permeability function
1.00E-05
N1
kw [m/s]

1.00E-06
N=1

1.00E-07
N=0

1.00E-08
20 30 40 50 60 70 80 90 100
(ua-uw)GS,0 [kPa]

Figure 3.3. Calculated permeability threshold (red line) and experimental permeability function
(blue line) versus suctions

Table 3.1. Examined cases: different permeability kw and hydraulic diffusivity D correspond to
each initial suction imposed at ground surface (ua-uw)GS,0

(ua-uw)GS,0 kw D
Case
[kPa] [m/s] [m2/s]
1 50 3.68E-07 2.25E-05
2 40 7.19E-07 3.32E-05
3 30 1.71E-06 5.51E-05

A B
zmax [m] 0.647 4.810
[] 66.32 54.99

Figure 3.4. Landslides triggered on 13/10/2004

34
Case 1: (ua-uw)GS,0 = 50 kPa
The calculated minimum factor of safety FS0,min is 1.21 at the end of the inizialization
phase (i.e. before the application of the rainfall history measured on 13/10/2004). In
particular, the cells A and B respectively show FSA,t0 = 3.272, FSB,t0 = 1.224 (Tab. 3.2).
As reported in Fig. 3.5, after the 7 hours of the applied rainfall period, failure verifies
only in correspondence of cell A. Factor of safety of cell B FSB,t7 is equal to 1.216 (Tab.
3.2) and the factor of safety calculated in correspondence of 36 cells is comprised
between 1.0 and FSB,t7.

Table 3.2. Factor of safety FS calculated at the end of the initialization phase (t=0) and after the
assigned rainfall history (t=7h) in correspondence of the cells A and B, where landslides
triggered on October 13th, 2004.

(ua-uw)GS,0 t=0 t=7h


Case
[kPa] FSA,t0 FSB,t0 FSA,t7 FSB,t7
1 50 3.272 1.224 0.370 1.216
2 40 3.142 1.187 0.072 1.171
3 30 2.982 1.140 0.071 1.105

Figure 3.5. Results of case 1 at the end of the 7th hour: cell A (red cell) is the only cell which
has a factor of safety FSA,t7 lower than 1.00; 36 cells (yellow cells) present a factor of safety
comprised between 1.00 and the factor of safety FSB,t7 of cell B (equal to 1.216)

35
Case 2: (ua-uw)GS,0 = 40 kPa
The calculated minimum factor of safety FS0,min is 1.17 at the end of the inizialization
phase; the cells A and B respectively show FSA,t0 = 3.142, FSB,t0 = 1.187 (Tab. 3.2).
As reported in Fig. 3.6, after the 7 hours of the rainfall period, failure verifies only in
correspondence of cell A. Factor of safety of cell B FSB,t7 is equal to 1.171 (Tab. 3.2)
and 35 cells present a factor of safety comprised between 1.0 and FSB,t7.

Figure 3.6. Results of case 2 at the end of the 7th hour: the factor of safety FSA,t7 of cell A (red
cell) is lower than 1.00; 35 cells (yellow cells) present a factor of safety comprised between
1.00 and the factor of safety FSB,t7 of cell B (equal to 1.171)

Case 3: (ua-uw)GS,0 = 30 kPa


The calculated minimum factor of safety FS0,min is 1.12 at the end of the inizialization
phase. In particular, the cells A and B respectively show FSA,t0 = 2.982, FSB,t0 = 1.140
(Tab. 3.2).
As reported in Fig. 3.7, at the end of the rainfall period, failure verifies not only in
correspondence of cell A, but also of other 5 cells. Factor of safety of cell B FSB,t7 is
equal to 1.105 (Tab. 3.2) and 30 cells present a factor of safety FS,t7 comprised
between 1.0 and FSB,t7.

