Académique Documents
Professionnel Documents
Culture Documents
ffi Slope angles are a key parameter in estimating susceptibility to developing earth flows.
In this paper, slope angles are used to estimate potential unstable areas in a pilot sector
of the Benevento province in (Southern ltaly).
Since the study area is characterized by a complex lithological setting, landslide
distribution was analyzed within four-groups of homogeneous litho-technical sequences.
Slope angle frequency distributions were obtained from a landslide sample in accordance
with the Weibull probability density distribution function. Their analysis shows that the
largest occurrence of landslides fall within an interval of slope angles ranging from 9' to
14". As filed surveys confirm, the low frequency of instabilities on steeper slopes can be
explained by a deficit of potentially involving materials, partially due to the presence
of stony sequences. Consequently, the probability of failure was calculated only on slope
angle ranges already affected by existing landslide phenomena.
itt.ffiÈ$
Earth flows . Weibull distribution . Southern Italy
Alluvial ssqu&nffi§
the Benevento province in Southern Italy, where considerable approximate area of 358 k^', approximately 18 7o of ::.
data relating to slope angles and lithological sequences was entire surface area of the province (Fig . 2). By compar-r":
available (Guadagno et aL.2006). Figs. l, and 2, one can note that landslide phenome:-
are mainly connected to Cl, Ar-Cl-Cg, Cl-S and C1-l :
litho-technical sequences.
Lands1idedistributionactivity(sensuwP/TVLI1993
Geological Features and Landslide Occurrence
generally advancittg, except in the source area where i:
in the Study Area
retrogressive.
Moreover, large earth flows usually affect the full slc:"
The Province of Benevento (approximately 2,A00 ktt') has
length, even though smaller and secondary phenomena, rn.l-': "
been historically affected by landslides prevalently of the
have often different directions and velocities, as highli-ehi.:
earth flow type. Its morphological pattern is controlled
in the literature for similar instabilities (e.g. Corominas eI :
by both the regional geo-structural setting, and the outcrop-
2005; Lollino et al. 2003), take place within the main lar :'
ping terrains which mainly consist of structurally complex
slide body. Usually , reacttvations involve the full lands-:.
formations (sensu Esu 1977).
bodies or limited sectors during long and intense storms.
Revellino et al. (2010) grouped these deposits into successions
on the basis of their lithological and engineering-geological/
geomechanical features (Fig. 1) in order to elucidate the
relationship between landslides and the geological forma- Probability of Failure and Landslide
tions. Within these groups, the analysis was carried out Susceptibility
only on the following sequences affected by earth
flows: (1) Clayey sequences (Cl); (2) Arenaceous-clayey- Scientific Background
conglomeratic sequences (Ar-Cl-Cg) ; (3 ) Clayey- silty sequences
(Cl-S) and Clayey-marly sequences (Cl-M) Parametric statistical analysis is well suited to add:.,, -
As regards landslides, recent studies carried out by sing landslide susceptibility assessments of geomorphoic " ;
Guadagno et al. (2006) and Revellino et al. (2010) led to environments dominated by recurrent slope instabilit-"
an inventory of more than 3,100 earth flows, covering an The experimental results obtained by PWRI (1916), Ohn - *
Slope Angle as Indicator Parameter of Landslide Susceptibility in a Geologically... 427
1*"3I Y.,t f,..,.-. n controls the expansion and the reduction of the
,iffi probability density function graphs, and moving the
NIffi
{#H
4,ÉM$*tf
peak position, n indicates the theoretical mode of a
failure. Iwahashi et al. (2003) clarifled that the more
fine-grained the geology, the smaller the value of q
would thus result;
y controls the parallel movement of probability density
function graphs.
The cumulative Weibull distribution function F(A)
(3) defines the probability of failure (Iwahashi et al. 2003).
F(A) is def,ned by the following equation:
F(A)-P(A<*) P(A
- o,) (3)
-@*-
& s
ffi *&ffiffi
Fit=94.4"/" Fit=95.1 %
0.01 0.01
.=
=at- a
cq)
c) 0.001 0.001
E E
.= =
-o
(§
-o
(s
-o 0.0001 -o 0.0001
o
L
o
L
o_ o_
0.00001 0.00001
Fit=96.9%
0.01 0.01
=aC .=
a
C
G)
ìo
0.001 o 0.001
E
.=
=
-o
(§ -o
(6
-o
o 0.0001 -o 0.0001
L
o
L
o_ o-
0.00001 0.0000'1
0.0000001 0.0000001
20 20
Slope, degrees Slope, degrees
Fig. 3 hobability density function versus slope angle in the landslide area for each litho+echnical sequence analyzed: (a) Clayey sequences.
