Sie sind auf Seite 1von 24

Landslide hazard

assessment: summary
review and new
perspectives
P. Aleotti 7 R. Chowdhury

Abstract This paper deals with several aspects of présente un bilan résumé et une classification des
the assessment of hazard and risk of landsliding. In principales approches qui ont été développées de par
recent years the interest in this topic has increased le monde. La première étape consiste à distinguer les
greatly and there are many technical papers dealing méthodes qualitatives et quantitatives. Le premier
with this subject in the literature. This article pres- groupe est basé principalement sur l’expérience
ents a summary review and a classification of the régionale d’experts, avec une notion de prédisposi-
main approaches that have been developed world- tion aux glissements déterminée directement sur le
wide. The first step is the subdivision between quali- terrain ou par combinaison de différentes cartes
tative and quantitative methods. The first group is d’indices. Les approches du second groupe sont
mainly based on the site-specific experience of théoriquement plus rigoureuses. Il est possible de
experts with the susceptibility/hazard determined différencier des analyses statistiques (bivariées ou
directly in the field or by combining different index multivariées) et des méthodes déterministes qui font
maps. The approaches of the second group are intervenir l’étude de pentes ou de sites particuliers,
formally more rigorous. It is possible to distinguish basée sur des modèles géomécaniques. De telles
between statistical analyses (bivariate or multivar- analyses peuvent être déterministes ou probabilistes.
iate) and deterministic methods that involve the Parmi les méthodes quantitatives présentées, on
analysis of specific sites or slopes based on geo-engi- trouve l’approche par les réseaux neuronaux qui n’a
neering models. Such analyses can be deterministic été que récemment appliquée à des problèmes de
or probabilistic. Among the quantitative methods géologie de l’ingénieur. Enfin, plusieurs considéra-
discussed is the Neural Networks approach which tions relatives au concept de risque acceptable et de
has only recently been applied to engineering gestion du risque sont présentées.
geology problems. Finally several considerations
concerning the concept of acceptable risk and risk Keywords Landslides 7 Hazard 7 Risk Assessment 7
management are presented. Neural Networks 7 G.I.S

Résumé Cet article traite de plusieurs aspects de Mots clés glissement de terrain 7 aléa 7 évaluation
l’évaluation de l’aléa et du risque de glissement de du risque 7 réseaux neuronaux 7 S.I.G.
terrain. Ces dernières années, l’intérêt pour ce sujet a
grandement augmenté et de nombreux articles tech-
niques s’y référant ont été publiés. Cet article

Landslides – significance and


Received: 30 November 1998 7 Accepted: 2 April 1999 impact
P. Aleotti (Y)
Via Tolstoj 86, 20098 San Giuliano Milanese (MI) – Italy In recent years the assessment of landslide hazard and risk
e-mail: pietro.aleotti@snamprogetti.eni.it has become a topic of major interest for both geoscientists
Fax: c39 02 52052586 and engineering professionals as well as for the community
R. Chowdhury and the local administrations in many parts of the world.
University of Wollongong, Department of Civil Mining and The main elements of a risk assessment system are shown
Environmental Engineering, Northfield Avenue, Wollongong, in Fig. 1. The majority of available literature is based on
New South Wales, 2522, Australia the technical terminology introduced at the beginning of

Bull Eng Geol Env (1999) 58 : 21–44 7 Q Springer-Verlag 21


P. Aleotti 7 R. Chowdhury

the socio-economic significance of landslides and secondly,


Landslide risk assessment
the increased pressure of development and urbanisation
on the environment. As development increases on sloping
Currently stable slopes Currently unstable slopes
urban areas, a higher incidence of slope instability and
landsliding is reported. A database of reported events of
slope instability extending back many decades has been
developed for the Greater Woolagong urban area in
History of No history of
Australia (Fig. 2). From this the effect of increasing devel-
landsliding landsliding
opment and urbanisation has been inferred by Flentje and
Chowdhury (1999). The increase of landslide risk is often
Quantifying extent of due to the increase in both the hazards and the elements at
Effect of development movement
risk (Fig. 3).
In 1981 Varnes estimated that from 1971 to 1974 nearly 600
Is risk people were killed every year world-wide as a consequence
low, moderate or high? of slope failures. Approximately 90% of these deaths
Effect of development
with and without occurred within the Circum Pacific Region. In 1979 the
Quantifying risk controls International Association of Engineering Geology Commis-
sion on Landslides estimated that some 14% of the lives
lost in natural catastrophes could be attributed to land-
slides. Alexander (1994), considering all natural
Cost of assessment phenomena in the period 1967–1991, reduced this to 1.2%.
He estimated, however, if the only causes considered are
volcanic eruptions, tsunami, earthquakes and landslides,
Cost of controls 5% of the total would be attributable to landslides.
In Italy, owing to its particular geological and geomorpho-
Fig. 1 logic setting (Fig. 4) landslides account for nearly 37%
Main elements of a landslide risk assessment for a given sloping (Catenacci 1992) of lives lost due to all natural phenomena.
site (after Chowdhury 1998c) Of this percentage value, 63% is accounted for by two
episodes, the 1992 Val di Stava landslide with a death toll
of 285 and the 1964 Vajont landslide with almost 2000
the 80’s (Varnes and IAEG 1984) and on minor variations victims. The latter is the most disastrous landslide which
proposed by other authors (Einstein 1988; Fell 1994). The has occurred in Europe in this century. A recent analysis of
reader is assumed to be familiar with such terms as suscep- this dramatic event, extending beyond a technical consid-
tibility, hazard, vulnerability and risk and their use with eration, is given by Paolini and Vacis (1997). Other signifi-
respect to slope instability and landsliding hence their defi- cant events took place in Tresenda in July 1983 (Polloni
nitions are not repeated here. For a discussion on mecha- and Aleotti 1998) and in Val Pola in July 1987, with 17 and
nisms and uncertainties with regard to landslides as 38 victims respectively (Alexander 1988; Cambiaghi and
natural hazards, see for example Chowdhury (1980). Schuster 1989). More recently, a major flood in the Pied-
The reasons for the increasing international interest in mont Region (November 1994) resulted in 70 deaths, of
landslides are twofold: firstly an increasing awareness of which 20 can be attributed to landslides (Aleotti et al,

50
Fig. 2
The increase with time of 45 Media references
instability in a sloping urban Data base references
40
Number of landslides recorded

area, the Greater Wollongong


Area of New South Wales,
35
Australia. The effect of
increased development of 30
sloping areas with time can
be inferred from such studies 25
(after Flentje and Chowdhury
20
1999)
15

10
5

0
1870

1880

1890

1900

1910

1920

1930

1940

1950

1960

1970

1980

1990

2000

Year

22 Bull Eng Geol Env (1999) 58 : 21–44 7 Q Springer-Verlag


Landslide hazard assessment

Risk = (Hazard) (Vulnerability) (Elements at risk) 2%/year. This implies that the economy should grow by at
least 6%/year to sustain economic development compared
with the 2–3% achieved recently.
Continued deforestation in landslide-prone areas
While landslides have a significant impact, their impor-
Increased regional and local precipitation tance is often under-estimated as the landslide damage is
Increased seismic activity
considered simply as a result of the triggering processes
Increased urbanization and development
and thus is included in reports of other phenomena such
as earthquakes, flood etc (Schuster 1996).
Fig. 3 Landslide costs in United States, Japan, the Alpine nations
The main reasons of the increase of landslide risk associated with (Austria, France, Italy and Switzerland) and India are quite
development similar in magnitude ($1–$5 billions per year) although, it
is likely that actual costs are highest in Japan. In Italy
natural disasters have resulted in damages of more than
1998a) while in May 1998, the disaster which that struck
$1.8 billion per year over the last 45 years and even when
Southern Italy (Campania Region) and took 161 lives was
costs resulting from loss of wealth are not included
caused by more than 100 soil slip-debris flows (Guadagno,
amounted to more than $4 billions/year during the decen-
del Prete and Hawkins 1998). This again brought landslides
nium 1980–1989 (Catenacci 1992). Data on the social and
and their catastrophic effects dramatically to the attention
economic effects of landslides are now becoming available
of not only the Italian public, but also to local and national
from other countries. In Australia for example, this
Governments and indeed to people throughout the world
problem is taking on increasing importance (Fell 1992;
(Onorati 1998). An interesting review of some historical
Chowdhury 1993; Chowdhury and Flentje 1996, 1997,
mountain landslides that affected Italy and, in general, the
1998b; Leiba 1998) and landslides are one of the primary
European Alps, has been presented by Skermer and
hazards being mapped by the Australian Geological Survey
VanDine (1992).
Organisation as part of its Cities Project (Granger 1998).
As regards the economic damage in many developing
countries, natural catastrophes account for 1–2% of the
gross national product (Hutchinson 1995). In many cases
these effects contribute to economic stagnation and lack of Basic concepts of hazard
development. A well-known example is that of Costa Rica
(Mora 1995) where population growth is 2.6%/year and the assessment including mapping or
decline in GNP caused by natural disasters is estimated at zonation
As a basis for landslide hazard/risk assessment and zona-
tion researchers generally make four fundamental assump-
tions (Varnes and IAEG 1984; Hutchinson 1995):
a) landslides will always occur in the same geological,
geomorphological, hydrogeological and climatic condi-
tions as in the past
b) the main conditions that cause landsliding are
controlled by identifiable physical factors
c) the degree of hazard can be evaluated
d) all types of slope failures can be identified and clas-
sified.
For an evaluation of landslide susceptibility it is necessary
to answer the following basic questions:
a) where will the landslides occur?
b) what type of failures will occur?
c) how will the landslides occur?
Only by answering these questions can one define the land-
slide-susceptibility in the study area. For a complete
hazard assessment, however, the questions of both when
the landslides will occur and the frequency with which they
will occur over time must be addressed.
There are a number of obstacles to be overcome
including:
a) the discontinuous nature (in space and time) of slope
Fig. 4 failures
Percentage of lives lost due to different natural phenomena (data b) the difficulty of identifying the causes, the triggering
by Alexander 1994; Catenacci 1992; Varnes and IAEG 1984) factors and the cause-effect relationships, and

Bull Eng Geol Env (1999) 58 : 21–44 7 Q Springer-Verlag 23


P. Aleotti 7 R. Chowdhury

c) the lack of complete historical data concerning the particular area on the basis of the local environmental
frequency of these geomorphologic processes. conditions.
This last aspect distinguishes landslides from other The highest hazard is generally, if not always, associated
phenomena such as floods and earthquakes, for which with known landslide areas, i.e. sloping areas which have
detailed information is often available both on the inten- suffered instability in the past. However, neither qualita-
sity of the past events and their frequency. It is only for tive nor quantitative methods of analysis have, until quite
certain types of slope failures such as soil slips and debris recently, addressed the question of differentiating the
flows that the difficulties mentioned in (b) and (c) can be hazard in existing landslide areas. It has been shown by
resolved to some extent. Chowdhury and Flentje (1998a) that the hazard may be
The cause-effect relationship is easily identifiable when, as significantly different from one existing landslide area to
in rainfall-induced landslides, the statistical analysis of another. They ranked the hazards associated with existing
rainfall data may enable both the definition of the trig- landslides in a specific study area in Australia and found
gering threshold (usually in terms of intensity and dura- this was validated on the basis of the recorded impact of
tion of rainfall) and calculation of the recurrence interval landsliding in that study area. These findings are impor-
of critical rainfall, even as a function of local characteristics tant for further development of reliable methods of land-
(thickness of surficial layers, soil permeability, etc.) slide hazard assessment.
(Aleotti et al. 1996b; Caine 1980; Cancelli and Nova 1985; Another aspect that should be mentioned here is the work-
Crosta 1998; Moser and Hohenssin 1983; Polloni et al. scale to be adopted for mapping, investigation and assess-
1996; Wilson and Weiczorek 1995); see Fig. 5. ment of hazard. It has been proposed that the work-scale
According to current trends, an ideal map of slope insta- be chosen on the basis of three factors (Aleotti et al.
bility hazard should provide information on the spatial 1996a):
distribution, type, volume, velocity, travel distance and 1. The purpose of the assessment
retrogression limit of the mass movements predicted in a In land use, for example, planning on a regional scale
certain area in a given period of time (Hartlen and Viberg could be adopted (1 : 100.000–1 : 500.000), whereas for more
1988). However, owing to the difficulties stated above, specific problems such as the implementation of large
examples of this type of map are extremely rare in the engineering structures or the definition of a plan of
literature and a complete risk assessment procedure such priority measures, the scale should be more detailed
as the 5 steps proposed by Einstein (1988) has limited (medium scale: 1 : 25.000–1 : 50.000). Frequently, these
practical application. analyses are preliminary to more specific studies to be
The probabilistic component in hazard assessment conducted on perhaps a single slope or small areas where
requires understanding and assessment of many uncertain- an even larger scale would be appropriate.
ties. This is often a difficult task. Moreover, there is the 2. The extent of the studied area
problem of quantitatively determining the vulnerability of 3. Data availability
the elements at risk. These complexities make it hard to This aspect is often directly related to the extent of the
achieve an accurate risk assessment. As a consequence, examined area, and therefore, to the possibility of
rather than hazard, it is frequently susceptibility which is collecting useful data for the assessment through an appro-
assessed, i.e. the possibility that a landslide will occur in a priate cost/benefit analysis. For large areas, hazard assess-
ment may be based on the analysis and interpretation of
100 available data, while for smaller areas the assessment of
Soil depth (cm) stability and hazard could be facilitated by specific
−3
1·10 Permeability (cm/s) geotechnical investigations including observational proce-
Rainfall (R.T.= 10, 25, dures requiring subsurface monitoring devices such as
Intensity (mm/h)

50, 100, 250, 1000 years) inclinometers and piezometers.