36
Figure 3.7. Results of case 3 at the end of the 7th hour: the factor of safety calculated at the end
of the rain event is lower than 1.00 in correspondence of cell A FSA,t7 and of other 5 cells (red
cells); 30 cells (yellow cells) present a factor of safety comprised between 1.00 and the factor
of safety FSB,t7 of cell B (equal to 1.105)

After a comparison among these three examined cases, we can make some
comments:
- all the three cases are able to forecast landslide A;
- no one case is able to forecast landslide B;
- case 1 and case 2 dont furnish false alarms;
- case 3 furnishes five false alarms.

We retain that case 2 is the most reliable case, because it assigns the minimum factor
of safety to cell B, without giving false alarms. In particular, the initial suction related to
that case, (ua-uw)GS,0 = 40 kPa, is also consistent with the values measured on October
13th, 2004 by a tensiometer installed in proximity of the Hermitage (Fig. 3.8) at different
depths next to the ground surface (30, 60 and 90 cm): according to these
measurements, the suctions range between about 40 and 55 kPa. A more detailed
description (hourly results) related to case 2 will be reported in the next paragraph.

Its evident that the susceptibility to failure of each cell would increase if we assigned
higher values to hydraulic conductivity. For example, maintaining the hypothesis (ua-
uw)GS,0 = 40 kPa, the cell B would attain failure after five hours from the beginning of the
rain event (time 21:00) imposing the saturated hydraulic properties kw,sat = 5.40E-05
m/s and Dsat = 5.88E-03 m/s: this result is unlikely, because 257 cells would present FS
1, producing 255 false alarms (Fig. 3.9).

37
T

(Minopoli, 2005)

Figure 3.8. Suctions measured at depths 30, 60 and 90 cm by a tensiometer (named T in the
figure) installed next to the Hermitage: suctions measured on 13/10/2004 range between about
40 kPa and 55 kPa (after Minopoli, 2005)

Figure 3.9. Factor of safety FS calculated after 5 hours (time 21:00) from the beginning of the
rain event and assigning saturated hydraulic properties: red cells have FS1

38
3.1.1 Hourly results

Factor of safety
As explained in the previous paragraph, we have decided to start the analyses with the
hypothesis that the initial suction at ground surface is equal to 40 kPa. According to this
choice, we have assigned to the loose pyroclastic soil a permeability kw = 7.19E-07 m/s
and a diffusivity D= 3.32E-05 m2/s (Tab. 3.1).
Table 3.3 reports the results at the end of each hourly step, starting from the beginning
of the imposed rain event, in terms of:
- number of cells N with FS 1;
- number of cells N with 1 < FS FSB,t7, where FSB,t7 = 1.171 is the factor of safety of
the cell B calculated at the end of the 7th hour;
- factor of safety of the cell A FSA;
- factor of safety of the cell B FSB.

According to these results (shown from Figure 3.10 to Figure 3.16), cell A maintains a
very high factor of safety during the first three hours (the ratio between FSA and FSB is
greater than 2). This is due to the great influence played by the cohesion on the factor
of safety at low depths (the depth z = 2m, where FSB is calculated, is about three times
greater than zmax,C, equal to 0.65 m). At the same time, its obvious that the reduction of
strength induced by the infiltration is more rapid just at the cells characterized by a
smaller thickness. The cell A attains the failure at the end of the 4th hour - time 20:00
(Fig. 3.13), that is 1 hour later than the time of rainy peak (I = 19.2 mm/h), which
verifies between 18:00 and 19:00, while FSB remains higher than 1.00. During the
entire period of the rain event, the cell A remains the only cell with FS<1. On the other
hand, its also interesting to note that the cell B attains its minimum value FSB = 1.171
at the end of the 7th hour - time 23:00 (Fig. 3.16), contemporarily to other 8 cells and
after that 28 cells already attained this value. Insofar, 28 cells present a susceptibility to
failure higher than cell B: these 28 cells are characterized by a higher slope and, as a
consequence of the linear relationship reported in Fig. 3.5, by a lower soil thickness.
Having assigned to these cells the same soil properties, this higher susceptibility is due
only to different geometrical characteristic.

