An Inventory-Based Approach to Landslide
Susceptibility Assessment and its Application to the
Virginio River Basin, Italy
N. CASAGLI
F. CATANI
Earth Sciences Department–University of Firenze, Via G. La Pira 4–50121, Firenze, Italy
C. PUGLISI
G. DELMONACO
ENEA: Italian National Agency for Energy, New technologies and Environment-Roma,
Via Anguillarese, 301, Santa Maria di Galeria–00100, Roma, Italy
L. ERMINI
Earth Sciences Department–University of Firenze, Via G. La Pira 4–50121, Firenze, Italy
C. MARGOTTINI
ENEA: Italian National Agency for Energy, New technologies and Environment-Roma,
Via Anguillarese, 301, Santa Maria di Galeria–00100, Roma, Italy
Key Terms: Landslide Susceptibility, Landslide Factors, Slope Instability Indicators, GIS
ABSTRACT
An inventory-based method for the assessment
of landslide susceptibility is presented in this article.
The method has been tested in the Virginio River
Basin, a tributary of the Arno River whose confluence
is located about 20 km downstream from Florence
(Italy). The scope of this study includes setting up
a procedure for landslide hazard zoning to be applied
by those urban planners typically working on small
areas at large scale. The proposed method deals with
traditional and well-known landslide hazard analyses,
based on geomorphological tools, and its most original
contribution is represented by the attempt to carry
out and apply a technique for landslide hazard assessment that takes into account two different scales
of analysis. The basis of a detailed landslide inventory
and the first phase of this research was an in-depth
geomorphological investigation at basin scale (1:
25,000–1:10,000). This was aimed at indicating the
most important factors influencing the landslide processes within the area, which also turned out to be the
most generally accepted factors: a) slope gradient in
which the landslide originated, b) lithology, and c)
land cover. Once those factors were defined as
thematic vector data, they were expressed using GIS
overlay mapping, allowing the identification, for the
entire Virginio River basin, of first-order homogeneous domains (Unique Condition Units, UCUs) that
contain, for each landslide type, unique combinations
(domains) of the selected hillslope stability factors.
The domains are the basic Terrain Units for the
subsequent landslide susceptibility assessment and
mapping, which was carried out at slope scale (1:
10,000–1:2,000). Landslide factors not identified in
the first phase of analysis, but considered to have
played an important role in contributing to the
activation of the mass movements, so-called secondorder landslide preparatory factors, have been taken
into account in the second phase of the analysis. Once
mapped and spatially referenced, these factors were
overlain by vector-based GIS techniques to define
second-order UCUs which, in turn, constituted the
basis of a landslide susceptibility function. Essentially,
this is a logic function based on the presence/absence
of preparatory factors and slope instability indicators
within previously selected Unique Condition Terrain
Units. The final mapping of the areas characterized by
different landslide susceptibility levels was performed
by vector- and raster-based GIS techniques on the
Environmental & Engineering Geoscience, Vol. X, No. 3, August 2004, pp. 203–216
203
Casagli, Catani, Puglisi, Delmonaco, Ermini, and Margottini
basis of the number and rank of the preparatory
factors. The final landslide susceptibility classes,
defined by a logical and easily replicable procedure,
are considered to be useful in the decision-making
procedures associated with territorial planning.
inventories are commonly available at scales ranging
between 1:25,000 to 1:10,000, while a digital terrain model
with a planimetric resolution of 20 m is available for the
whole territory.
INTRODUCTION
Existing Methodologies for Landslide
Susceptibility Assessment
Landslide hazard and risk analysis has been gaining
interest from urban planners and policy makers because of
the increasing socioeconomic impact of landslides on
human activities (Schuster, 1996). In 1998 the Italian
Government, following the landslide disaster that affected
the Sarno area (Campania Region), promoted new policies
on landslide and flood risk assessment that require
National and Regional River Basin Authorities to identify
and map areas subject to landslide risk on the basis of four
different levels.