−4
1·10 The choice of the work-scale affects the selection of the
10
−5 approach: thus, a statistical approach may not be suitable
5·10 1000 for studies concerning individual slopes or small areas
while a geotechnical engineering approach based on the
−5 40 80
calculations of safety factor and/or associated failure prob-
1·10 ability, would not be suitable at the regional scale.
10
1
1 10 100 1000
Duration (h)

Fig. 5
Rainfall intensity vs rainfall duration plot for different return
Essential information and input
time intervals (10, 25, 50, 100, 250, 1000 years) for the Langhe data
Area, Piedmont Region (NW Italy). Four different permeability
thresholds, with five soil thickness curves (40, 50, 60, 70, 80 cm),
are traced on the plot. This kind of analysis may be useful in the Specific facts and generalised information are required for
preliminary phases of hazard assessment of areas prone to surfi- many components of hazard assessment and risk manage-
cial landslides (soil slips) (after Aleotti et al. 1996b) ment of slopes and landslides. These include:

24 Bull Eng Geol Env (1999) 58 : 21–44 7 Q Springer-Verlag


Landslide hazard assessment

a) the assessment of performance over time geomorphological time scale. Factors to be considered
b) the consequences of failure including environmental have often been listed in the literature (Amanti et al. 1992;
impacts and Brabb 1984; Carrara and Merenda 1976; Cotecchia 1978;
c) economic aspects. Cruden and Varnes 1996; Hansen 1984; Hutchinson 1995;
It is important therefore to have a clear idea of the aims of Popescu 1998; Terzaghi 1950; Varnes et al. 1984) but the
the hazard assessment. Factors to be considered include effective use of such lists requires firstly effective data
whether it should be a qualitative or a quantitative assess- storage and secondly, appropriate selection of information
ment, whether the basis of assessment should be determin- useful for the particular analysis. This will depend on the
istic or probabilistic, the extent to which reliance should be type of problem under consideration (type of landslides
placed on historical records of instability, whether obser- present), the extent of the study area, the work-scale and
vational approaches should be part of continuing assess- the approach that will be adopted as well as the reliability
ment and to what extent the vulnerability of elements at and accuracy of the data.
risk and the impact of landsliding should be taken into
account. Data management: collection, storage and
There are essential pre-requisites to the development or selection
application of any method of hazard assessment. These The first step in every assessment consists of collecting all
include an understanding of geology, hydrogeology and available information and data on the study area. The
geomorphology from available maps, reports and field importance of accurate collection and storage of informa-
surveys, preliminary geological and geotechnical investiga- tion in the database is widely acknowledged (Amanti et al.
tion, access to historical information on landsliding, rain- 1996; Carrara 1982; Cruden and Brown 1992; Chowdhury
fall records and historical seismicity. Information on the and Flentje 1998a; Wieczorek 1984) but this is undoubtedly
extent of success or failure of any hazard assessment and one of the most burdensome operations in the task of
management strategies previously undertaken will also hazard assessment, regardless of the particular approach
prove to be useful in most cases. that is being adopted and the extent of the study area.
In developing a method of landslide hazard assessment it Some researchers estimate that the cost of data collection
is of fundamental importance to identify the causes of and management accounts for 70–80% of the total cost,
potential instability and landslides. Indeed, by determining including review and updating (Leroi 1996).
the factors that have caused landslides in the past it may be The achievement of the objective has been greatly
possible to forecast where and when future events could enhanced by the introduction of Geographical Information
occur. In other words, whatever methodology that is Systems (GIS).
adopted, the input data for a hazard assessment must only Two fundamental rules must be observed when creating a
be selected after carefully considering the causes of past database (Leroi 1996):
instability and the likely causes of potential future insta- a) the information must be homogeneous, i.e. it must
bility. Analysis of cause-effect relationships is not always have the same work-scale and the same geographic
easy however as a landslide is seldom linked to a single projection system
cause. It is often difficult to identify all the factors involved b) the database must be organised into basic mono-
and in many cases it is hard to define the relationships thematic layers, each of which contains homogeneous
between different causes. As will be detailed in later data.
sections of this paper, a bivariate analysis, for example, In addition, a database should include at least the
could lead to inaccuracies as two or more factors consid- following basic information (Chowdhury and Flentje
ered separately could lead to a different assessment than 1998a):
when they are examined together while the same factor a) a census of existing landslides including their nature,
could have different consequences depending on the type size, location and history
of process/mechanism that is considered relevant. b) a reliable site reference code
From the well-known limit equilibrium model used for c) any available information from previous site investiga-
geotechnical analysis, the factors likely to generate slope tions (aerial photo interpretation, laboratory testing,
instability can be divided into two groups: those which field analyses including back analyses of failures)
increase the shear stresses or disturbing forces and those d) any remedial or preventive measures installed and their
which reduce the shear strength or the stabilising forces. It effectiveness
is also helpful to distinguish between pre-disposing factors e) data from any installed instrumentation (inclinometers,
and triggering factors (an analysis of the latter was recently piezometers).
proposed by Wieczoreck 1996). Some authors distinguish The selection of data that could or should be utilised for
these factors from the so-called revealing factors (Vaunat the assessment of a given area depends essentially on:
et al. 1992). Siddle et al. (1991) proposed a bipartition a) the size of the study area
between extrinsic and intrinsic factors. Extrinsic factors b) the work-scale
may change over a historical time scale and are likely to be c) the technique adopted and
associated with the reactivation of old landslide areas. The d) the type of landslide.
intrinsic factors are usually related to features of the land- An important part of this process is the selection of reli-
scape and may be expected to change only within a able data from what may be available. This is facilitated by

Bull Eng Geol Env (1999) 58 : 21–44 7 Q Springer-Verlag 25


P. Aleotti 7 R. Chowdhury

examining the source of the data, the purpose of the Outline of various methods
original investigation and data collection and the form in
which data are available. In preparing GIS-based maps of of approaches to landslide
geology, Flentje (1998) obtained access to a number of susceptibility and hazard
borehole records from mining companies concerned with
coal exploration and production in the study area, public assessment
authorities concerned with development of transportation
links such as roads and railways and geotechnical Broadly speaking, methods of landslide hazard assessment
consulting firms. The quality of data was very variable, as may be qualitative or quantitative. The detailed classifica-
was the level of quantification and detail. Consequently, tion of assessment methodologies is somewhat subjective
careful judgement, supported by discussion with other and depends upon the choice of the aspect/s to be empha-
professionals, was required to identify data which were sised. A number of methods have now been developed as
unreliable. A great deal of the remaining information was summarised in Fig. 6.
found to be reliable and some information which was The different approaches can be classified on the basis of
initially considered to be less accurate was subsequently how the assessment area units are defined. It is possible to
included when other supporting information was found. differentiate between:
The whole of the data could then be further reviewed in the a) the adoption of regular grids of identical size and shape
light of validation processes which included reference to (square, rectangular, hexagonal and so on) (Carrara
existing maps and careful field checking. 1983, Anbalagan and Singh 1996)
b) the adoption of individual hillslopes or other landform
elements (morphological units) (Carrara et al. 1992;
Kienholz 1978) or
Data accuracy and reliability c) a consideration of units automatically derived from
overlays of each parameter map (Ives and Bovis 1978).
As stated above, the reliability and accuracy of data during Another classification is based on distinguishing between
collection and storage is very important. Nevertheless, its the different categories of hazard. Hartlen and Viberg
reliability and accuracy should subsequently be reviewed (1988) identify:
from time to time. Hazard assessment should be an on- a) relative hazard, where different slopes are compared
going process which can be updated as frequently as without expressing a precise value or quantification
required. Additional information often becomes available b) absolute hazard calculated either deterministically
with the discovery of new historical sources or as a result (safety factor) or on the basis of statistical analysis
of additional investigations. For example, data on
frequency and spatial distribution of past landsliding must
Field
always remain open to revision and expansion. The occur- geomorphological
rence of new instabilities or the reactivation of old land- analysis
slides may provide detailed information on the failure
Qualitative
mechanisms or further details of the relationship between methods
Combination or
rainfall and landsliding may become available. Similarly, overlay of index
maps
additional information will become available on the Use of index or
seismic response of slopes and escarpments and the devel- parameter maps Logical
opment of earthquake-induced landsliding. Every effort analytical M.
Assessment
methods

should be made to continually increase the accuracy and


Hazard

reliability of the data, checking the validity of assumed


mechanisms of failure and further refining the relation- Bivariate A.
ships between causative and influencing factors and on the Statistical
analysis
landsliding phenomena themselves. It will therefore be Multivariate A.
necessary to develop procedures and methods for updating
of data concerning both currently stable areas and areas Deterministic A.
which have already been subject to landsliding. Accurate (safety factor
Quantitative Geotechnical
information on geological details, geotechnical parameters engineering calculation)
methods
and pore water pressures is of paramount importance for approaches
detailed studies concerning individual sites. The target or Probabilistic
approaches
desirable level of detail and reliability concerning all this
information will be determined not only by the purpose
and importance of the project but also by the availability of Neural network
analysis
the financial and other resources to carry out the relevant (black box)
tasks.
Fig. 6
Proposed classification of landslide hazard assessment methods

26 Bull Eng Geol Env (1999) 58 : 21–44 7 Q Springer-Verlag


Landslide hazard assessment

c) empirical hazard based on the development of curves and hence do not require a basic conceptual knowledge
which relate several parameters such as slope height of these.
and angle
d) monitored hazard where deformation and/or rainfall
data are compared with predetermined critical or
threshold magnitudes considered to trigger land- Qualitative methodologies
sliding.
However, the most common classification is based on the In general qualitative approaches are based entirely on the
method by which landslide susceptibility and hazard are judgement of the person or persons carrying out the
assessed. Carrara (1982) distinguished between a geomor- susceptibility or hazard assessment. The input data are
phological and a quantitative/index approach. In the usually derived from assessment during field visits,
former, a geomorphological map, supplemented by addi- possibly supported by aerial photo interpretation. These
tional data, provides all the information needed to define methodologies, also defined as Expert Evaluation
the conditions of stability or the degree of instability in a Approaches (Leroi 1996), can be divided into two types:
given area. In the quantitative/index approach, base maps field geomorphological analysis and the combination or
(also called parameter maps or index maps) are overlain overlaying of index maps with or without weighting.
and suitably weighted and the area is divided into homoge-
neous zones. Subsequently Carrara (1988) introduced a Field geomorphological analysis
third group of methods, the so-called statistical approach This group is probably the simplest of the qualitative
in which statistical techniques are used to evaluate the methods. The assessment and/or zonation is carried out
influence of each contributory factor in producing insta- directly in the field by the earth scientist, based on his/her
bility. experience in other similar situations, with no clear indica-
From the beginning of the 80’s, the enormous potential of tion of any “rules” which have led to the assessment and/or
computer-based tools (Personal Computer and Geogra- zonation. In this case the stability maps are directly
phical Information System) enabled storage and proc- evolved from detailed geomorphological maps (Panizza
essing of a huge amounts of data through the use of 1975). The main disadvantages of such approaches are
complex techniques (multivariate statistics). The classifica- (Leroi 1996):
tion put forward by Hutchinson (1995) can be compared to a) the subjectivity in the selection of both the data and the
that proposed by previous authors, in which the statistical rules that govern the stability of slopes or the hazard of
approach was defined as a direct comparison technique instability. This fact makes it difficult to compare land-
based on the relationship between various fundamental slide hazard maps produced by different investigators
maps (usually lithology and slope inclination) with land- or experts
slide distribution maps. b) use of implicit rather than explicit rules hinders the
Some conceptual differences can be recognised in the clas- critical analysis of results and makes it difficult to
sification suggested by Crozier (1992). He proposed the update the assessment as new data become available
distinction between inductive, deductive and geotechnical c) lengthy field surveys are required.
approaches, although the difference between inductive and Nevertheless, they allow a rapid assessment of stability in a
deductive methods is not always clear. In essence, it given area, taking into consideration a very large number
appears that with the deductive approach the selection of of factors. They can be used successfully at any scale, and if
parameters and the choice of their relative weighting is necessary, adapted to specific local requirements. The field
determined empirically (post-event), whilst in the deduc- geomorphological analysis does not require the use of a
tive approaches such determinations are made on the basis Geographical Information System which, in this case, is
of experience and intuition (pre-event). simply a drawing tool.
In this paper a new classification is proposed (Fig. 6), Examples of studies using these methods are very frequent
based on those of Carrara (1982), Hansen (1984), Leroi in literature of the 70s and 80s (Bosi et al. 1982; Carrara
(1996), Soeters and Van Western (1996). The main varia- and Merenda 1976; Fenti et al. 1979; Guerricchio and Meli-
tions consist in doro 1979; Kienholz 1978; Ives and Messerli 1981; Rupke et
a) the introduction of the so-called logical-analytical al. 1988). Brugner and Valdinucci (1972) propose a stability
models, where susceptibility is expressed by an equa- map at a scale of 1 : 50.000 in which landslide susceptibility
tion in which the independent variables are tied is divided into five classes. At the two ends of the spectrum
together by various weights, depending upon the of classes or categories are the stable areas and active land-
influence they exert on the dependent variable (slope slides, between which are three categories of varying
deformation) (Bughi et al. 1996) and susceptibility coloured in different tones of yellow. The
b) the introduction of new techniques that utilise neural legend indicates the factors that have influenced the alloca-
networks (Aleotti et al. 1996c 1998b; Mayoraz et al. tion of an area within a given category: the inclination of
1996). These techniques (neural networks) do not fall slope; the soil stratification-slope angle; the extent of frac-
entirely under any traditional assessment technique. turing in lithotypes and the morphological regularity.
The distinctive feature of such methods is that they do Perhaps one of the most comprehensive projects reported
not depend upon the physical aspects of the problem in the literature is the French ZERMOS procedure (Antoine