Table 3.3. Results at the end of each hourly step of the precipitation event: number of cells N
with FS 1 and with 1 < FS FSB,t7 (FSB,t7 = 1.171) ; factor of safety in correspondence of cell
A (FSA) and of cell B (FSB)

N
t I (mm/h) FSA FSB
FS 1 1 < FS 1.171
1h (17:00) 0.6 0 8 3.105 1.187
2h (18:00) 6.8 0 12 2.922 1.186
3h (19:00) 19.2 0 17 2.637 1.184
4h (20:00) 4.4 1 20 0.962 1.181
5h (21:00) 10.8 1 23 0.072 1.177
6h (22:00) 4.0 1 27 0.072 1.174
7h (23:00) 1.8 1 35 0.072 1.171

39
Figure 3.10. Factor of safety FS calculated after 1 hour (time 17:00) from the beginning of the
rain event: 8 yellow cells are characterized by 1<FS1.171

Figure 3.11. Factor of safety FS calculated after 2 hours (time 18:00) from the beginning of the
rain event: 12 yellow cells are characterized by 1<FS1.171

40
Figure 3.12. Factor of safety FS calculated after 3 hours (time 19:00) from the beginning of the
rain event: 17 yellow cells are characterized by 1<FS1.171

Figure 3.13. Factor of safety FS calculated after 4 hours (time 20:00) from the beginning of the
rain event: 1 red cell attains a value lower than 1.00 (cell A), 20 yellow cells are characterized
by 1<FS1.171

41
Figure 3.14. Factor of safety FS calculated after 5 hours (time 21:00) from the beginning of the
rain event: 1 cell attains a value lower than 1.00 (cell A); 23 yellow cells are characterized by
1<FS1.171

Figure 3.15. Factor of safety FS calculated after 6 hours (time 22:00) from the beginning of the
rain event: 1 cell is characterized by FS<1 (cell A); 27 yellow cells are characterized by
1<FS1.171

42
Figure 3.16. Factor of safety FS calculated at the end of the rain event (time 23:00): 1 cell is
characterized by FS<1 (cell A); 35 yellow cells are characterized by 1<FS1.171 (cell B attains
the value 1.171)

Runoff
The code uses a simplified method for the calculation of the surface runoff from cells
where precipitation exceeds the infiltrability to the bordering downslope cells. In
particular, runoff is assumed to verify instantaneously from cell to cell and it doesnt
receive any contribution from subsurface flow. This assumptions allow to obtain a not
too reliable estimation of the flow. Nevertheless, the model could be used in order to
have useful information about the preferential paths of the flow, which can help, for
example, to forecast where the maximum downslope discharges have to be expected
in correspondence of the urbanized area.
Results are reported from Figure 3.17 to Figure 3.23. They are strictly linked to the
assigned rainfall history, showing, for example minimum and little diffuse values (about
10 mm/h) at the end of the 1st hour and maximum values (runoff overcomes 1000
mm/h at many cells) at the of the 3rd hour (Fig. 3.19), which respectively correspond to
the minimum and maximum intensities of the rain event.
As a general result, its interesting to note that the highest values are registered along
the natural channels, which bring the higher portion of the not infiltrated rain water from
upslope to the urban district set downslope.

43
Figure 3.17. Runoff calculated after 1 hour (time 17:00) from the beginning of the rain event

Figure 3.18. Runoff calculated after 2 hours (time 18:00) from the beginning of the rain event

44
Figure 3.19. Runoff calculated after 3 hours (time 19:00) from the beginning of the rain event

Figure 3.20. Runoff calculated after 4 hours (time 20:00) from the beginning of the rain event

45
Figure 3.21. Runoff calculated after 5 hours (time 21:00) from the beginning of the rain event

Figure 3.22. Runoff calculated after 6 hours (time 22:00) from the beginning of the rain event

46
Figure 3.23. Runoff calculated after 7 hours (time 23:00) from the beginning of the rain event

47
3.2 Test case nr. 2

Following the model proposed in the precedent paragraph, the permeability of each cell
has been linked to the initial suction imposed at the ground surface, through the
experimental permeability curve (Fig. 2.9). In the follow we will compare the different
results corresponding to three examined cases (Tab. 3.4), obtained at the end of the
assigned rainfall history (Fig. 2.11), with particular attention to the cells C, D and E,
where landslides triggered on September 17th, 2005 (Fig. 3.24).