(b) Arenaceous-clayey-conglomeratic sequences,Ar-Cl-Cg; (c) Clayey-silty sequences, C/-S; (d) Clayey-marly sequences, C/-M
correlation. In particular, by approaching the peak, the values As shown in Fig. 4a, b, which refer to sequences u lie:-
strongly decrease. In this way, it was possible to identify the the stony component is lacking or poorly present. C;l
slope sectors where landslides are spatially distributed. appears to be univocally distributed in terms of trend a::
Moreover, the PDFs obtained for each litho-technicatr peak values, being the curves fully overlapped (the avera;-
sequence show peak values between 9-11o,11-13" and shift is: 5.91 7o and 1.52 7o, taspectively). This result cou.:
14-16o where landslides involved Cl and Cl-M (Fig. 3a, d), highlight the fact that the slope steepness was reduced :
Ar-C1-Cg (Fig. 3b) and Cl-S (Fig. 3c) respectively. extreme landslide processes. Indeed, in these areas. r,.
Landslide Index, calculated as the percentage of the er;.i
affected by landslide events for a 1km2 grid for the n'h.,,.
Probability Distribution of the Slope Angle (a) province, is even more than l0 7o (Revellino et al. 201{l
On the contrary, the data-set distributions of Fig -lc" :
In order to analyze the evolutive trend of slopes induced show a poor agreement of the peak values, which i:"-
by instabilities, the spatial distributions of u were compared significantly shifted, and a poor overlapping of the cun. ;:
in both landslide and stable areas. Data was normaltzed (the average shift is: 26.19 7o and 13.18 7o, respectir er",,
in relation to the max of the two curves, allowing the range above all in the case of Cl-S sequences (Fi g. 4c). Moreor, ;.
of the probability distribution variability to be restricted distribution tails also display a different behavior, rr,hi:-
between 0 and 1. might be influenced by the poor number of e\-ùr--:
5,ope Angle as lndicator Parameter of Landslide Susceptibility in a Geologically... 429
b Stable Area
Stable Area
E.
&
s
à1 &
fu *.$
% .=
= a ##
ioB #o I 0.8
-w
5 #
.=
#$
:3- 06 # E
-o
0.6
o Landslide Area
! Landslide Area o- dd
i 0.4 # g 0,4
e$
E
!
=
Z-
0.2
d .N
(É
E
zo
0.2
0
i
5 10 15 20 25 30 10 15 20 25 30
Slope Angle (degrees) Slope Angle (degrees)
rea
=a à1
a
C
I 0.8 I 0.8
.= È
-o -o
E 0.6 .8 0.6
o
o-
Landslide Area Io-
E 0.4
.N
8
.N
0.4 Landslide Area
(6 (§
E tr
zE
0.2
zb
0.2
15 20 25 30 10 15 20 25 30 35 40
Slope Angle (degrees) SIope Angle (degrees)
Fig. 4
Area probability density distribution, normalized in relation to the peak, vs slope angles of both landslides and stable area: (a) Clayey
sequences, C/; (b) Arenaceous-clayey-conglomeratic sequences, Ar-Cl-Cg; (c) Clayey-silty sequences, C/-S; (d) Clayey-marly sequences, C/-M
inventoried in this geological context too. As regards Cl-S respect to the slope angle distribution of the entire landslide
and Cl-M sequences, landslide area distributions show area.In addition, the peak values of the u probability density
prevailing slope angles around 10*12 . On the other hand, of the source areas are signiflcantly higher (13-15") with
distributions of the stable areas are characterrzed by angles respect to those of the channel areas (8-9"). This could
betweenT" and 9o. probably be related to the geo-mechanical resistance offered
conflrmed by fleld surveys, it should be noted
As by the stony component that, in these sectors, constrains
that landslide source areas are slowly evolving where the failure along the pre-existing weak surfaces such as fractures
stony terrains are prevailing at the top of the sequence or bedding joints (Grelle et al. 20I1b). The result is a
involved in landsliding. This f,eld evidence is supported by slow and irregular retrogression, locally influenced by the
the distributions of Fig. 5, where slope angles of the land- fracture pattern of the stony terrain. The consequence is high
slide source areas are compared with those of the respective evolutive control on the distribution of the activity as
channels. demonstrated by Revellino et al. (2010).