Bull Eng Geol Env (1999) 58 : 21–44 7 Q Springer-Verlag 27


P. Aleotti 7 R. Chowdhury

1977; Humbert 1977; Meneroud and Calvino 1976) which Amadesi and Vianello (1978) proposed a scheme involving
involves two main phases: analysis and extrapolation. In several steps in which the initial susceptibility map,
the first phase, all the factors which may influence the obtained by overlaying the lithological maps with the slope
stability are examined, both permanent (topography, maps and the structural maps, is progressively refined by
geology, hydrogeology, hydrology, etc.) and temporary considering first the land use map and then the geomor-
(climate, land use and other man-made factors). Active phological map (Fig. 7). The most interesting aspect lies in
and/or inactive landslides may be analysed. In the next the last phase when the second stability map is compared
phase all the factors are extrapolated by the author to areas with the geomorphological map. No numerical weighted
with similar physical conditions, thus enabling zonation of value is assigned to the geomorphological data and the
the area into three sections with varying degrees of hazard operator must quantify its influence on stability based on
(“risque”): his/her personal experience. This step introduces a similar
a) null or low hazard, areas in which no instability should element of subjectivity as is found with the direct geomor-
occur phological approach.
b) potential or uncertain hazard, areas with potential To some extent these limitations have been overcome by
instability of uncertain nature and extent recent techniques such as that proposed by Anbalagan
c) ascertained hazard, areas with declared instability and (1992), Abu-zeid et al (1994) Anbalagan and Singh (1996)
certain threat of failure. Gupta and Anbalagan (1997) Visintainer and Turrini
The hazard calculated is of a relative nature and the (1995), at scales in the range 1 : 25.000–1 : 50.000. The
authors acknowledge that the various hazard categories hazard (TEHD) is calculated by breaking down the study
cannot be compared from one area to another. area into facets delimited by ridges, spurs, gullies and
Hearn (1995) has recently proposed mapping at a scale of rivers, followed by the overlay of index maps. However,
1 : 10.000 based on a hazard classification sheet compiled with the scales adopted and for large areas, it appears diffi-
directly in the field for every geomorphological unit (Gus) cult to apply the correction factor for the weathering of the
identified previously. An interesting example of a suscepti- lithotypes and to accurately determine the structural
bility map based on field analysis is reported by Carmassi discontinuity-slope angle. Moreover, the criterion used to
et al. (1992) who used this method to identify the most divide the area determines the presence of various-sized
favourable zones for the construction of power plants. cells that could contain inhomogeneous parameters. This
could be overcome by attributing to each parameter in a
Synthesis based on overlay or combination of given cell, a value corresponding to the weighted mean of
index maps or parameter maps with weights the partial contributions, or more conservatively, the
In this approach, the expert selects and maps the factors highest value among those present (Abu-zeid et al. 1994).
that affect slope stability and, based on personal experi- In the same line of research, there are the so-called logical
ence, assigns to each a weighted value that is proportionate analytical models which have the same conceptual back-
to its expected relative contribution in generating failure. ground but a different scale of application. Figure 8 shows
The following operations should be carried out (Soeters the basis of a logical analytical model. The first stage
and Van Westen 1996): consists of a tentatively proposed relationship which links
a) subdivision of each parameter into a number of rele- some experience-based weighted factors. Using the
vant classes proposed relationship it is possible to predict the slope
b) attribution of a weighted value to each class displacements of some landslides and, by comparing the
c) attribution of weighted values to each of the parame- results with the monitoring data, to define the degree of
ters
d) overlay mapping of the weighted maps and Operations
e) development of the final map showing hazard classes. Lithological (1) overlay and sum
The advantages of such a methodological approach are map (2) overlay and sum or subtraction
that it considerably reduces the problem of the hidden (3) expert qualitative analysis
rules and enables total automation of the operations listed
above through appropriate use of Geographical Informa- (1) First
tion Systems. Furthermore, it enables the standardisation Slope map stability
map
of data management techniques, from acquisition through Second
(2)
to final analysis. This technique can be applied at any stability
scale. The major disadvantage is the lengthy operations map
Structural (3) Final
involved, especially where large areas are concerned. The Land use map hazard
map
problem of subjectivity in attributing weighted values to map
each parameter and to the different factors also remains, as Geomorphic
well as the difficulty of extrapolating a model developed in map
a particular area to other sites or zones (Carrara 1983).
An early and well-known example of this methodology is Fig. 7
that reported by Stevenson (1977) who developed a similar Schematic representation of the method proposed by Amadesi
approach for hazard zonation for an area in Tasmania. and Vianello (1978)

28 Bull Eng Geol Env (1999) 58 : 21–44 7 Q Springer-Verlag


Landslide hazard assessment

Physical by gas pipelines (Bughi et al. 1996). The field surveys


knowledge enabled the recognition of a number of factors, including
Monitoring the different types of landslides; translational slides and
Relationship and
rotational slides being the most frequent) and the litho-
weights selection types involved – usually clay and silt. The subsurface
instrumented data refer mainly to inclinometric and
piezometric measurements. From Fig. 9 it may be noted
Calibration

Slope that the depth of the slip surface ranges between 1 and
displacements 16 m and that the rate of movement is extremely slow or
very slow.
General
poor Degree good
application and
of agreement
hazard
classification Quantitative methodologies
Fig. 8 Statistical analyses
Example of a logical analytical model
The attribution of weighted values on a subjective basis to
the numerous factors that govern slope stability represents
agreement. At this stage the established relationship can be the main limitation in all the methods described above.
calibrated by varying the weights of the elements. Once the The solution to this problem could be to adopt a statistical
degree of agreement is good, the relationship can be used approach that compares the spatial distribution of land-
to classify all the failures, even those for which subsurface slides with the parameters that are being considered. The
instrumented data are not available. results could then be applied to areas currently free of
A logical analytical model has been used to predict the landslides but where conditions may exist for susceptibility
slope displacement in geologically unstable areas crossed to future instability. The major difficulty consists in estab-
lishing the slope failure processes and in systematically
identifying and assessing the different factors related to
Fig. 9 landsliding (Carrara 1988). One of the principal advantages
General characteristics of instability or movements related to is that the investigator can validate the importance of each
pipe-slope interaction in some geologically unstable areas within factor and decide on the final input maps in an interactive
Italy (modified after Scarpelli et al. 1995) manner. The use of G.I.S. makes these operations much

40 60
FA Fall SA Sandstone
CM Complex CO Conglomerate
30 FW Flow MA Marls
LI Limestone
Percentage (%)

Percentage (%)

TS Translational slide 40
RS Rotational slide GR Gravel
TO Topple SD Sand
20 ST Silt
CL Clay
20
10

0 0
FA CM FW TS RS TO SA CO MA LI GR SD ST CL
Landslide type Lithology

60 60
Percentage (%)
Percentage (%)

40 40

20 20

0 0
0.05 0.05 - 0.10 0.10 - 0.20 > 0.20 <3 3-6 6-9 9 - 12 12 - 15 > 15
Rate of movements (mm/day) Depth of slip surface (m)

Bull Eng Geol Env (1999) 58 : 21–44 7 Q Springer-Verlag 29


P. Aleotti 7 R. Chowdhury

easier and to a large extent explains the increasing “popu- The bivariate statistical approach has been successfully
larity” of the statistical approach which closely parallels the employed by researchers mapping hazard of surficial fail-
ever-increasing application of G.I.S. techniques. ures. In some cases, the principal pre-disposing factor was
Statistical analyses can be either bivariate or multivariate. again found to be slope steepness. This parameter was
analysed by several authors together with the sharpness of
ridge crest (Ellen et al. 1982), soil type and vegetation
Bivariate statistical analyses covers (Hastie 1990) and slope aspect (Aleotti et al. 1996b).
In bivariate statistical analyses each individual factor is Aleotti et al. (1996b 1998a) applied such an approach for
compared to the landslide map. The weighted value of the the evaluation of landslide hazard in a region in Northwest
classes used to categorise every parameter is determined Italy. In 1994, an unprecedented rainstorm triggered many
on the basis of landslide density in each individual class. surficial landslides, mainly soil slips and soil slip-debris
The following operations are required: flows, causing 20 fatalities (Polloni et al. 1996). From a
a) selection and mapping of significant parameters and statistical analysis it was possible to determine the critical
their categorisation into a number of relevant classes geometric conditions for these landslides in terms of
b) landslide mapping morphological context, slope angle and other factors.
c) overlay mapping of the landslide map with each param- Aleotti et al then calculated the critical depth of failure as a
eter map function of both the slope angle and the internal friction
d) determination of density of landslides in each param- angle. The values obtained were in good agreement with
eter class and definition of weighted values the field observations. Following this, they used the Green
e) assignment of weighting values to the various param- and Ampt model (Green and Ampt 1911, Pradel and Raad
eter maps 1993), to estimate the minimum intensity/duration rainfall
f) final overlay mapping and calculation of the final values capable of saturating soil at different depths. Having
hazard or susceptibility value of each identified land completed the examination of the boundary conditions, a
unit. susceptibility map was prepared by overlaying the pre-
Although the bivariate statistical analysis is considered to disposing factor maps. An analysis of the historical rainfall
be a quantitative approach to landslide hazard assessment, data should then be added. In fact it is possible to obtain
a certain degree of subjectivity exists, particularly in step 5. the permeability threshold values as a function of precipi-
In addition, it must be appreciated that in many situations, tation with different return times. This corresponds to a
the analysed factors are not independent and may show probabilistic assessment of the triggering cause. Hence, in
either high or low correlation (Leroi 1996). addition to susceptibility, a hazard assessment is also
The most simple models include the determination of a possible (Fig. 10).
threshold value of slope angle or inclination, above which
the site or area is considered unstable in relation to the Multivariate statistical analyses
presence of a given lithotype (Brabb et al. 1972; Nielsen et Although the multivariate statistical approach had already
al. 1979; Pomeroy 1979). The models are based on a series been successfully applied in several areas of applied
of empirical curves (Hartlen and Viberg 1988) that relate geology, such as petroleum exploration, the application of
the two or more significant parameters such as slope this technique to landslide hazard evaluation began at a
height and slope angle for failed slopes (Viberg 1984). As later date (Neuland 1976; Carrara et al. 1978). Indeed, it
an example, Lucini (1979) records that in research was not until the 80s that detailed analyses on this basis
conducted in Southern Italy, in the area where the “Argille began, following the work accomplished by Carrara
Varicolori” clays outcrop, it was noted that all the land- (1983).
slides occur in zones with slope inclinations higher than The procedure involves several preliminary steps which are
15%. As a consequence, it was considered appropriate to undertaken in a test area. Once the results achieved have
distinguish only two hazard classes for this study area. been verified (see step 6 below), they are extended to the
The bivariate statistical approach is widely employed by entire area under examination. The following steps are
the earth scientists and numerous parameters may be required:
taken into consideration: lithology, slope angle, slope 1. Classification of the study area into land units.
height, land use (Bortolami et al. 1994), distance from In the first applications (Carrara et al. 1978; Carrara 1983
major structures, drainage density (Meherotra et al. 1994), 1988) the land units were square with sides of varying
relief morphology (Sorriso Valvo et al. 1996), closeness of dimensions (usually 100 or 200 m), depending upon the
the facet to a river (Choubey et al. 1992), attitude of litho- extent of knowledge of the area. Subsequently they have
types (Guida et al. 1979). Kingsbury et al. (1992) based been defined on a morphometric basis (Carrara et al. 1991
their hazard classification and mapping entirely on slope 1992);
inclination as they found that additional factors (soil type, 2. Identification of significant factors and creation of input
land-use, slope aspect, proximity to water courses) did not maps
improve the final result. Such a conclusion can only be The input variables include information concerning the
applicable to a particular area or region and should not be landslides (type, degree of activity, etc.) and geo-refer-
extended to another area or region without a thorough encing. Several attributes are automatically derived from
analysis which is region- or area-specific. statistical operations performed on the basic parameters

30 Bull Eng Geol Env (1999) 58 : 21–44 7 Q Springer-Verlag


Landslide hazard assessment

Analysis of the In general, the decision to carry out the analysis on two
Statistical analysis
of the landslides Stabilty analysis historical groups only (unstable and stable land units) simplifies the
(bivariate) meteorologic data problem from a statistical viewpoint but hinders the iden-
set
tification of various combinations of factors related to
different hazard types;
Calculation of the
5. Combination of the parameter maps with the land unit
Determination of the Definition of map and creation of an absence/presence matrix of a given
permeability
geometrical critical meteorologic critical
conditions
thresholds and of
thresholds class of a given parameter within each land unit;
the critical depth
6. Multivariate statistical analysis:
Given the large amount of data, this analysis should be
carried out using specific software. Fortunately these pack-
ages are now included as part of the G.I.S. software
Knowledge of the
programs. The statistical analyses most frequently used are
Knowledge of the
factors of triggering cause
discriminant analysis (Baeza and Corominas 1996; Carrara
predisposition 1983; Carrara et al. 1990) or regressive multiple analysis
(Bernknopf et al. 1988; Jade and Sarkar 1993; Wieczorek et
al. 1996) which are often employed in parallel within the
same project (Baldi et al. 1990; Carrara 1983). It is prefer-
Susceptibility map able to apply discriminant analysis (stepwise or canonical
Soil slip hazard discriminant analysis) with continuous variables, while the
assessment
regressive analysis can be used even with nominal varia-
bles (Clerici and Dall’Olio 1995);
Fig. 10 7. Reclassification of land units based on the results
Flow chart for surficial failures susceptibility map (after Aleotti et achieved in the previous phase and their classification into
al. 1996b and 1998a) susceptibility classes.
In the discriminant analysis for example, the inspection of
the standardised discriminant coefficient allows the contri-
bution of various parameters in causing slope instability to
(mean, standard deviation, maximum or minimum be quantified and, as a result, enables objective reclassifi-
values). An important aspect is the conversion of various cation of the study area. By transforming the classification
parameters from nominal to numeric, eg, rock composi- function scores into probabilities, the susceptibility map
tion or vegetation cover. This can be done through the can then be converted into a hazard map (Carrara 1983);
creation of dummy variables or by coding and ranking the
classes based on the relative percentage of the area affected
by landsliding. The two methods are similar but the latter
is to be preferred for two reasons. It avoids the creation of Geotechnical models –
an excessively high number of dummy variables and allows
consideration of the so-called “previous knowledge” of deterministic and probabilistic
landslide susceptibility (Carrara 1983);
3. Construction of a landslide map Deterministic framework
In this phase the collection (on special sheets) and/or This type of approach involves analysing specific sites or
storage of all information concerning landslides (Carrara slopes in engineering terms. The main physical properties
1982) into specific databases is very useful. In fact, if syste- are quantified and applied to specific mathematical models
matic, complete and homogeneous databases are and the safety factor is calculated. These models (mono-.
constructed at world-wide level, they could be applied to bi- and tri-dimensional) are commonly used in soil engi-
other regions having similar characteristics. Unfortunately, neering for slope-specific stability studies. The approach is
to date such nation-wide databases are rare (Leroi 1996; widely employed in civil engineering and engineering
Cruden and Brown 1992); geology and has been applied to landslide hazard assess-
4. Identification of the percentage of landslide-affected ment and mapping, especially after the introduction of
areas in every land unit and their classification into un- G.I.S. Accuracy and reliability is improved as detailed
stable and stable units. knowledge of the area of application increases. A deter-
The threshold value of this classification is fixed every time ministic approach was traditionally considered to be suffi-
on the basis of two requirements: cient for both homogenous and non-homogenous slopes.
(a) in areas with high landslide density, the threshold The index of stability is the well known safety factor, based
should be based on relatively high percentages in order on the appropriate geotechnical model. The calculation of
to achieve two statistically representative groups the safety factor, F, requires geometrical data, data on
(b) however, even a relatively low landslide density must shear strength parameters (cohesion and angle of internal
be taken into consideration as it could represent a risk friction) and information on pore water pressures. More-
for human activities (Carrara 1983). over, decisions must be made on whether to use peak shear