C D E
zmax [m] 0.647 4.200 2.938
[] 66.32 56.64 60.08

D E
C

Figure 3.24. Landslides triggered on 17/09/2005

Table 3.4. Factor of safety FS calculated at the end of the initialization phase (t=0) and after the
assigned rainfall history (t=7h) in correspondence of the cells C, D and E, where landslides
triggered on September 17th, 2005.

(ua-uw)GS,0 t=0 t=7h


Case
[kPa] FSC,t0 FSD,t0 FSE,t0 FSC,t7 FSD,t7 FSE,t7
1 50 3.272 1.213 1.206 2.559 1.204 1.193
2 40 3.142 1.176 1.167 0.072 1.160 1.141
3 30 2.982 1.129 1.119 0.071 1.100 1.065

Case 1: (ua-uw)GS,0 = 50 kPa


At the end of the inizialization phase (i.e. before the application of the rainfall history),
the cells C, D and E respectively show FSC,t0 = 3.272, FSD,t0 = 1.213 and FSE,t0 = 1.206
(Tab. 3.4).

48
As reported in Fig. 3.25, after the 7 hours of the applied rainfall period, failure doesnt
verify in correspondence of any cell. Factor of safety of the cells C, D and E are
respectively FSC,t7 = 2.559, FSD,t7 = 1.204 and FSE,t7 = 1.193 (Tab. 3.4). There are 11
cells external to the triggering zone with a factor of safety comprised between 1.00 and
the factor of safety FSD,t7.

Case 2: (ua-uw)GS,0 = 40 kPa


At the end of the inizialization phase, the cells C, D and E respectively show FSC,t0 =
3.142, FSD,t0 = 1.176 and FSE,t0 = 1.167 (Tab. 3.4).
As reported in Fig. 3.26, at the end of the rain event the cell C attains failure, while cells
D and E have respectively FSD,t7 = 1.160 and FSE,t7 = 1.141 (Tab. 3.4). There are 12
cells external to the triggering zone with a factor of safety comprised between 1.00 and
the factor of safety FSD,t7.

Case 3: (ua-uw)GS,0 = 30 kPa


At the end of the inizialization phase, the cells C, D and E are characterized
respectively by FSC,t0 = 2.982, FSD,t0 = 1.129 and FSE,t0 = 1.119 (Tab. 3.4).
As reported in Fig. 3.27, only the cell C shows a factor of safety lower than 1.00 at the
end of the rain event. The cells D and E present respectively FSD,t7 = 1.100 and FSE,t7
= 1.065 (Tab. 3.4). There are 13 cells external to the triggering zone with a factor of
safety comprised between 1.00 and the factor of safety FSD,t7.

Figure 3.25. Results of case 1 at the end of the 7th hour: no one cell has a factor of safety lower
than 1.00; 11 cells (yellow cells) external to the triggering zone present a factor of safety
comprised between 1.00 and the factor of safety FSD,t7 of cell D (equal to 1.204)

49
Figure 3.26. Results of case 2 at the end of the 7th hour: cell C is the only one with a factor of
safety lower than 1.00; 12 cells (yellow cells) external to the triggering zone present a factor of
safety comprised between 1.00 and the factor of safety FSD,t7 of cell D (equal to 1.160)

Figure 3.27. Results of case 3 at the end of the 7th hour: cell C is the only one with a factor of
safety lower than 1.00; 13 cells (yellow cells) external to the triggering zone present a factor of
safety comprised between 1.00 and the factor of safety FSD,t7 of cell D (equal to 1.100)

50
Comparing the three examined cases, we have to observe that:
- no one case furnishes false alarms;
- case 1 doesnt report any failure;
- case 2 and case 3 are able to forecast landslide C;
- no one case is able to forecast landslides D and E.