The law, even though representing an important step
toward the establishment of correct land planning policies,
disregards some important factors that require consideration. In particular, mass movements are not differentiated
on the basis of type or intensity, which should be carefully
taken into account for the prediction of possible damages.
A basic requirement of the law is that ‘‘a key action for
mapping areas characterized by different landslide
susceptibility levels is the correct recognition of events
that have occurred in the past.’’
The first step of a landslide risk analysis is a landslide
hazard assessment. In most cases it is not possible to
evaluate the landslide hazard in absolute terms because of
the difficulties associated with acquiring a statistically
significant amount of data regarding the temporal frequency of movements or a sufficient understanding of the
relationships associated with activation and triggering
events (e.g., intense rainfall and earthquakes). For this
reason, small-scale landslide hazard analyses are conducted to evaluate how, in relative terms, the territory is
prone to landsliding (Varnes and IAEG, 1984; Einstein,
1988; Canuti and Casagli, 1996). This information is
commonly acquired by merging slope processes (effects of
instability) with analysis of physical terrain features that
are responsible for instability (causes). As emphasized by
Sorriso Valvo (2002), this kind of landslide hazard may be
identified by the term ‘landslide susceptibility,’ commonly
defined as the landslide hazard zoning in space with no
reference to time (Brabb, 1984; Sorriso Valvo, 2002). The
quality of susceptibility analyses is often controlled by the
scale at which phenomena occur, compared to that at which
they are surveyed in the analysis. If the former is smaller
than the latter, a misleading factor is introduced that can
affect the correctness of the entire assessment procedure.
Another important limitation is represented by the
availability of spatial terrain data. For most parts of Italian
territory, geological maps, land cover maps, and landslide
Conventional methods to assess landslide susceptibility, according to Soeters and Van Westen (1996) and
Aleotti and Chowdhury (1999), can be classified into four
broad categories: a) landslide inventories, b) heuristic
methods, c) statistical analyses, and d) deterministic
approaches.
The landslide inventory represents the most basic and
simple method of landslide hazard assessment. In its
basic form, the hazard map is derived directly from the
landslide inventory map. This method is only partially
satisfactory because attributing null hazard levels to
areas outside the landslide boundaries excludes areas
in which landslides have not currently been recognized.
For this reason, this method is suitable only to areas in
which easily recognizable landslides are prevalent.
Heuristic approaches are usually based on geomorphological analyses aimed at recognition and correct interpretation of the factors that control landslide
occurrences. The hazard assessment is carried out by
a geomorphologist, through both fieldwork and aerial
photointerpretation. In some applications, the analysis is
accomplished by combining several thematic maps in
which factors affecting landslide occurrence are weighted
on the basis of the skills and experience of the earth
scientists responsible for the analysis. Heuristic approaches have been criticized repeatedly by several
authors because of their highly subjective nature (Soeters
and Van Westen, 1996).
Several authors agree that statistical methods are more
appropriate for hazard zoning, since the degree of
subjectivity is reduced to a minimum. Results of the
inventory are compared with the physical terrain factors
influencing landslides. In particular, multivariate statistical approaches (black-box models such as factor analysis
or discriminant analysis) have been successfully applied
in landslide hazard mapping (Carrara et al., 1978;
Carrara, 1983, 1984, 1989; Carrara et al., 1990, 1991,
1995). However, such approaches, although conceptually
simple, also have some limitations because of the great
complexity in identifying the slope-failure processes,
systematically collecting and representing all predisposing factors related to landsliding, and applying geomorphological predictive modeling of failure over large
areas (Guzzetti et al., 1999). Theoretical limitations could
arise from transforming nonparametric variables, usually
codified on the basis of an ordinal scale, into absolutescale measured variables (Sorriso Valvo, 2002). Often
204
Environmental & Engineering Geoscience, Vol. X, No. 3, August 2004, pp. 203–216
Landslide Susceptibility Assessment
Table 1. Hypsography of the Virginio River basin.
Catchment Area
Maximum elevation a.s.l.
Minimum elevation a.s.l.
Mean elevation a.s.l.