Bull Eng Geol Env (1999) 58 : 21–44 7 Q Springer-Verlag 31


P. Aleotti 7 R. Chowdhury

strength values or residual shear strength values (or values the aspect map and the elevation map. Based on the anal-
in between) for specific parts of the slip surface (Chowd- ysis of the physical characteristics of the geological units, a
hury 1976; Chowdhury and Bertoldi 1977). For these geotechnical map was produced. Finally, the geomorpho-
reasons such methods are normally applied only in small logical map provided the location of both active and
areas and at detailed scales. Calculations of the safety dormant landslides. For each lithological unit the critical
factor must be made for each individual slope or area geometrical conditions were determined by means of pseu-
before a hazard map can be prepared. dostatic stability analyses (input matrix). The final suscep-
The use of G.I.S. facilitates the simulation of multiple scen- tibility map (at a scale of 1 : 10,000) was derived by overlap-
arios based on variable factor hypotheses (usually the trig- ping the geotechnical map, the landslide map, the slope
gering factors), as well as the construction of reliable map and the relief map (Fig. 12).
hazard maps. The studied area can be divided in six classes of suscepti-
A summary of deterministic approaches for susceptibility bility, but without rigid distinctions. Flat areas, obviously,
and hazard assessment is provided in Fig. 11 which shows will not have stability problems. In areas with rock
the procedures of type (A), (B) and (C) discussed below. outcrops, stability must be verified by taking into account
the rock characteristics (fracturing, stratification and
Safety factor as a basis for susceptibility and hazard weathering). In stable slopes it is necessary to verify the
assessment impact produced by earth removal or overloading, while in
In the simplest model (A), the results obtained from the moderately stable slopes the work must be designed on the
stability analyses for various combinations of parameters basis of geotechnical investigations and slope stabilisation
are qualitatively re-converted into various degrees of measures will sometimes be necessary. In potentially unst-
susceptibility and spatialized using a standard G.I.S. able slopes these measures are essential. Finally for
overlay mapping procedure to achieve a final assessment existing landslides, one should have planning solutions
map (Fig. 11). An example of this kind has been recently which will not adversely affect these areas. Moreover, it
proposed by Gokceoglu and Aksoy (1996). In general this should be noted that the feasibility of any preventive or
approach can be defined as semi-quantitative. remedial measures depends upon the geometry and kine-
Baldelli et al (1996) applied this type of technique during matics of movements.
the Environmental Impact Assessment of the Messina By adding some variable data (for example the depth of the
Straits Crossing Project. This consists of three submerged piezometric line or an earthquake of given magnitude) the
floating tunnels anchored to the sea floor running at a previous susceptibility map can be modified to reflect
depth of 50 m below sea level. The base maps are the different scenarios (B-type models in Fig. 11). Finally, by
Digital Terrain Model (DTM), the geological map and the introducing a probabilistic analysis of the input data a
geomorphological map. The DTM provided the slope map, hazard map can be produced (C-type models). In the more

Fig. 11 A B C
Summary of deterministic
methods for susceptibility and Varying Varying
hazard assessment Fixed input (*) Fixed input Fixed input
input (*1) input (*2)

Profile Profile Profile Probab.


selection selection selection analysis

Stability analysis Stability analysis Stability analysis

Spatialization Spatialization Spatialization

Susceptibility Susceptibility map


Hazard
map for different
map
scenarios

(*) precautionary (*1) triggering factors (*2) triggering and predisposing


(conservative) values (groundwater level, earthquake, etc.) factors

32 Bull Eng Geol Env (1999) 58 : 21–44 7 Q Springer-Verlag


Landslide hazard assessment

R1 – the testing equipment and methods are not perfect.


R2 In addition to real parameter variability and systematic
R3 errors, there are often significant uncertainties associated
Rn with geotechnical models. Finally uncertainties are also
associated with the mechanisms of failure, occurrence of
γ β α failure and its impact.
unit 1, 2, ... n The recognition of uncertainties has led to the develop-
Stability ment of methods of analysis within a probabilistic frame-
analysis work while maintaining the basic geotechnical models. The
Input matrix Unit n probability of failure is defined as the probability that the
Unit 2 performance function has a value below the threshold
Unit 1 value. Considering the factor of safety (F) as the perform-
Landslides map Relative relief (R) ance function, the threshold value is 1 and probability of
Lithological map SF SF SF SF
failure (pf) may, therefore, be defined as:
inclination
Slope

SF SF SF SF
D.T.M. pfpP[F~1]
SF SF SF SF
• Slope map
SF SF SF SF
• Relative relief map The probability of success or the reliability (ps) is therefore
the complement of pf. Hence:
Overmapping
Susceptibility map psp1Ppf
In order to calculate the probability of failure, the proba-
Fig. 12
Synthetic representation of the semi-quantitative approach
bility density function of the performance function is
proposed by Baldelli et al. 1996 during the Environmental Impact required. Thus it is recognised that the factor of safety (F)
Assessment for the Messina Straits Crossing Project is not a single-valued function. Its probability distribution
may be characterised by means of at least two statistical
parameters, the mean or central value F and the standard
deviation sF.
complex models the safety factor becomes a random vari- It is often useful to define a reliability index b which
able, a function of the probabilistic distribution of parame- combines the mean and standard deviation of the perform-
ters which, in the mathematical model adopted, contribute ance function. Thus, if F is the performance function:
to defining this factor (Berggren et al. 1992, Chowdhury
1984, Genevois and Tecca 1987, Hammond et al. 1992, FP1
bp
Mulder and Van Asch 1988, Ward et al. 1982, Christian et sF
al. 1992, Tang et al. 1976, Wu and Kraft 1970, Wu and
Abdel Latif 1994). The change from A-type to C-type The numerator gives the extent to which the average value
models corresponds to the transition from deterministic is above the threshold value and the denominator reflects
methodologies to probabilistic approaches. the dispersion from this average value.

Probabilistic approaches Calculation Methods


For decades, geotechnical modelling and analysis within a Three commonly used methods of probability calculation
deterministic framework has facilitated the quantification are:
of safety or reliability. However, performance indicators (i) First Order Second Moment Method (FOSM)
such as the factor of safety, F, do not take into considera- (ii) Point Estimate Method
tion the variability of geotechnical material parameters (iii) Monte Carlo Simulation Method
such as cohesion (c), angle of internal friction (f) and the Brief descriptions of these methods and their application
undrained shear strength (su) some of which may also vary to slope stability may be found in Chowdhury (1984) and
in magnitude with time. The spatial and temporal varia- Nguyen and Chowdhury (1984, 1985).
bility of pore water pressures is again very important but is
not reflected in the calculated values of the conventional System reliability, updating, progressive failure
factor of safety. The probabilistic framework for reliability analysis can
There is now an increasing recognition of the importance offer much more than the replacement of the conventional
of uncertainties in geotechnical engineering. Parameter safety factor (F) by the probability failure (pf) or the relia-
variability is only one example; there are also systematic bility index (b).
uncertainties which arise from the fact that: A number of approaches have been developed in recent
– a soil mass can only be investigated at a finite number of years which do not have a direct counterpart within a
points deterministic framework. These include methods for the
– the number of field and laboratory tests conducted to assessment of the reliability of a geotechnical system
determine soil parameters is limited by financial and consisting of a number of elements or a number of
time constraints possible modes of failure or both. For example, the relia-

Bull Eng Geol Env (1999) 58 : 21–44 7 Q Springer-Verlag 33


P. Aleotti 7 R. Chowdhury

bility assessment of embankment or natural slope stability combined with other factors, including the stated or
is enhanced by including multiple slip surfaces rather than implied frequencies of occurrence of the time-dependent
by relying only on the conventional critical slip surface parameters, such approaches can facilitate hazard assess-
(Chowdhury and Xu 1994a, 1995). Often reliability is over- ment on a quantitative or semi-quantitative basis. Calcu-
estimated in complex systems if only a single mode of lated probabilities of failure based on a geotechnical model
failure is taken into consideration even though that mode may be regarded as conditional probabilities as these
may be the critical mode in conventional terms. calculations are valid for the chosen values of geotechnical
It is widely recognised that the initial assessment of material parameters (cb, fb, g) and for a chosen distribu-
geotechnical properties and parameters may not be accu- tions of pore water pressures along the relevant potential
rate. In particular, systematic uncertainties influence the slip surfaces. The frequency of occurrence of the assumed
results of reliability calculation. Therefore, updating of conditions should be estimated using other appropriate
reliability with the availability of additional information is methods. For example, selected pore water pressures
very important (Zhang 1990, Chowdhury 1998c). A proba- would relate to specified seepage conditions or ground
bilistic framework facilitates this. The Bayesian approach water levels which may correspond to rainstorms of a
has been used by Chowdhury and Zhang (1989) to update particular return period or frequency. Unless analyses
the reliability of a deep excavation considering perform- concerning intensity, duration and frequencies of rainfalls
ance during the process of construction/excavation. have been carried out, the temporal aspect of rainfall-
Expert judgements can also be incorporated in the induced landslide hazard will not be clear even if a fully
updating of reliability assessments within a Bayesian prob- quantitative geotechnical analysis has been performed
abilistic framework. within either a deterministic or a probabilistic frame-
Modelling of progressive failure within a soil mass or a work.
geotechnical system is very important given that the distri- The application of a geotechnical model for quantitative
bution of stress and strain is often non-uniform and that analysis requires that reliable values of the material param-
soils are often strain-softening. Innovative approaches eters be available. The purpose of a quantitative approach
developed for simulation of failure progression within a may be defeated if such data are not available. This aspect
probabilistic framework have no counterpart in determin- is even more important from the perspective of probabil-
istic soil mechanics (Chowdhury and Grivas 1985, Chowd- istic analyses. For such analyses, the data must be more
hury and Tang 1987, Chowdhury et al. 1987). By combining detailed or numerous than those for deterministic
probabilistic and observational approaches, powerful analyses. Each parameter must be represented by a proba-
methods can be developed for updating reliability and for bility distribution or at least by two statistical parameters,
decision-making (Chowdhury 1992, Chowdhury and Zhang the mean and the standard deviation. In the absence of
1993, Chowdhury 1994). detailed information, when resources are not available to
In order to make decisions based not only on performance obtain the required data or when there are time
indicators but on the consequences of failure or of inade- constraints, a subjective probability assessment may be
quate performance, it is again necessary to work within a made or it may be necessary to rely on expert judgements
probabilistic framework. Amongst the techniques available as to the probability values to be used for assessment of
are those which use event trees for consequence-based risk landslide susceptibility, hazard or risk.
analysis. Event trees are often used for consequence-based
risk analysis of dams, a recent example being BC Hydro Importance of correct mechanism or model
(1993). Extension to landslide hazard and risk studies Assuming that sufficient information on geotechnical
could be useful. parameters is available it is important that appropriate
failure mechanisms relevant to the study area are iden-
Application to landslide hazard analysis tified and that geotechnical models are selected which
The availability of geographical information systems correspond as closely as possible to these mechanisms.
(G.I.S.) can facilitate the use of a deterministic or a proba- Unfortunately there has been an increasing tendency to
bilistic geotechnical approach as part of a methodology of adopt an ‘infinite slope’ model which is well known as the
landslide hazard assessment. For example, after subdivi- ‘one-dimensional model’ representing potential failure of
sion of an area or region into elements or small areas, the long natural slopes along slip surfaces parallel to the
factor of safety of individual sloping areas may be ground surface. Such a model may be justified for shallow
computed and then mapped. Alternatively, using a proba- slope failures and also for some parts of a study area
bilistic framework, the probability of failure of individual involving deep-seated failures. However, it is not advisable
sloping areas could be computed and then mapped. to use such a model for areas with a variety of failure
Depending on the adopted hazard assessment approach, mechanisms. The use of a G.I.S.-based approach for accu-
this information could be used on its own or combined rate mapping cannot compensate for grossly inaccurate
with other factors or factor maps to produce susceptibility estimates of failure probability due to the selection of
maps and/or hazard maps. geotechnical models corresponding to inappropriate
The calculation of values of the safety factor (F) or the failure mechanisms. Recent studies of slope displacements
probability of failure (pf) may be regarded as part of the under seismic conditions (Crespellani et al. 1998) have
process of quantitative susceptibility assessment. When provided solid support for using, as far as possible, only