In order to have FS 1 also in correspondence of cells D and E, an initial suction equal


to 20 kPa should be imposed at ground surface. The result connected to this
hypothesis is not reliable, because, as shown in Fig. 3.28, 246 cells external to the
triggering zone would be activated, generating in this way too many false alarms.

Figure 3.28. Results at the end of the 7th hour, obtained assigning (ua-uw)GS,0 = 20 kPa: 246
cells (red cells) external to the triggering zone present a factor of safety lower than 1.00

We retain that case 3 is the most reliable case, because it forecasts landslide C and
contemporarily calculates a factor of safety very next to 1.00 in correspondence of cell
D and E. A more detailed description (hourly results) related to case 3 will be reported
in the next paragraph.

51
3.2.1 Hourly results

According to the hypothesis (ua-uw)GS,0 = 30 kPa, the pyroclastic soil is characterized


by the permeability kw = 1.71E-06 m/s and the diffusivity D= 5.51E-05 m2/s (Tab. 3.1).
Table 3.5 reports the results at the end of each hourly step starting from the beginning
of the imposed rain event, in terms of:
- number of cells N with FS 1;
- number of cells N with 1 < FS FSD,t7, where FSD,t7 = 1.1 is the factor of safety of
the cell D calculated at the end of the 7th hour;
- factor of safety of the cell C, FSC;
- factor of safety of the cell D, FSD;
- factor of safety of the cell E, FSE.

Table 3.5. Results at the end of each hourly step of the rain event: number of cells N with FS
1 and with 1 < FS FSD,t7 (FSD,t7 = 1.1) ; factor of safety in correspondence of cell C (FSC), cell
D (FSD) and cell E (FSE)

N
t I (mm/h) FSC FSD FSE
FS 1 1 < FS 1.1
1h (22:00) 0.4 0 0 2.959 1.128 1.118
2h (23:00) 8.0 0 0 2.468 1.125 1.114
3h (24:00) 44.2 1 4 0.071 1.117 1.102
4h (01:00) 4.0 1 12 0.071 1.109 1.088
5h (02:00) 2.0 1 17 0.071 1.103 1.076
6h (03:00) 0.2 1 21 0.071 1.101 1.069
7h (04:00) 0.4 1 21 0.071 1.100 1.065

According to these results (represented from Figure 3.29 to Figure 3.35), cell C has the
highest factor of safety during the first two hours. As already observed for test case
nr.1, this is due to the great influence initially played by the cohesion on the factor of
safety at low depths (zmax,C, equal to 0.65 m, is the most small between them). At the
same time, the reduction of strength induced by the infiltration is inevitably more rapid
just at the cells characterized by smaller thickness. In fact, the cell C attains the failure
at the end of the 3rd hour - time 24:00 (Fig. 3.31), that is at the end of the interval
characterized by the most intense rainfall (I = 44.2 mm/h), while FSD and FSE are still
greater than 1.00. During the entire period of the rain event, the cell C remains the only
cell with FS<1.
On the other hand, we have to note that FSD is always higher than FSE. In particular,
FSE attains a value lower than 1.1 at the end of the 4th hour time 01:00 (Fig. 3.32),
contemporarily to other 8 cells and after that 5 cells already attained this value. FSD
attains its minimum value (FSD = 1.1) at the end of the 7th hour - time 04:00 (Fig. 3.35),
contemporarily to other 4 cells and after that 18 cells already attained this value. With
regard to the discussed analyses of the test case nr.1, these different susceptibilities to
failure are clearly due to different geometrical characteristics.