Excluding the analysis on the CL-S sequence, which
could be influenced by the poor statistic representativeness
of data due to the exiguous number of landslides recorded Probability of Failure
(# l4), the remaining sequences are charactenzed by
distributions (Fig. 5) that are almost unchanged with respect A statistical analysis of the frequency-mean slope angle
to those in Fig. 4. relationship was carried out in order to deflne the probability
In the case of the Cl-M sequence, the cr distribution of of failure. The data f,tting was performed by using the
the source area is only slightly overlapped and shifted with 3-parameter Weibull density function (Fig. 6). The result
430 A. Donnarumma et a
a
1
=aC
o à
c 0.8 ?
o 0.8
= o
-o
(U -_
-o 0.6 4§ 0.6
o
L
o- -o
o o
L
o o-
0.4
.N
(6 E
.N
0.4
E
o
E
z 0.2 E
zò 0.2
0
15 20 25
Slope angle (degrees) 15 20 25
Slope angle (degrees)
b d
=, 0.8 .à @
0.8
o C
E o
E
-o
(§
I t i!
à 0.6
-o
(§
-o
o _o
o_ o
.o 0.4 o- o.4
o Eo
.N
6
C
.N
6
l-
L
o
,_\ ,l
v-4 E 0.2
z o
z
15 20 25 30 15 20 25
Slope angle (degrees) Slope angle (degrees)
Ffig, S Area probability density distribution, normahzed in relation to the peak, vs slope angles of both landslides and source areas: (a) C-,
sequenc es, C l; (b) Arenaceous-clayey-conglomeratic sequences, Ar-C l-C g; (c) Clayey-silty sequences, C/-,S; (d) Clayey-marly sequences. -
h
I
B*t*t flt p*ram*t*n; i ffi**t *it p#flffi §?,l#t*rti:
{fi*& r**e
$* ?"?S i F* 2'fr$
.f* *"*5 y* *'#*
** '
-*
ffiq
q*l*.33 dt!
t,*Lt,tI
b* FB
$ §;3
L *
ss%
t&é
s&
-
$ry*
f,
{i.
IL TL
f il.sr **4;3#r/e
*%
fr* W Ii§ fr* ** d#
frdasn slspe e*6k ffirwx},
d t**
ffisxr {ir ffiEm}?trtsrx; ffis t fit p#r$§§€É6rs:
f}= 3.?* $* ?'*$
Y* #'?il ?§ I '#*
-6
8" 4* I3"${} -* t*%
# tl*.$.ffi
3t Hl*
{,} t;
q3 &
fil' §%
& a
[L
1*
TL s%
Fig. 6
Mean slope-angle frequency distributions of landslides for each geological sequence: (a) Clayey sequences, C/; (b) Arenaceous-:-:
conglomeratic sequences, Ar-Cl-Cg; (c) Clayey-silty sequences, CÀS; (d) Clayey-marly sequences, C/-M
Slope Angle as lndicator Parameter of Landslide Susceptibility in a Geologically.. 431
ffi
IÉ*.TL 1t-*fi
*I r"/
dr
*1"ffi
*t"S
or#
r f§ ..?
c,
*
#
ri rp
$
L $ ,{ o. r u.02
t r §o,
*s%
=(!
th
J
tj "Y,
(a
*- r# ; g"
c ,t ,# #l (§
(l)
gh iJ, f
rs
0)
1f-s3
-o 40%
G
ldj
rf -"
c!.
a
U)
.EÌ r §..'
o
L
f#:
1r-04
,
o- I
t;
,J[
f#Ys t§
.§i
tg
.§ 1E-0S
,H ro !m 4s sù B* r*s
r"d irn
tr-{17'}
su s0 70 $*
Slop* *r*g** {d*gr**e}
*tL
0 § t$ rs rs gs ** §s 4*
Ffi#" $ Distribution curve of the landslide area in relation to the
SIrp*, degn**r
slope angle
Fig" F Cumulative distribution functions of mean slope-angle distributions
rf landslides for each geological sequence
in gently slope shapes alternated by rough and steep
IEétJB
morphologies. The latter arc characterized by a low spatial
ON#ffHYtr landslide frequency and, therefore, a lack of materials prone
o
H;B&
*.n * '"-
o3 to developing landslides.
ffiC& +
B{}% .o *ou
*(l Therefore, when operating in geologically complex
ùJ
** areas, the probability computed by the Weibull model must
€
{r+ be calibrated and interpreted in the specific morphological
G
§) *s context. In fact, the cumulative distribution function C(A)
L tr{JH} e*
{§
0, **+ (6) shows that more than 95 7o of the outcropping of the
^
Él
o
a
À
four litho-technical sequences has slope angles lower than
o mH* 17" (Fig. 8).