34 Bull Eng Geol Env (1999) 58 : 21–44 7 Q Springer-Verlag


Landslide hazard assessment

the realistic failure mechanism as a basis for mathematical alter their initial state, something no conventional algo-
and numerical analyses. This important point has also rithm does: they can learn” (Lees 1996).
been highlighted by Chowdhury (1998b).
The most common configuration of the neural network is
composed of three layers. In the feed-forward network, the
The performance function in terms of layers are completely connected to one another. Figure 13
deformations is a schematical representation of an “element of calculus”
It is also important to consider the selection of the that receives the weighted input from its connections. The
performance function itself. In many instances, probabili- procedure consists, firstly, of attributing the weights to the
ties of failure or non-performance may have to be based on connections in a casual manner and of choosing the input
different criteria. For example, there are many applica- parameters (Fig. 14). The calculated output is compared to
tions, such as high speed railways through hilly areas, in that expected and the error is determined. The procedure
which deformations must be kept low. Satisfactory progresses in an iterative manner until convergence of the
performance in such cases should be specified in terms of calculated and expected output is reached. This is the
lateral and vertical slope deformations below a tolerable learning phase, in which the function of the neural
limit, which would be the basis of the pf. In such circum- network is created.
stances, although a complete or catastrophic slope failure Aleotti et al. (1996c, 1998b) have applied this kind of tech-
is unlikely, a sloping area may nevertheless have a high nology to divide some low-speed landslides in Northern
hazard because of deformation levels which constitute Italy according to their degree of hazard. The aim was to
non-performance or failure in terms of that particular establish the priority for the management of landslides or
application. For a road or highway embankment, the same potential landslides.
deformation levels may not constitute a high or even a The procedure consists of five stages:
moderate hazard. 1. Field analysis
For earthquake induced landsliding, the use of appropriate A special field survey report is drawn up for each landslide
deformation analyses is always preferable to pseudo-static describing the characteristics of every examined failure
analyses. Thus a sliding block model (such as that (geometry, kinematics, activity) and the general physical
proposed by Newmark 1965) may be used as the basis for a setting (geology, geomorphology, hydrology, hydrogeo-
deterministic approach (Chowdhury and Xu 1994b, Cres- logy, soil, etc.). In addition to providing low-cost detailed
pellani et al. 1998, Chowdhury 1997, Chowdhury 1998a, information, these reports allow the coding and standar-
Sarma 1975, Sarma and Jennings 1980). Categories of disation of available data;
hazard would correspond to different categories of esti- 2. Laboratory preparatory phase
mated deformation. The same type of model may be used The number of parameters to be considered is reduced and
for a probabilistic approach. The performance function a numeric index assigned to each parameter. For the non-
would be formulated in terms of the permanent deforma- numeric variables an index is defined to express the rela-
tion of a sliding block subjected to a particular ground
motion or a particular design earthquake. Based on that
performance function, values of the probability of failure
(pf) would then be calculated. The use of sophisticated x0
wj0
numerical methods for the siesmic response of escarp-
ments has also been explored by a number of research x1 wj1
workers as discussed by, for example Chowdhury and Sum fTRASF Output
Tabesh (1998). x2
wj2

wjn
xn
Use of neural networks
Carrara (1983) in his classification of box models distin- Output layer
guished between:
– white box models – based on physical analyses Hidden layer
– grey box models – only partly based on physical
analyses
– black box models – not based on physical analyses Input layer
Neural networks analysis must be considered as a black
box model. Lees (1996) describes it as follows:
Data
“Neural network is a processing device, implemented as Fig. 13
an algorithm or in hardware, whose design was moti- Neural network geometry; the network used by Aleotti et al.
vated by the design and function of mammalian brains; 1996c 1998b consists of three layers (input, hidden and output
they react to training data input in such a way as to layers)

Bull Eng Geol Env (1999) 58 : 21–44 7 Q Springer-Verlag 35


P. Aleotti 7 R. Chowdhury

of the test set have not been entered directly into the
training procedure of the neural network, they represent a
test area that closely resembles the actual conditions in
which the evaluator will be used.
Even considering the sites of the test set it is possible to
observe that the neural output closely approaches the
assessments provided by the experts (Fig. 15). This denotes
that the neural network was capable of apprehending,
based on the examples of the learning set, an assessment
function that proved to be highly effective in the sites of
the test set.
In conclusion, it may be observed that neural networks
could represent an effective approach when dealing with
landslide hazard assessments that would be difficult to
achieve by means of standard mathematical models.
Fig. 14
The neural network procedure consists of 5 stages. This proce-
dure progresses until convergence of the calculated and expected
output
Acceptable risk of landsliding
tive physical phenomena (Table 1). For example for the
factor “landslide visibility” which is how a landslide is Importance of location and consequences of
recognisable in the field, an index equal to 0.25 may be failure
assigned for the class “poor” and an index equal to 1 for Landslide hazard is the most important component of risk
the class “excellent”. associated with landsliding and, therefore, it is most
3. Preliminary assessment important to consider concepts and criteria concerning
For every landslide, three experts with extensive knowl- acceptable risk. The traditional view was that no level of
edge of the examined area provide a separate hazard risk, however small, should be tolerated for major projects
assessment based on a numerical scale. In addition, the such as large dams, nuclear power plants, major buildings
same experts provide a “blindfold” assessment based on and bridges and that even for relatively small projects, a
the values derived from the seven examined parameters level of risk resulting in loss of life would be regarded as
(hydrological features, landslide typology, landslide visi- unacceptable. However, it is now widely recognised that
bility, length, breadth, depth, speed) without disclosing the some degree of hazard and, therefore, risk, however small,
name of the site which they refer to. Then a mean value is is unavoidable in any project. Risk can often but not
calculated from the 6 assessments provided for each area. always, be reduced to a very low or extremely low levels
4. Application of neural network to the learning set but can never be eliminated.
The landslides are divided randomly into two groups, the The main principle which applies to slopes and land-
first is called the “learning set”, the second group is called sliding, and indeed to other structures or projects, is that
the “test set”. The neural evaluator is obtained by training the level of risk which is tolerated or accepted should be
the neural network on the sites of the learning set: the inversely proportional to the consequences of failure
function obtained in this learning phase is very close to the (Whitman 1984).
assessments provided by the experts on landslides of the A probability of failure of 10 –4 is often considered reason-
learning set. able for dams and the US Department of Energy considers
5. Application of neural evaluator to the test set 10 –5 as the target mean annual probability of exceedance
The effectiveness of the neural evaluator obtained in this for civilian nuclear power plants.
training phase can be verified only by applying it to the It is useful to quote here from Peck (1980) on the accept-
landslides that compose the test set; in fact, since the sites able risk for dams.

Table 1
Non-numeric parameters indices

Numeric index

Parameter 0.25 0.5 0.75 0.8 1

Hydrologic features* none (i) (ii) (iii) – (iv)


Landslide typology falls/topples – slides flows complex
Landslide visibility poor fair good – excellent

* hydrologic features: (i) water table (ii) springs (iii) stagnant waters (iv) runoff

36 Bull Eng Geol Env (1999) 58 : 21–44 7 Q Springer-Verlag


Landslide hazard assessment

Fig. 15 1.00
Comparison between neural Experts (mean value) LEARNING SET TEST SET
and expert (mean value)
Calculated index with NN
outputs (after Aleotti et al.
1998b)

0.75

Susceptibility index
0.50

0.25
3 4 5 6 7 9 1 1 1 1 1 1 1 1 2 2 2 2 2 2 2 3 3 3 1 2 8 1 1 2 2 2 3 3
1 2 3 4 6 7 8 9 0 1 3 5 7 8 9 0 1 8 0 5 2 4 6 2 6
Studied landslides ( identification number)

“From a probabilistic point of view it is logical to The third point concerns sloping areas which are part of a
assume a base level of probability of failure of 10 –4 per major project such as a dam. The reservoir-rim stability
dam year. There is no reason, however, why engineers for a dam-reservoir system can be a critical factor in deter-
should be satisfied to consider such a failure rate as the mining the overall risk. The destruction and loss of life
norm. Dams should be designed and constructed not to which resulted from the 1964 Vajont slide in Italy is a well-
fail, even if a probability of failure is incorporated into a known and stark example. The fact that the Vajont dam
cost-benefit analysis. Since we know where the greatest itself remained totally stable during the catastrophic failure
weaknesses lie, we should be able to devise the means proved that the risk of failure of one element of the system
for applying judgement to avoid these weaknesses. If we was extremely low whereas for the system as a whole the
succeed we should be able to justify a base level of risk was extremely high. From observations and investiga-
failure no more than 10 –5 per dam year. Such an tions over several years prior to the slide, it was known
improvement is now within the state-of-the-art. Its that continuing slope instability and potential for a large
achievement does not depend on the acquisition of new landslide was the predominant factor in the high overall
knowledge....” risk even though the speed with which the slide occurred
and the enormous consequences could not be foreseen.
There are several important points which must be made
It is clear from the above that:
about this challenging statement from an acknowledged
a) In large and complex systems, the acceptable risk of
leader of the engineering profession. Firstly, natural slopes
landsliding should be comparable with the risk asso-
are quite different from man-made structures like dams.
ciated with other elements of the system considering all
Therefore, our knowledge of the state of their nature is
the failure modes. For example, in a dam reservoir
usually far from perfect even after extensive investigations
system the risk of landsliding along the reservoir rim
– which are rarely available except on major projects. It is
should be as low, or at least of the same order of
unlikely that all the possible mechanisms of failure will be
magnitude, as the risk associated with the failure of the
appreciated or that where all the weaknesses lie will be
dam itself, which may be due to a number of causes
known. Judgement may be helpful but it is questionable
such as structural or foundation inadequacy, overtop-
whether very low failure probabilities should generally be
ping during floods and subsurface soil erosion.
assumed.
b) The concept of acceptable risk is closely related to the
The second point, however, concerns those sloping areas
location of a slope or landslide area, its proximity to
for which it is critically important to maintain stability
constructed facilities, the functions of those facilities
because the consequences of failure are considered enor-
and their operational life.
mous. For such sloping or landslide areas, it is necessary to
c) The consequences of failure of the slope itself and facil-
install effective preventive or remedial measures with a
ities that its failure will impact are extremely important
view to reducing failure probability to very low levels. The
for adopting or specifying levels of acceptable risk. The
extent to which such a goal can be achieved would depend
possible impact on human life, other life, property,
on the extent of resources and the time available for the
infrastructure, economy and the environment must be
project. On this basis it is possible to work towards low to
considered hence location with respect to human settle-
very low target risk levels.
ments and transportation routes is also critically
important.

Bull Eng Geol Env (1999) 58 : 21–44 7 Q Springer-Verlag 37


P. Aleotti 7 R. Chowdhury

Observed frequencies and triggering agents Extent of slope deformations


Calculated failure probabilities for individual slopes must In the past, attention has often been concentrated on slope
be distinguished from the observed frequencies of the failures which result in catastrophic landslides. Such land-
occurrence of landslides. These frequencies are often slides are obviously of enormous significance. Modelling
dominated by the triggering agent. For example, rainfall- these movements in terms of well known performance
induced landslides affect many countries and vast regions functions such as the factor of safety is often quite
in some countries. The damage associated with such land- adequate. However, there are also slopes and landslides
slides is often enormous. The intensity, duration and which move slowly and sometimes imperceptibly and
frequency of rainstorms which cause landsliding must be sensitive subsurface instruments, such as inclinometers,
studied to learn about the correspondence between rainfall are required to monitor these. In many situations, these
magnitudes/intensities and risk levels. In many areas, rain- slow moving instabilities do not cause notable disturbance.
fall may lead only to shallow landslides and there may be a However, over a significant period of time, or after several
predominance of debris flows in some of these areas. Deep rainstorms, the cumulative deformation may be enough to
seated landslides also occur in many areas as a conse- cause considerable damage. Sometimes such cumulative
quence of rainfall. However, such landslides often correlate deformations can result in destruction of houses in urban
with cumulative rainfall for significant antecedent periods areas as well as disruption of roads, highways, railways and
(e.g. 15 days, one month, two months or more) rather than pipelines. On some occasions, complete collapse of a slope
to rainfall intensity over short periods (e.g. several or full reactivation of landslide may take place after a
hours). period of time. but even if catastrophic failure does not
Observed frequencies are thus extremely important in occur, the socio-economic impact may nevertheless be
understanding hazard and risk and in setting target or unacceptable.
acceptable levels of risk for many areas. The frequencies of Even relatively small movements can be significant for the
the triggering agent (eg, rainfall) must therefore be care- railways operating at normal speeds, let alone the high-
fully evaluated. speed systems. In such cases, acceptable levels of risk
Landsliding is often triggered by earthquakes. Such land- related to slope instability should be based on target levels
slides may have different mechanisms and characteristics of lateral and vertical deformations rather than on the
and may occur at different locations from those where probability of complete or catastrophic failure. Geotech-
rainfall-induced landslides occur. The return period of nical modelling should include deformation in addition to
earthquakes of different magnitudes must be estimated in stability represented by the factor of safety F while proba-
order to understand the level of hazard and to specify bilistic modelling should consider deformation as the
target or acceptable levels of risk. performance function in addition to the factor of safety F.
Based on the above the following important points can be
made:
a) Different levels of acceptable risk should apply to
different types of landslides e.g. shallow slides and Summary and conclusions
deep-seated slides.
b) Landslide movements can range from very slow to very This paper considers several aspects of landslide hazard/
rapid. Different levels of acceptable risk should apply risk assessment, the main ones being:
depending on the speed of movement. The impact of a – why is landslide hazard/risk assessment important?
landslide increases significantly with the speed and the – which methods can be used?
travel distance or run-out distance. As a consequence, – how can an acceptable risk threshold be defined?
it would be anticipated that levels of acceptable risk Reference has been made to the human fatalities as well as
would be lower for potential landslides which are likely the enormous economic consequences of landslides every
to be of higher velocity and which would travel farther year. Although landslide-related damage appears to be
compared with those for landslides which would be more extensive in industrialised countries (United States,
relatively less rapid and have a smaller travel distance. Japan, Italy, France, etc.), in view of the increasing pres-
c) In setting the target or acceptable levels of risk, careful sure on the environment, in developing countries it could
attention should be given to the triggering agents (e.g. even lead to economic recession and stagnation. Experts
rainstorms, earthquakes), their frequencies, and their and planners all over the world are therefore becoming
relationship to historical landsliding. increasingly aware of the importance of predicting where,
d) Different levels of acceptable risk may also be appro- how and why landslides will occur in a given area and
priate depending on the triggering agent. For example, numerous approaches to landslide hazard/risk assessment
it may be appropriate to distinguish between target have been developed in recent years.
levels of risk associated with seismically induced land- The benefit of using hazard maps is clear: the logical,
sliding and target levels of risk associated with rainfall- timely and appropriate use of such maps can considerably
induced landsliding. reduce the damage costs and other consequences of land-
sliding. However, less than 25% of the general hazard maps
of large areas have been fully utilised from a practical point
of view(Brabb 1996) largely because many decision-makers