52
Figure 3.29. Factor of safety FS calculated after 1 hour (time 22:00) from the beginning of the
rain event: every cell is characterized by FS>1.1

Figure 3.30. Factor of safety FS calculated after 2 hours (time 23:00) from the beginning of the
rain event: every cell is characterized by FS>1.1

53
Figure 3.31. Factor of safety FS calculated after 3 hours (time 24:00) from the beginning of the
rain event: 1 red cell attains a value lower than 1.00 (cell C), 4 yellow cells are characterized by
FS1.1 (3 of them external to the triggering zone)

Figure 3.32. Factor of safety FS calculated after 4 hours (time 01:00) from the beginning of the
rain event: 1 red cell has a value lower than 1.00 (cell C), 12 yellow cells are characterized by
FS1.1 (6 of them external to the triggering zone)

54
Figure 3.33. Factor of safety FS calculated after 5 hours (time 02:00) from the beginning of the
rain event: 1 red cell has a value lower than 1.00 (cell C), 17 yellow cells are characterized by
FS1.1 (11 of them external to the triggering zone)

Figure 3.34. Factor of safety FS calculated after 6 hours (time 03:00) from the beginning of the
rain event: 1 red cell has a value lower than 1.00 (cell C), 21 yellow cells are characterized by
FS1.1 (14 of them external to the triggering zone)

55
Figure 3.35. Factor of safety FS calculated after 7 hours (time 04:00) from the beginning of the
rain event: 1 red cell has a value lower than 1.00 (cell C), 21 yellow cells are characterized by
FS1.1 (14 of them external to the triggering zone)

Runoff
Results are reported from Figure 3.36 to Figure 3.42. The maximum and very diffuse
values of runoff are shown at the end of the 3rd hour (Fig. 3.38), underlining that a not
negligible portion of the rainfall, whose intensity is almost one order of magnitude
greater than the hydraulic conductivity of soils, is not able to infiltrate. In particular, a lot
of cells present a runoff higher than 1000 mm/h, which, starting from upslope, flows
along the natural channels and reaches the urban district set downslope. However, its
worth noting that the code tends to overestimate the runoff calculated at the end of
each period, because, as already discussed, the phenomenon is assumed to verify
instantaneously from cell to cell.
Runoff smaller than 10 mm/h are calculated at the end of the 1st, 6th and 7th hour: the
corresponding rainfall intensities are not so strong and so they are able to practically
full infiltrate into the soil.

56
Figure 3.36. Runoff calculated after 1 hour (time 22:00) from the beginning of the rain event

Figure 3.37. Runoff calculated after 2 hours (time 23:00) from the beginning of the rain event

57
Figure 3.38. Runoff calculated after 3 hours (time 24:00) from the beginning of the rain event

Figure 3.39. Runoff calculated after 4 hours (time 01:00) from the beginning of the rain event

58
Figure 3.40. Runoff calculated after 5 hours (time 02:00) from the beginning of the rain event

Figure 3.41. Runoff calculated after 6 hours (time 03:00) from the beginning of the rain event

59
Figure 3.42. Runoff calculated after 7 hours (time 04:00) from the beginning of the rain event

4. Conclusions

This report contains the results of some numerical analyses aimed to simulate the
rainfall-induced shallow landslides, which occurred on October 13th, 2004 and on
September 17th, 2005 along the south-eastern side of the Camaldoli hill, a slope set
inside the western sector of the urban district of Naples. The analyses have been
carried out thanks to the use of the Transient Rainfall Infiltration and Grid-based
Regional Slope-stability (TRIGRS) program (Baum et al., 2002). This is a GIS-based
code which evaluates the slope stability, combining the solution coming from the
infinite-slope model with a one-dimensional analytical solution for transient pore
pressure response to rainfall infiltration. Like every GIS-based landslide model, the
program operates on a map area that is subdivided into a 3D-cells grid. As a matter of
fact, the quality of the response of this kind of analyses is strictly related to the
representativity of the input data, which may strongly differ over the entire examined
area. With regard to this topic, TRIGRS requires detailed information about:
- rainfall records;
- topography;
- soils properties;
- boundary condition (depth of the lower impervious bedrock);
- initial hydraulic conditions.
The assignment of the characteristics to the rain events which triggered the landslides
events, has been possible thanks to the hourly rainfall measurements made by a