't#'t&
.J
G {j'
rt
sffi # f;l
E s. :
(6)
J
& fi* *il-Ht
u da§&
6 iF , a;1a
4o sl_§
*
s & &r-SI*#g
sl
t* This value represents the point of intersection (roll-over, fl*)
si,&
#* of two data sets in the bi-logarithmic graph of the slope angle
# # -**. t* tr* È,§, *ttÉ
(Fig. 9). In particular, the spatial analysis of q shows that the
§kpe, d*gr***
areal frequency distribution beyond the roll-over is controlled
F*9. & Cumulative slope angle distribution by the following power law equation:
shows a unimodal asymmetric shape with peak values gen- y 2E+ 08 - 1.909u(r'
- - 0.96)
erally around 10' .
The cumulative Weibull distribution function F(A) was This correlation is valid for slope angles ranging between
thus computed for the four litho-technical sequences
(Fig. 7). Although the preliminary analysis of the probability
of failure curves shows that the model seems to describe well old deflning slope angles influenced by the outcropping
the variability of experimental data, some interpretations of stony sequences. It is assumed that slopes having a
should be necess artly made. First, the morpho-structural slope angle higher than fl* coffespond to sectors where
setting of large-scale areas may influence the slope angles stony layers are outcropping.
distribution, increasing its variability. Second, the heteroge- In order to adapt the model to the speciflc study
neous nature and setting of the outcropping sequences play case, the probability of failure curves were recalculated
a dominant role in the slope gradient pattern, resulting considering the influence of u*.
432
"* tu s st_F#
-fif; study site and a post-processing analysis addresse:
s*?t **-s contextuahze the output data are needed.
@ Af*#***g
&,
The results obtained indic ate that accvrate slope rr: r
t*
t analyses considerably improve landslide susceptibilin s:- --
ffi
f§
u-
6t*
s*?a ies in areas charucterrzed by complex terrains, evalu; ",- *
t3
},}
its effective contribution as landslide predisposing fac: - :
H
,ffi
{s 4S3;
#* References
f
l*
tu
Corominas J, Moya J, Ledesma A, Lloret A, Gili JA (2005) Pre;-: :
ft,
à*9a +{ §}
of ground displacements and velocities from groundwater :
changes at the Vallcebre landslide (Eastern Pyrenees. S, , "
n,6
-tr #qt
,Ér Landslides 2(2):83-96
r" -f"*"{{
**L
Environmental Systems Research Institute (ESRI) (1999) A:: :
the study area the mainly clayey sequences show a good 2001 .01.015
regular Gaussian disposition with peak values, 9-14", Iwahashi J, Watanabe S, Furuya T (2001) Landform analysis c- , "
which are included in the shear strength fully-softened movements using DEM in Higashikubiki area, Japan. Ct --:
Geosci 27 (1):851-865
to residual range obtained by means of ring shear tests on Iwahashi J, Watanabe S, Furuya T (2003) Mean slope-angle frec*:-
tipical clayey samples (Grelle and Guadagno 2010; distribution and size frequency distribution of landslide ma-...
Grelle et al. 2011a). Higashikubiki area, Japan. Geomorphology 5A349-364
However, the fltting Weibull functions for the four Korup O (2005) Distribution of landslides in southwest New Ze :--"'
Landslides 2:43-51
study sequences, Cl, Ar-Cl-Cg, Cl-S and Cl-M, show a Lee CT, Huang CC, Lee JF, Pan KL, Lin ML, Dong J (2008) Stai-, "
different pattern which is referable both to the interaction approach to storm event-induced landslides susceptibility. Na, .. -
between the geo-structural sequence and the slopes and ardBarth Syst Sci 8:941-960
the retrogressive evolution of landslides, which is typical Lollino G, Lollino P, Bertolino G (2003) Analysis of the beha'v'io:-
*
large landslide in structurally complex soils by means of monl: . ' .
of source areas.
f,eld data. In: Proceedings of the international conference - - "
As reported in the literature, the probability failure slope movements prediction and prevention for risk miti-ea:- - -
analysis by considering slope angles only should take into Naples, May 11-13. Vol 1. Ed. Patron. Bologna,pp 317-3):
account the real lithology of the slope. In this study the Ohmori H, Sugai T (1995) Toward geomorphometric mode-,
estimating landslide dynamics and forecasting landslide . - -
threshold angle between the low and high stony abundance
"
slope angle. Consequently, the suitable slope failure analy- landslides no.2): Technical Memorandum of PWRIIIZ1 (in Jap--. ,