38 Bull Eng Geol Env (1999) 58 : 21–44 7 Q Springer-Verlag


Landslide hazard assessment

and politicians are concerned regarding their possible the authors have proposed a classification of landslide
implications for business and wealth creation as they may hazard assessment methodologies starting from a prelimi-
well indicate, for example, land-use restrictions. In addi- nary bi-partition between qualitative and quantitative
tion, members of a community may have difficulty in approaches according to a scheme similar to those
properly understanding landslide hazard maps. proposed by such authors as Carrara (1982 1988),
Engineers and earth scientists have two important ways to Hutchinson (1995), Leroi (1996) and Soeters and van
overcome these problems: Westen (1996).
a) they must first make certain that the language on maps No single approach is recommended as the most suitable:
is simple, direct and easily understood indeed only a few papers have reported comparative
b) they must take positive steps to ensure planners and studies of different approaches (Carrara et al. 1990;
politicians appreciate the enormous benefits of using Jennings et al. 1991). In addition, selection of the approach
these maps for planning, conservation and the encour- to be adopted for a given project is very often based as
agement of sustainable development. much on economic concerns (cost/benefit analysis, polit-
On the basis of a critical analysis of available literature and ical convenience, etc.) as on the technical nature of the
experience obtained in several projects at various scales, problem.

Table 2
Advantages and disadvantages of different methods of landslide hazard assessment, relationship between type of approach and scale
of work; role of GIS PAGINA 43, PAGINA 37

Scale

Methods Advantage Disadvantage S M L Role of GIS

field geomorphologic allow a rapid assessment taking totally subjective methodology R Y Y only as a drawing
analyses into account a large number of use of implicit rules that hinder tool
factors the critical analysis of the
results
combination of index they solve the problem of the subjectivity in attributing R Y Y overlay of different
maps hidden rules weighted values to the single maps
total automation of the steps classes of each parameter
standardisation of data manage-
ment
logical analytical allow the comparison of require monitoring data prefer- R R Y –
models different slopes ably from installed instruments
mathematically rigorous and applicable mainly to slow-speed
perfectible landslides
statistical analyses objective in methodology systematic collection and anal- Y Y R analyses
(bivariate and multi- total automation of the steps ysis of data concerning the
variate) standardisation of data manage- different factors is quite cumber-
ment some
safety factor – deter- objective in scope and method- the need for detailed knowledge R R Y analyses
ministic approaches ology of the area
quantitative in scope use of appropriate geotechnical
encourages investigation and model requires a lot of experi-
measurement of geotechnical ence
parameters in detail does not take into account
various uncertainties
probabilistic allows consideration of different requires comprehensive data Y R R –
approaches uncertainties otherwise subjective probabili-
quantitative in scope ties required
objective in scope and method- probability distributions difficult
ology especially for low level of hazard
provides new insight which is and risk
not possible in deterministic
methods
neural networks objective in methodology difficulty in the verification of R Y Y –
do not require theoretic knowl- the results when instrumental
edge on physical aspects of the data are not available
problem

R: restricted use; Y: yes

Bull Eng Geol Env (1999) 58 : 21–44 7 Q Springer-Verlag 39


P. Aleotti 7 R. Chowdhury

b) Great attention must be given to triggering factors.


Whenever possible, these must be analysed statistically
in order to: (i) define the critical thresholds above
which landslides are triggered; (ii) define the frequency
with which these thresholds are exceeded. It is also
important to distinguish between levels of risk asso-
ciated with seismically-induced landslides and rainfall-
induced landslides.
Many recent papers deal with risk management problems
and with principles of hazard reduction (Canuti and
Casagli 1994; Fell 1994; Fell and Hartford 1997; Finlay and
Fell 1997; Keaton 1994; Morgenstern; 1995; Wu et al. 1996).
Broadly speaking, risk management can follow two distinct
guidelines which may not be mutually exclusive (Aleotti
and Polloni 1998c; Canuti and Casagli 1996) (Fig. 16). It is
possible to increase the acceptable risk threshold by means
of an information campaign for the population threatened
by a given hazard. This facilities the passage from involun-
tary to voluntary risk, for which a higher threshold is
Fig. 16 usually accepted. Alternatively it may be possible to vary
Landslide risk management (after Aleotti and Polloni 1998c) the components of the risk, eg by reducing hazard (Kock-
elman 1986; Schuster and Kockelman 1996), vulnerability
and/or the number of risk elements.

The paper points out that all the available techniques


present both advantages and disadvantages. Table 2,
although neither complete nor exhaustive, gives a
summary of these techniques and their effectiveness. The References
table also presents some considerations regarding the scale
of applicability of the various approaches, as well as the Abu-zeid N, Mazzini E, Semenza P, Turrini MC (1994) Appli-
role of G.I.S. in development of the procedures. G.I.S is cazione di un metodo cartografico-numerico al bacino
applied extensively in hazard mapping and has enabled the dell’Alpago (BL) per la zonazione della pericolosità potenziale
development of statistical or multi-scenario deterministic da frana. Geologia Tecnica e Ambientale, 3 : 45–55
techniques which would not otherwise have been success- Aleotti P, Baldelli P, Polloni G, De Marchi D (1996a) La
valutazione della pericolosità da frana: esempi metodologici a
fully implemented (Carrara and Guzzetti 1995). The use of varia scala. In: Proc Conf “La stabilità del suolo in Italia: zona-
neural networks (Mazzetti 1991) is less common. These zione della sismicità – frane”. Accademia Nazionale dei Lincei,
mathematical models attempt to simulate the functionality Roma, 30–31/05/1996 pp 19–23
of the human brain in a simplified manner and in recent Aleotti P, Baldelli P, Polloni G (1996b) Landsliding and
years have been found to be effective in civil engineering flooding event triggered by heavy rains in the Tanaro basin.
problems (Pande and Pietruszczak 1995; Raha and Dhar- (Italy). Proc Int Congr Interpraevent 1996, 1 : 435–446,
mappa 1998) and slope stability (Mayoraz et al. 1996; Garmisch-PartenKirchen, 24–28/06/96
Aleotti P, Baldelli P, De Marchi D (1996c) Le reti neurali
Aleotti et al. 1998b). Recent studies have confirmed that nella valutazione della suscettibilità da frana. Geologia Tecnica
neural networks represent a very powerful, versatile and & ambientale, 4 : 37–48
promising computer-based tool. Aleotti P, Baldelli P, Polloni G (1998a) Soil slips, rock-block
It is now widely recognised that a certain level of risk must slides and stream hydraulic processes caused by heavy rains:
be contemplated in any activity or project (Whitman their interaction and relevant hazard. Proc. of the Second Conf
1984). In the case of landslides, the definition of acceptable Environ Management ICEM2, Wollongong, 10–13 Feb 1998,
risk must take into consideration the following: 1 : 553–564
Aleotti P, Baldelli P, Polloni G, Puma F (1998b) Keynote
a) Different levels of risk must be assigned according to paper: Different approaches to landslide hazard assessment.
the velocity of slides (slow or fast) and their dimensions Proc. of the Second Conf Environ Management (ICEM-2),
(shallow failures or deep-seated slides). If possible, Wollongong (Australia) 10–13 Feb. 1998, 1 : 3–10
these two parameters should be analysed simulta- Aleotti P, Polloni G (1998c) Risk scenario in mountain alluvial
neously. For example, soil slips, despite their small fans: some examples in Italian Central Alps. Proc of the 8th
dimensions, are highly destructive and travel at a very Congr Int Assoc Engin Geol, Vancouver – Canada, 21–25
fast speed. On the other hand slow, large-sized failures September 1998
Aleotti P, Baldelli P, Govi M, Polloni G, Puma F, Villani
can cause cumulative deformations that result in severe B (1999) Hazard and risk assessment in the Po River Basin
damage to houses, roads, pipelines, etc. (a recent paper (North Italy). (in press)
on this type of landslide in clayey soils was published Alexander D (1988) Valtellina landslide and flood emergency,
by Baldelli et al. 1998). Northern Italy. Disaster, 12 : 212–222

40 Bull Eng Geol Env (1999) 58 : 21–44 7 Q Springer-Verlag


Landslide hazard assessment

Alexander D (1994) Geomorphology and hazards of landslides. Brabb EE (1996) Hazard maps are not enough. In: Proc of the
Proc. Conf held in Aquater. June, 28 1994, San Lorenzo in 6th Spanish Congress and Int Conf Environ Geol Land-Use
Campo (PS), Italy Planning, Granada 22–25/4/1996, 1 : 331–336
Amadesi E, Vianello G (1978) Nuova guida alla realizzazione di Brand EW (1988) Landslide Risk Assessment in Hong Kong. In:
una carta di stabilità dei versanti. Mem Soc Geol It 19 : 53–60 Proc of 5th Int Symp Landslides, Laussanne, 2 : 1059–1074
Amanti M, Carrara A, Castaldo G, Colosimo P, Gisotti G, Brugner W, Valdinucci A (1972) Schema di cartografia tema-
Govi M, Marchionna G, Nardi R, Panizza M, Pecci M, tica relativa alla stabilità dei terreni ed esempio della sua appli-
Vianello G (1992) Linee guida per la realizzazione di una cazione nel territorio di Alpago (Prov di Belluno). Boll Serv
cartografia della pericolosità geologica connessa ai fenomeni di Geol d’It, 93 : 189–194
instabilità dei versanti alla scala 1 : 50.000 (versione prelimi- Bughi S, Aleotti P, Bruschi R, Andrei G, Milani G, Scar-
nare). Presidenza del Consiglio dei Ministri. Servizio pelli G (1996) Slow movements of slopes interferring with
Geologico, Progetto CARG, 53 pp pipelines: modelling vs monitoring. In: Proc 15th Int Conf
Amanti M, Casagli N, Catani F, D’Orefice M, Motteran G OMAE, Firenze, June 1996
(1996) Guida al censimento dei fenomeni franosi e alla loro Caine N (1980) The rainfall intensity duration control of shallow
archiviazione. Miscell Serv Geol d’It, 7, 109 pp landslides and debris flows. Geogr Annaler, 62 : 23–27
Anbalagan R (1992) Terrain evaluation and landslide hazard Cambiaghi A, Schuster RL (1989) Landslide damming and
zonation for environmental regeneration and land use plan- environmental protection: a case history from Northern Italy.
ning in mountainous terrain. Proc. VI Int Symp on Landslides, In: Proc of Second Int Symp on Env Geotechnology, Shangai,
Christchurch, New Zealand, 2 : 861–868 May 15–19, 1989
Anbalagan R, Singh B (1996) Landslide hazard and risk assess- Cancelli A, Nova R (1985) Landslides in soil debris cover trig-
ment mapping of mountainous terrains – a case study from gered by rainstorm in Valtellina (Central Alps, Italy). In: Proc
Kumaun Hymalaya, India. Eng Geol, 43 : 237–246 4th Int Conf on Landslides, Tokyo, 1 : 267–272
Antoine P (1977) Rèflexions sur la cartographie ZERMOS et Canuti P, Casagli N (1994) Considerazioni sulla valutazione del
bilan des expériences en cours. BRGM Bull, Sect III, 1–2 : 9–20 rischio di frana. In: Atti del Convegno “Fenomeni franosi e
Baeza C, Corominas J (1996) Assessment of shallow landslide centri abitati”, Bologna 27 maggio 1994, Regione Emilia
susceptibility by means of statistical techniques. Proc VII Int Romagna CNR-GNDCI, pp 29–130
Symp on Landslides, Trondheim, June 1996, 1 : 147–152 Canuti P, Casagli N (1996) Previsione e prevenzione del rischio
Baldelli P, Aleotti P, Polloni G (1996) Landsliding-suscepti- di frana. In: Proc Conf “La stabilità del suolo in Italia: zona-
bility numerical mapping at the Messina Strait crossing site, zione della sismicità – frane”. Accademia Nazionale dei Lincei,
Italy. In: Proc VII Int Symp on Landslides, Trondheim, June Roma, 30–31/05/1996, pp 47–58
1996, 1 : 153–158 Carmassi F, Liberati G, Ricciardi C, Sciotti M (1992)
Baldelli P, Aleotti P, Polloni G, Tosti CA, Andrei G, Stability evaluation for unified power plant siting in
Milani G (1998) Fenomeni franosi lenti in terreni argillosi: un geothermal areas. In: Proc VI Int Symp on Landslides, Christ-
contributo alla classificazione e al monitoraggio. In: Proc of the church, 1 : 893–898
Congress “Il rischio idrogeologico e la difesa del suolo”, Accad- Carrara A, Merenda L (1976) Landslide inventory in northern
emia Nazionale dei Lincei, Roma 1–2 ottobre 1998 Calabria, southern Italy. Geol Soc Am Bull, 87 : 1153–1162
Baldi G, Beilin P, Carrara A, Cardinali M, Cavallin A, Carrara A, Catalano E, Sorriso-Valvo M, Reali C, Ossi I
Chiesa S, Crosta G, Detti R, Dovzak C, Floris B, Fras- (1978) Digital Terrain Analysis for land evaluation. Geologia
soni A, Guzzetti F, Marchetti M, Nozza G, Pasqui V, Applicata e Idrogeologia, 13 : 69–127
Reichenbach P (1990) Geograph Inform System Appl (GIS): Carrara A (1982) Cartografia tematica, stoccaggio ed elabora-
landslide hazard maps. Mem Soc Geol It, 45 : 253–258 zione dati. In: atti del Convegno Conclusivo P.F. Conser. Suolo,
Hydro BC (1993) Guidelines for consequence-based dam safety Sottopr. Fenomeni Franosi, Relazione Generale, 9–10 giugno
evaluations and Improvements. Report of August 1993, 1982, Roma, pp 265–281
Hydroelectric Division Carrara A (1983) Multivariate methods for landslide hazard
Berggren B, Fallsvik J, Viberg L (1992) Mapping and evalua- evaluation. Mathematical Geol, 15 : 403–426
tion of landslide risk in Sweden. In: Proc Sixth Int Symp Land- Carrara A (1988) Landslide hazard mapping by statistical
slides, Christchurch, New Zealand, 2 : 873–878 methods:a black box model approach. In: Proc of the Work-
Bernknopf RL, Campbell RH, Brookshire DS, Shapiro CD shop on Natural Disaster in European Mediterranean Coun-
(1988) A probabilistic approach to landslide hazard mapping in tries. Perugia, Italy. Consiglio Nazionale delle Ricerche
Cincinnati, Ohio, with application for economical evaluation. Carrara A, Cardinali M, Detti R, Guzzetti F, Pasqui V,
Bull of the Assoc Engin Geologists, 25 : 39–56 Reichenbach P (1990) Geographical Information System and
Bortolami GC, De Luca DA, Masciocco L (1994) Studio sulla multivariate models in landslide hazard evaluation In: Proc of
realizzazione di carte della franosità a scala regionale. In: Proc the 6th Int Conf Field Workshop on Landslides, Switzerland-
of IV Geoeng Int Congr “Soil and Groundwater Protection”, Austria-Italy. Aug 31st–Sept 12th 1990, 1 : 17–28
Torino Carrara, A Cardinali M, Detti R, Guzzetti F, Pasqui V,
Bosi C, Dramis F, Gentili B (1982) Carte geomorfologiche di Reichenbach P (1991) GIS techniques and statistical models
dettaglio e carte di stabilità: esempi nel territorio marchigiano. in evaluating landslide hazard. Earth Surface Processes and
Geol Appl ed Idrogeol, 20 : 53–62 Landforms, 16 : 427–445
Brabb EE, Pampeyan EH, Bonilla MG (1972) Landslide Carrara A, Cardinali M, Guzzetti F (1992) Uncertainty in
susceptibility in San Mateo County, California. Misc Field assessing landslide hazard and risk. ITC Journal, 2 : 172–183
Studies Map MF360. US Geological Survey, Reston, Va Carrara A, Guzzetti F (1995) Geographical Information
Brabb EE (1984) Innovative Approaches to Landslide Hazard Systems in assessing natural hazard. Selected Contribution
and Risk Mapping, In: Proc 4th Int Symp Landslides, Toronto, from the International Workshop held in Perugia on Sept
1 : 307–323 20–22 1993, Kluver Academic Publisher, Dordrecht/Boston/
Brabb EE, Harrod BL (1989) Landslides: Extent and Economic London, pp 354 (Advances in Natural Technological Hazard
Significance AA Balkema, Rotterdam/Brookfield Research Series, Volume 5)