River length
60.3 km2
411 m
43 m
233 m
19.7 km
a.s.l. ¼ above sea level.
‘microzoning’ landslide susceptibility at local scales
(reference scale 1:10,000–1:2,000). The landslide susceptibility maps could be useful for end users, such as the
National and Regional River Basin Authorities, who are
responsible for land planning activities. The methodology
consists of three main phases: a) analysis of a landslide
inventory, b) production and use of thematic maps of
landside factors, and c) specific field investigations to
assess locally variable susceptibility levels.
Geomorphological and Geological Settings of the
Virginio River Basin
Figure 1. Location map of Virginio River basin, Italy (inset), showing
topographic relief.
the field mapping scale of variables considered in the
analysis is different from the scale at which slope instability phenomena occur or have to be recognized. In addition, there are several local slope instability indicators
(crevices, trenches) that play an important role in the forecasting of a landslide event but are not usually included in
statistical analyses. Statistical methods performed
best when applied at the scale of regional land planning (;1:100,000–1:25,000), but encountered major
limitations when used for urban planning purposes at
the local scale (;1:10,000–1:2,000).
Deterministic approaches consist of slope stability
analyses generally designed to determine a safety factor.
Application of deterministic models requires detailed
geotechnical and hydrological data and the correct
knowledge of the failure mechanisms affecting the
investigated slopes. Except for failure mechanisms that
can be interpreted through infinite slope models, a twodimensional problem easily managed by GIS (Montgomery and Dietrich, 1994; Terlien et al., 1995), deterministic
models are suitably applied only to small areas, at the
scale of a single slope.
The main goal of this research is to develop a userfriendly methodology for landslide susceptibility assessment, based on field geomorphological reconnaissance. In
particular, the proposed criteria are for the performance of
The Virginio River Basin was selected as the test
site for the application of this method because it is
representative of lithological, climatic, vegetational, and
land cover conditions common to several large regions in
central Italy and the Mediterranean. The Virginio River is
located in the middle portion of the Arno River basin
(Figure 1) and it is a southern tributary of the Pesa River,
which is, in turn, a tributary to the Arno River about 20
km downstream from Florence, Italy. The most important
hypsographic characteristics of the Virginio River basin
are summarized in Table 1.
According to rainfall data gathered over a 30-year
period by the rain gauges of Tenuta il Corno and
Montespertoli, located, respectively, along the eastern
and western water divides of the catchment area, the
Virginio River basin receives a mean annual rainfall of 895
mm, concentrated in autumn and spring (with maximums
in November and April). Summer is usually very dry.
From a geological point of view, the basin was completely dissected by Pliocene marine succession. The
outcrops in the area were mapped in detail by Canuti and
others (1982) and Amato (2001). The sediments were
deposited mostly during the lower Pliocene; fill basins
developed after the end of the main compressional phase
of this inner part of the Apennines (Martini et al., 2001).
Traditionally these basins have been interpreted to be
a consequence of an extensional tectonic phase (Merla,
1952) that affected the whole of the Apennines from the
Tortonian. These depressions went through phases of high
subsidence with rates that overcame the deposition rates.
They are typically characterized by an eastward asymmetry because of the greater development of normal faults on
Environmental & Engineering Geoscience, Vol. X, No. 3, August 2004, pp. 203–216
205
Casagli, Catani, Puglisi, Delmonaco, Ermini, and Margottini
Figure 2. Schematic cross section of the Virginio River basin showing stratigraphy, landslides affecting the valley slopes, and the lithology control on
slope instability phenomena.
the eastern side with respect to the antithetic faults that
border the basins along its western side. The outcropping
terrains are marine and lacustrine sediments (Pliocene to
Pleistocene) in almost horizontal beds, over which recent
fluvial deposits of the Virginio River are deposited
(Figure 2). Following what was proposed by Canuti and
others (1982), the main terms of this cyclic series are,
from the surface down: Pcg, weakly cemented conglomerates with a sandy to silty matrix, representing the most
western deposits of the more widespread outcrops that
are present in the Pesa River Basin (S. Casciano
pebbles), interpreted as deltaic succession deposits
(Canuti et al., 1966); Pcg-S, conglomerates and weakly
indurated sands, representing the passage in the deposition between pebbles and sands, with lateral
alternations resulting from the several pulses of transgression and retreat of the sea.