60
pluviometer installed by the Civil Protection of the Campania Region, which registers
without any interruption from October 2000.
Topographic information have been derived from a high-resolution Digital Elevation
Model (DEM), created using the available topography map of the area, realized by the
local Provincial administration.
Soils properties have been assigned taking into account the results of some
experimental laboratory tests.
Due to the availability of few measurements in the study area, the depth of the
impervious boundary depth has been considered as a function of the topographic slope
data.
With regard to the initial hydraulic conditions, we have decided to assign the initial
depth of the water table as a function of the hypothesized suction at ground surface, in
turn varied within a range of plausible values.
According to the analyses regarding both the examined test cases, the sensitivity of the
program to soil properties and the initial conditions (in turn not independent between
themselves) seems rather high. In particular, the susceptibility to failure of the cells grid
considerably grows with the permeability and decreases with initial suction imposed at
the ground surface. That means that it has not been possible to forecast the landslides
that effectively have been recorded, without giving false alarms (i.e. observing failure
in correspondence of other cells). Its likely that an accurate knowledge of local
conditions, which could be made possible by more detailed surveys, could obviate this
not reliable result. Nevertheless, we retain that TRIGRS is able to furnish useful
indications in order to individuate, during a rainfall event and over large areas, the
landslides risk slopes that, for this reason, need to be analyzed in depth through the
use of more sophisticated models. Besides, the runoff calculations supported by the
code may furnish also an estimation of the not infiltrated rainfall water and
contemporary of the preferential paths covered by the flow, helping to recognize if there
are flood risk areas.

5. References

Baum RL, Savage WZ and Godt JW (2002). TRIGRSA fortran program for transient
rainfall infiltration and grid-based regional slope-stability analysis. U.S. Geological
Survey Open-File Report 02-0424

Brooks RH and Corey AT (1964). Hydraulic properties of porous media. Hydrology


Paper No.3, Colorado State Univ., Fort Collins, Colorado

Croce A and Pellegrino A (1967). Il sottosuolo di Napoli - Caratterizzazione geotecnica


del territorio urbano. Proc. VIII Convegno di Geotecnica, Cagliari, Vol. 3

Di Crescenzo G, Rotella M and Santo A (2007). Il contributo della geologia per lo


studio dei meccanismi di innesco di colate rapide di fango al campo sperimentale di
Monteforte Irpino (primi risultati). In Piattaforme Evolute di Telecomunicazioni e di

61
Information Technology per l'Offerta di Servizi al settore Ambiente PETIT-OSA a cura
di C.Nunziata Aracne editrice srl Roma ISBN 978-88-548-1184-3.

Iverson RM (2000). Landslide triggering by rain infiltration. Water Resour Res, 36, pp.
18971910

Jenson SK and JO Domingue. (1988). Extracting Topographic Structure from Digital


Elevation Data for Geographic Information Systems Analysis. Photogrammetric
Engineering and Remote Sensing 54(11), pp. 1593-1600.

Mualem Y (1976). A new model for predicting the hydraulic conductivity of unsaturated
porous media. Water Resources Research, 12, pp. 513-522.

Savage WZ, Godt JW and Baum RL (2003). A model for spatially and temporally
distributed shallow landslide initiation by rainfall infiltration. Proc. III Int. Conf. on Debris
flow hazards mitigation: mechanics, prediction, and assessment, Davos, Switzerland,
pp 179187

Vanapalli SK, Fredlund DG, Pufahl DE and Clifton AW (1996). Model for the prediction
of shear strength with respect to soil suction. Canadian Geotechnical Journal, 33, pp.
379-392

62

You might also like