Bull Eng Geol Env (1999) 58 : 21–44 7 Q Springer-Verlag 41


P. Aleotti 7 R. Chowdhury

Catenacci V (1992) Geological and geoenvironmental failure Chowdhury RN (1998b) Discussion on the paper “Earthquake
from the post-war to 1990, Italy. Servizio Geologico Nazionale destructiveness potential factor and slope stability” by Crespel-
– Memorie Descrittive della Carta Geologica d’Italia, pp 301 lani et al., Geotechnique, 40, discussion submitted to Geotech-
Choubey VD, Chaudhari S, Litoria PK (1992) Landslide nique (in print)
hazard zonation in Uttarkashi and Tehri districts, UP Chowdhury RN (1998c) Evolving geotechnical reliability assess-
Himalaya, India. In: Proc of the Sixth Int Symp Landslides, ment strategies towards? Invited expert paper, Special Volume
Christchurch, New Zealand, 2 : 911–917 on year 20th Geotechincs, AIT, Bangkok, Dec 1998
Chowdhury RN (1976) The Mechanism of Natural Slope Fail- Chowdhury RN (1998d) Hazard and risk assessment with parti-
ures in the Greater Wollongong Area of New South Wales. cular reference to landslide. In: Workshop on Landslide Hazard
Search, Sydney, 7, No.9 : 396–97 Assessment – Progressing from Qualitative to Quantitative
Chowdhury RN, Bertoldi C (1977) Residual Shear Strength of Approaches, Wollongong February 8–9, 1998
Soil from Two Natural Slopes. Australian Geomechanics Chowdhury RN, Flentje P (1998a) A landslide database for
Journal, G7 : 1–10 landslide hazard assessment. In: Workshop on Landslide
Chowdhury RN (1980) Landslides as Natural Hazards – Mecha- Hazard Assessment – Progressing from Qualitative to Quanti-
nisms and Uncertainties. Geotechnical Engineering, tative Approaches, Wollongong February 8–9, 1998
11 : 135–180 Chowdhury RN, Flentje PN (1998b) The Thredbo landslide of
Chowdhury RN (1984) Recent developments in landslide 30 July 1997, New South Wales, Australia – an overview based
studies: probabilistic methods – state-of-art-report. In: Proc of on media reports. Landslide News, 11 : 26–27
the IV Int Symp on Landslides, Toronto, 1 : 209–228 Chowdhury RN, Tabesh A (1998) Seismic response of coastal
Chowdhury RN, Grivas DA (1985) Probabilistic Model of escarpments. In: Proc (CDROM) Sixth US Nat Conf on Earth-
Progressive Failure of Slopes, J Geotechn Eng Div, ASCE, 108, quake Engineering, Seattle, Washington, USA, May 31 – June 4,
No GT 6 : 803–820 pp 1–8
Chowdhury RN, Tang WH (1987) Comparison of Risk Models Christian JT, Ladd CC, Baecher GB (1992) Reliability and
for Slopes. In: Proc Fifth Int Conf on Applications of Statistics Probability in Stability Analysis. In: Raymond B Seed, Ross W
and Probability in Soil and Structural Engineering, Vancouver, Boulanger (eds) Proc ASCE Speciality Conf Stability and
Canada, pp 863–897 Performance of Slopes and Embankments-II, A 25 Year
Perspective, 2 : 1071–1111
Chowdhury RN, Tang WH, Sidi I (1987) Reliability Model of
Progressive Slope Failure, Geotechnique, 37 : 467–481 Clerici A, Dall’Olio N (1995) La realizzazione di una carta
della stabilità potenziale dei versanti mediante tecniche di
Chowdhury RN, Zhang S (1989) Bayesian Updating for Open analisi statistica multivariata e un Sistema d’Informazione
Pit Slopes. In: Proc of the Second Large Open Pit Conf, Geografica. Geologia Tecnica & Ambientale, 4 : 49–57
Aus.I.M.M., Latrobe Valley, Victoria Australia, April, 9–13
Cotecchia V (1978) Systematic reconnaissance mapping and
Chowdhury RN (1992) Modelling the Risk of Progressive Slope registration of slope movements. Bull of Int Ass Eng Geol,
Failure. Can Geotechn J, 29, 1 : 94–102 17 : 5–37
Chowdhury RN (1993) Field Guide for Technical Tour on Crespellani T, Madiai C, Vannucchi G (1998) Earthquake
Urban Slope Stability. In: Proc Int Conf on Environ Manage- destructiveness potential factor and slope stability. Geotec-
ment, Geo-Water and Engineering Aspects, 58 pp nique, 40 : 411–420
Chowdhury RN, Zhang S (1993) Modelling the Risk of Progres- Crosta G (1998) Regionalization of rainfall thresholds: an aid to
sive Slope Failure: A New Approach. Reliability Engineering landslide hazard evaluation. Environ Geology (in print)
and System Safety, 40 : 17–30 Crozier MJ (1992) Landslide hazard assessment: theme report.
Chowdhury RN (1994) Evaluating Geotechnical Risk. Ground In: Proc of 6th Int Symp on Landslides, Christchurch,
Engineering. 27 : 35–40 3 : 1843–1848
Chowdhury RN, Xu DW (1994a) Slope System Reliability with Cruden DM, Brown WM (1992) Progress towards the World
General Slip Surfaces. Soils & Foundations, 34 : 99–105 Landslide Inventory. In: Proc of 6th Int Symp on Landslides,
Chowdhury RN, Xu D (1994b) Simulation of embankment Christchurch, 1 : 59–64
response to earthquakes. In: Proc Fifth US Nat Conf on Earth- Cruden DM, Varnes DJ (1996) Landslide Types and Processes.
quake Eng, Chicago, Illinois, July, 3 : 507–516 Chapter 3 in Landslides-Investigation and Mitigation, Ed AK
Chowdhury RN, Xu DW (1995) Geotechnical System Reliability Turner and RL Schuster, Transportation Research Board
of Slopes, Reliability Engineering & System Safety, Special Report 247, National Research Council, USA, pp 36–75
47 : 141–151 Cruden D, Fell R (1997) Landslide Risk Assessment. IUSG
Chowdhury RN, Flentje PN (1996) Geology and land insta- Working Group on Landslides, Committee on Risk Assessment
bility mapping using a GIS package as a building block for the – Workshop, Honolulu 19–21 February 1997, Balkema, 371 pp
development of a risk assessment procedure. In: Proc VII Int Einstein HH (1988) Special lecture: Landslide risk assessment
Symp on Landslides, Trondheim, June 1996, 1 : 177–182 procedure. In: Proc V Int Symp on Landslides, Lausanne,
Chowdhury RN (1997) Comparison of approaches for seismic 2 : 1075–1090
performance evaluation of earth structures. Keynote paper. In: Ellen S, Peterson DM, Reid GO (1982) Map showing areas
Proc Third Int Geotech Conf, Cairo, Egypt, pp 171–191 susceptible to different hazards from shallow landsliding,
Chowdhury RN, Flentje PN (1997) Relevance of mapping for Marin County and adjacent parts of Sonoma County, Cali-
slope stability in the Greater Wollongong area, New South fornia. US Geological Survey Miscellaneous Field Studies Map
Wales, Australia. In: Proc Int Symp Engineering Geol Environ, MF-1406
Athens (Greece), 23–27 June 1997, 1 : 569–574 Fell R (1992) Landslides in Australia. In: Proceedings Sixth Int
Chowdhury RN, (1998a) Sliding block approach for seismic Symp Landslides, Christchurch, New Zealand, Balkema,
stability analysis of earth dams. In: Proc Second Int Conf 3 : 2059–2100
Environ Management (ICEM2), Elsevier (Oxford), Fell R (1994) Landslide risk assessment and acceptable risk. Can
2 : 1029–1036 Geotechn J 31 : 261–272