Ps, sands, sometimes weakly indurated, with levels
composed of loose materials that prevail in the western
and more distal area of the deltaic system; Ps-Pag, levels
of sands and clay materials, sometimes with poor mechanical properties, representing the passage to a deep-water
depositional environment; and Pag, clays of marine origin
outcropping at the base of the succession only in the
Montespertoli area.
Land cover is represented mainly by vineyards and
olive groves. These agricultural activities have long been
the principal cause of hillside modification (Canuti et al.,
1979a, 1979b; 1982). In the flatlands, the principal crops
are wheat, corn, and sunflower, which are generally sown
at the end of the winter.
The relief is smooth and somewhat gentle, but the
general slope gradients are about 108, especially where the
soils are more suitable for the cultivation of grapevines
and olives. Often slopes show stepped profiles controlled
206
by the horizontal bedding and differences in lithology that
strongly control the hillslope processes.
Outline of the Methodology and its Application
The proposed methodology is an attempt to overcome
some of the previously mentioned limitations in terms of
field representativeness and degree of subjectivity when
dealing with landslide susceptibility mapping at a local
scale. The application of detailed stability slope scale
analyses to a geographical context (such as the Virginio
River basin), in which instability processes are very
common, is impractical, if not unfeasible. For this reason
we decided to set up a double-level investigation to
represent the most innovative feature of this method. The
analysis consists of a detailed geomorphological field
investigation, combining the results of geomorphological,
geological, hydrogeological, and land cover surveys at
hillslope scale (1:10,000–1:2,000), on those portions of
the territory previously defined at basin scale (1:25,000–
1:10,000), as being prone to landslides.
Application of the present methodology requires
knowledge of physical terrain features that became
standard requirements in most landslide susceptibility
analyses (Aleotti and Chowdhury, 1999), including:
landslide inventory maps;
lithological maps (at a reference scale of 1:10,000);
Digital Elevation Models with adequate resolution (in
the Apennine area the maximum pixel size should be
100 m2); and
land cover maps.
The different concepts highlighted in the Introduction
section of this article are briefly summarized in the flow
chart in Figure 3. Setting up a landslide inventory is the
Environmental & Engineering Geoscience, Vol. X, No. 3, August 2004, pp. 203–216
Landslide Susceptibility Assessment
Figure 3. Flow chart describing the proposed methodology.
preliminary requirement for application of the methodology, followed by other phases that, with increasing detail
in the analysis, lead to the mapping of landslide
susceptibility.
In this approach, causes and effects of landslides are
ranked on the basis of two broad categories, considered
significant at different scales (Table 2): first-order causes
are landslide preparatory factors (LPF), well known to be
the most important physical terrain features in predisposing the occurrence of a landslide, while secondorder causes are represented by all the other LPFs that
locally control the dynamics of slope instability. At the
same time, the inventoried landslides can be considered
as first-order effects, available for describing slope
processes at basin scale, while local indicators, such as
formation of trenches and superficial creeping, can be
considered effects at the slope scale. Finally, first-order
and second-order causes and effects constitute input data
for the implementation of different susceptibility functions (Table 2).
Analysis of the Landslide Inventory
An inventory of active and dormant landslides has
been compiled in the Virginio River area through
a standard field data form set up by SGN (Italian
Geological Survey) and the CNR-GNDCI (National
Research Council–National Group for Hydrogeological
Disasters Reduction) (Amanti et al., 1996). The reference
mapping scale was 1:10,000 and the inventory was
performed following three major phases:
analysis of available references and previous investigations;
aerial photointerpretation carried out at different scales
as a function of the availability of aerial photographs;
and
fieldwork designed to survey and validate the results
obtained during the previous phases.