42 Bull Eng Geol Env (1999) 58 : 21–44 7 Q Springer-Verlag


Landslide hazard assessment

Fell R, Hartford D (1997) Landslide risk management. In: Jennings PJ, Siddle HJ, Bentley SP (1991) A comparative
IUSG Working Group on Landslides, Committee on Risk study of indirect methods of landslip potential assessment. In:
Assessment – Workshop, Honolulu 19–21 February 1997 Proceedings Conf “Slope stability engineering developments
Fenti V, Silvano S, Spagna V (1979) Methodological proposal and applications”, 15–18 April 1991, Isle of Wight, pp 127–132
for an engineering geomorphological map. Forecasting rock- Keaton JR (1994) Risk-based probabilistic approach to site
falls in the Alps. Bull of the Int Assoc Eng Geol 19 : 134–138 selection. Bull of the Assoc Eng Geol, 31 : 217–229
Finlay PJ, Fell R (1997) Landslide: risk perception and accept- Kienholz H (1978) Maps of Geomorphology and Natural Hazard
ance. Can Geotech J, 34 : 169–188 of Griendelwald, Switzerland, scale 1 : 10.000. Artic and Alpine
Flentje PN (1998) PhD thesis, University of Wollongong, Research, 10 : 169–184
Australia Kingsbury PA, Hastie WJ, Harrington AJ (1992) Regional
Flentje PN, Chowdhury RN (1999) Quantitative landslide landslip hazard assessment using a Geographical Information
hazard assessment in an urban area. In: Proc of 8th Australia – System. In: Proc Sixth Int Symp Landslides, Christchurch, New
New Zealand Conf Geomechanics, Hobart, Tasmania, Zealand, 2 : 995–999
Australia, February 1999 Kockelman WJ (1986) Some techniques for reducing landslide
Genevois R, Tecca PR (1987) Analisi probabilistica della stabil- hazards. Bull of the Assoc Eng Geol, 23 : 29–52
ità dei versanti: applicazione per la realizzazione di una carta di
Lees BG (1996) Neural networks applications in the geosciences:
pericolosità nella media Valle del Fiume Tammaro (Benev-
an introduction. Computer and Geosciences,22 : 955–957
ento). Mem Soc Geol It, 37 : 157–170
Gokceoglu C, Aksoy H (1996) Landslide susceptibility mapping Leiba MM (1998) Human and financial impact of landslides in
of the slopes in the residual soils of the Mengen Region Australia. In: Proc Second Conf Environmental Management
(Turkey) by deterministic stability analyses and image proc- (ICEM-2), Wollongong (Australia) 10–13 Feb 1998
essing technique. Engineering Geology, 44 : 147–161 Leroi E (1996) Landslide hazard – Risk maps at different scales:
Granger K (1998) Managing the risks to urban communities objectives, tools and developments. In: Proc VII Int Symp
posed by environmental hazards. Proc Second Conf Environ Landslides, Trondheim, June 1996, 1 : 35–52
Management (ICEM-2), Wollongong (Australia) 10–13 Feb Lucini (1979) Un metodo grafico per la valutazione della frano-
1998, 1 : 11–20 sità. Mem. Note Istituto Geol Appl Napoli, 11
Green WH, Ampt GA (1911) Studies on soil physics 1. The flow Mayoraz F, Cornu T, Vulliet L (1996) Using neural networks
of air and water through soils. J Agric Sci, 4 : 1–24 to predict slope movements. Proc VII Int Symp Landslides,
Guerricchio A, Melidoro G (1979) Fenomeni franosi e neotet- Trondheim, June 1996, 1 : 295–300
tonici nelle argille grigio-azzurre calabriane di Pisticci Mazzetti A (1991) Reti Neurali Artificiali: introduzione ai prin-
(Lucania). Geol Appl e Idrogeol, 14 : 105–138 cipali modelli e simulazione su Personal Computer. Apogeo –
Guida D, Guida M, Iaccarino G, Lambiase S, Metacalf G, Editrice Informatica, Milano, 108 pp
Salzano G, Vallario A, Vecchio V, Zicari G (1979) Meherota GS, Dharmaraju R, Kanungo DP (1994) Landslide
Proposta per l’elaborazione di cartografia tematica inerente alla hazard zonation – a tool for environmental development and
dinamica dei versanti. Mem Soc Geol It 19 : 170–198 protection of hilly terrain. In: “Man and Mountain 94”: primo
Gupta P, Anbalagan R (1997) Slope stability of Theri Dam convegno internazionale per la protezione e lo sviluppo
Reservoir Area, India, using landslide hazard zonation (LHZ) dell’ambiente montano, pp 101–113
mapping. Quarterly Journal of Eng Geology, 30 : 27–36 Meneroud JP, Calvino A (1976) Carte ZERMOS, Zones Expo-
Hammond C, Prellwitz R, Miller S (1992) Landslide hazard sées à des Risques liés aux Mouvements du Sol et du Sous-Sol à
assessment using Monte Carlo simulation. In: Proc Sixth Int 1 : 25.000, Region de la Moyenne Vesubie (Alpes Maritimes).
Symp Landslides, Christchurch, New Zealand, 2 : 959–964 Bureau de Recherches Géologiques et Minières, Orléans,
Hansen A (1984) Landslide hazard analysis. In: Brundsen D, and France
Prior DB (eds) ”Slope Instability“, Wiley, New York, Mora S (1995) The impact of natural hazard on socio-economic
pp 523–602 development in Costa Rica. Environmental & Engineering
Hartlen J, Viberg L (1988) General Report: Evaluation of land- Geoscience, 1 : 291–298
slide hazard. In: Proc V Int Symp on Landslides, Lausanna, Morgenstern NR (1995) Managing risk in geotechnical engi-
2 : 1037–1057 neering. In: 10th Panamerican Regional Conf on Soil
Hastie WJ (1990) Landslip hazard modelling, Wellington Mechanics and Foundation Engineering, Guadalajara, Mexico
Region, New Zealand. In: Proc of the Inaugural Colloquium of Moser M, Hohensinn F (1983) Geotechnical aspects of soil slips
the Spatial Information Research Centre, University of Otago, in Alpine regions. Eng Geol 19 : 185–211
Dunedin, pp 96–105
Mulder HFHM, Van Asch TWJ (1988) A stochastical approach
Hearn GJ (1995) Landslide and erosion hazard mapping at Ok
to landslide hazard determination in a forested area. In: Proc V
Tedi copper mine, Papua New Guinea. Quarterly Journal of
Int Symp Landslides, Lausanne, 2 : 1207–1210
Engineering Geology, 28 : 47–60
Humbert M (1977) Risk mapping of areas exposed to move- Neuland H (1976) A prediction model of landslips. Catena,
ments of soil and sub-soil: French ”Zermos“ maps Bull of Int 3 : 215–230
Assoc of Eng Geol, 16 : 80–82 Newmark NM (1965) Effect of earthquakes on dams and
Hutchinson JN (1995) Landslide hazard assessment. In: Proc VI embankments. Geotechnique, 15 : 139–159
Int Symp on Landslides, Christchurch, 1 : 1805–1842 Nguyen VV, Chowdhury RN (1984) Probabilistic analysis of
Ives JD, Bovis MJ (1978) Natural hazard maps for land-use plan- mining special piles – Two techniques compared. Int J Rock
ning, San Juan Mountains, Colorado, USA. Artic and Alpine Mechanics and Min Science, 21 : 303–312
Research, 10 : 185–212 Nguyen VV, Chowdhury RN (1985) Risk analysis with corre-
Ives JD, Messerli B (1981) Mountain Hazard Mapping in Nepal: lated variables. Geotechnique, 35 : 47–59
Introduction to an Applied Mountain Research Project. Moun- Nielsen TH, Wrigth RH, Vlasic TC, Spangle WE (1979) Rela-
tain Research And Development, 1 : 223–230 tive slope stability and land-use planning in the San Franicisco
Jade S, Sarkar S (1993) Statistical models for slope stability Bay region, California. US Geological Survey Professional Paper
classification. Engineering Geology, 36 : 91–98 944

Bull Eng Geol Env (1999) 58 : 21–44 7 Q Springer-Verlag 43


P. Aleotti 7 R. Chowdhury

Onorati G (1998) Relazione tra gli eventi franosi del maggio Skermer NA, VanDine DG (1992) Catastrophic impact of some
1998 e la piovosità dell’area di Sarno. In: Atti del Convegno ”Il historical mountain landslides. In: Proc of First Can Symp
rischio idrogeologico e la difesa del suolo“, Accademia Geotechnique Natural Hazards (Geohazard ’92), Vancouver,
Nazionale dei Lincei, Roma, 1–2 ottobre 1998 (in press) May 6–9 1992, pp 91–98
Pande GN, Pietruszczak S (1995) Numerical models in geome- Soeters R, Van Westen CJ (1996) Slope stability: recognition,
chanics. In: Proc of the 5th Int Symp Numerical Models in analysis and zonation. In: Turner AK, Shuster RL (eds) “Land-
Geomechanics, Davos, Switzerland, 6–8 sept 1995 slides: investigation and mitigation”. Transportation Research
Panizza M (1975) Ricerche di geomorfologia applicata alla pian- Board – National Research Council, Special Report 247,
ificazione territoriale. Mem Soc Geol It, 14 : 109–112 pp 129–177
Paolini M, Vacis G (1997) Il racconto del Vajont Garzanti Sorriso-Valvo M, Tansi C, Antronico L (1996) Relazione tra
Editore frane, forme del rilievo e strutture tettoniche nella media valle
Peck RB (1980) Where has all the judgement gone? Laurits del Fiume Crati (Calabria). Geogr Fis Dinam Quat 19 : 107–117
Bjerrum Mineforedrag No 5, Norges Geotekniske Institutt, Stevenson PC (1977) An empirical method for the evaluation of
Oslo, 5 pp, also Can Geotechn J 1980, 17 : 585–590 and Norwe- relative landslide risk. Bull Int Ass Eng Geol, 16 : 69–72
gian Geotechnical Institute Publn, 134 : 1–5 Tang WH, Yucemen MS, Ang AH-S (1976) Probability-based
Polloni G, Aleotti P, Baldelli P, Nosetto A, Casavecchia short term design of soil slopes. Can Geotechn J, 13 : 201–215
K (1996) Heavy rain triggered landslides in the Alba area Terzaghi K (1950) Mechanism of landslides. In: Paige S (ed)
during November 1994 flooding event in the Piemonte Region ”Application of geology to engineering practice“. Geol Soc Am,
(Italy). In: Proc VII Int Symp Landslides, Trondheim, June Vol: 83–123
1996, 3 : 1955–1960 Turner AK, Schuster RL, (1996) Landslides-Investigation and
Polloni G, Aleotti P (1998) Terraced slopes in mountain areas: Mitigationb, Special Report 247, Transportation Research
sliding risk and protective measures in Valtellina (Italy). Proc Board, National Research Council, USA, 675 pp
of the 8th Congr Int Assoc Engineering Geol, Vancouver – Varnes DJ, IAEG Commission on Landslides (1984) Landslide
Canada, 21/25 September 1998 hazard zonation – a review of principles and practice.
Pomeroy JS (1979) Map showing landslides and areas susceptible UNESCO, Paris, 63 pp
to sliding in Beaver County, Pennsylvania. US Geological Vaunat J, Leroueil S, Tavenas F (1992) Hazard and risk anal-
Survey Miscellaneous Investigation Map I-1160 ysis of slope instability. In: Proc First Canadian Symposium on
Popescu ME (1998) An integrated approach to geohazard in Geotechnique and Natural Hazards (Geohazard ’92),
Romania on a regional scale. In: Proc Second Conf Environ- Vancouver, May 6–9 1992, pp 397–404
mental Management (ICEM-2), Wollongong (Australia) 10–13 Viberg L (1984) Landslide risk mapping in soft clay in Scandi-
Feb 1998, 2, pp 1269–1278 navia and Canada. In: Proc IV Int Symp Landslides, Toronto,
Pradel D, Raad G (1993) Effect of permeability on surficial 1 : 325–348
stability of homogeneous slopes. J Geotech Eng, ASCE, 119 Visintainer P, Turrini MC (1995) Carta della pericolosità di
Raha D, Dharmappa HB (1998) Use of artificial neural eventi naturali della Val Duron (Trentino Alto Adige). Geologia
networks (ANN) for predicting the performance of activated Tecnica e Ambientale, 2 : 17–33
sludge process. In: Proc Second Conf Environmental Manage- Ward TJ, Ruh-Ming L, Simons DB (1982) Mapping landslide
ment (ICEM-2), Wollongong (Australia) 10–13 Feb 1998 hazard in forest watershed. J Geotechn Eng Div, ASCE, 108
Rupke J, Cammeraat E, Seijmonsbergen AC, van Westen CJ (GT2):319–324
(1988) Engineering geomorphology of Widentobel Catchment, Whitman RV (1984) Evaluating Calculated Risk in Geotechnical
Appenzell and Sankt Gallen, Switzerland: a geomorphological Engineering. J Geotechn Eng, ASCE, 110 : 145–188
inventory system applied to geotechnical appraisal of slope Wieczorek GF (1984) Preparing a detailed landslide-inventory
stability. Engineering Geology, 26 : 33–68 map for hazard evaluation and reduction. Bull of the Assoc Eng
Sarma SK (1975) Seismic stability of earth dams and embank- Geol, 21 : 337–342
ments. Geotechnique, 25 : 743–761 Wieczorek GF (1996) Landslide triggering mechanisms. In:
Sarma SK, Jennings DN (1980) A dynamic pore water pressure Turner AK, Shuster RL (eds) “Landslides: investigation and
parameter. Proc Conf on Dams and Earthquakes, Institution of mitigation”. Transportation Research Board – National
Civil Engineers, October 1981, London, Thomas Telford, Research Council, Special Report 247, pp 76–90
pp 155–160 Wieczorek GF, Gori PL, Jager S, Kappel WM, Negussey D
Scarpelli G, Aleotti P, Baldelli P, Milani G, Brambati E (1996) Assessment and management of landslide hazards near
(1995) Field experience of pipeline in geologically unstable Tully Valley landslide, Syracuse, New York, USA. Proc VII Int
areas. Proc of the 14th Int Conf Offshore Mechanics and Arctic Symp Landslides, Trondheim, June 1996, 1 : 411–416
Engineering (OMAE 95), Copenhagen, June 1995, 5 : 87–96 Wilson RC, Wieczorek GF (1995): Rainfall thresholds for the
Schuster RL (1996) Socio-economic significance of landslides. initiation of debris flow at La Honda, California. Environ-
In: Turner, Shuster (eds) “Landslides: investigation and mitiga- mental & Engineering Geoscience, 1 : 11–27
tion”. Transportation Research Board – National Research Wu TH, Kraft LM (1970) Safety Analysis of Slopes. J of Soil
Council, Special Report 247, pp 12–35 Mech and Foundation Eng Div ASCE, 96 : 609–630
Schuster RL, Kockelman WJ (1996) Principles of landslide Wu TH, Abdel-Latif MA (1994) Prediction, Mapping and
hazard reduction. In: Turner AK, Shuster RL (eds) “Landslides: Updating of Landslide Hazard, Proc Conf on Deep Foundation
investigation and mitigation”. Transportation Research Board Ground Improvement, AIT Bangkok, Balkema
– National Research Council, Special Report 247, pp 91–105 Wu TH, Tang WH, Einstein HH (1996) Landslide Hazard and
Siddle HJ, Jones DB, Payne HR (1991) Development of a meth- Risk Assessment. In: Turner AK, Shuster RL (eds) “Landslides:
odology for landslip potential mapping in the Rhondda Valley. investigation and mitigation”. Transportation Research Board
In: Proceedings of the Conf “Slope stability engineering devel- – National Research Council, Special Report 247, pp 106–118
opments and applications”, 15–18 April 1991, Isle of Wight, Zhang S (1990) Evaluation and Updating of Slope Reliability,
pp 121–126 Ph.D thesis, University of Wollongong

44 Bull Eng Geol Env (1999) 58 : 21–44 7 Q Springer-Verlag

Das könnte Ihnen auch gefallen