The literature review took into account both the scientific
reports and the documents issued by the Public Administrations for urban planning purposes. The most
important references were the geomorphological analyses
of a portion of the Virginio River basin carried out by
Canuti and others (1982). This was one of the first
examples of multitemporal analysis of landslides aimed at
the evaluation of slope activity, based on the Geomorphological Map produced by the Province of Firenze
in 1995 and the preliminary investigations carried out by
the Arno River Basin Authority.
Most of the aerial photointerpretation was carried out
on the photographs from SOREM, 1998, elevation 1,100
m (proximal scale 1:7,500) and was limited to the
southern sector of the basin to CGR Parma 1993/1994.
The digital ortho-photos (1:10,000 scale) from AIMA1997 were used to reconstruct the activity of the
landslides and to apply a multitemporal analysis.
Most slopes of the Virginio River basin show the
typical forms of a landscape affected by landslide
activity. Although farming activities partly hide superficial evidence of landslides, scars and counterslopes can
Environmental & Engineering Geoscience, Vol. X, No. 3, August 2004, pp. 203–216
207
Casagli, Catani, Puglisi, Delmonaco, Ermini, and Margottini
Table 2. Ranking of the causes and effects of landslides.
Causes
Effects
Results
First
Lithology, slope
Landslide
Macroareas
order
angle, land cover
inventory
Second Local
Slope instability
Susceptibility
order
geomorphological
indicators
ranking
factors
(crevices, trenches,
local failures)
be found, usually close to the main water divide.
Hummocky forms prevail along the median part of the
slopes. The toe area is recognizable as a pronounced
lobe.
Fieldwork carried out during the month of April 2001
confirmed (as stated, for example, by Ardizzone et al.
(2002) and Sorriso Valvo (2002)) that even though the
aerial photointerpretation is a cost-effective procedure for
mapping landslides, it has a low degree of accuracy,
especially when a subsequent detailed field survey is not
conducted.
During the field survey phase, it was important to
recognize that several landslides classified as dormant by
the aerial photointerpretation showed signs of recent
reactivation, probably due to severe rainfall events in the
area between November 2000 and April 2001 (with
a peak of 167 mm of rainfall in the northern part of the
Arno River basin on November 20 and 21, 2000).
Rotational and translational slides prevail, mainly along
deep-seated rupture surfaces. They sometimes evolve into
earth flows at the toe (Figure 4). With reference to the
scheme proposed by Cruden and Varnes (1996), recorded
velocities range between slow and moderate. Rupture
surfaces are often located along overconsolidated clays.
Results of geotechnical tests carried out on this material
(Caioli, 1993), and shown in Table 3, give high values of
cohesion and peak friction angle indicating that, for the
most part, rotational and translational slides affecting the
Virginio area first developed under climatic and geomorphological conditions different from those observed
at present. For example, they initiated at higher slope
angles and, at present, they move essentially as
reactivations characterized by residual shear-strength
parameters. This assumption is confirmed by the fact
that new generation rupture surfaces are usually small in
size and shallow seated, mainly affecting artificial cuts
excavated in cohesionless terrains for both road construction and farming activities. As previously mentioned, rainfall is the most important cause of landslides
in this area, followed by fluvial erosion processes which
are very active where clayey terrain outcrops in the valley
bottom. Human activities are also important and are
mainly connected to agriculture practices.
In the Virginio area, falls of poorly cemented soils and
first-time debris slides are also present, mainly affecting
the steep slopes that develop as a result of geological (e.g.,
outcrop of sands and weakly cemented conglomerates)
or human (e.g., slope cutting, anthropogenic terraces)
causes. Falls and first-time debris slides may occur also
as a consequence of the increase in stress conditions associated with unloading, foot erosion, and retrogressive
activity of reactivated rotational slides located in the
middle part of the slopes.
The volumes of falls are usually small (largest volume
approximately 1,000 m3), with run-outs that are generally
limited to the height of the slopes (5–15 m), since
detached material simply piles up at the base of the slope
as a consequence of the abrupt change of slope gradient
produced by the change in lithology.
In summary: (a) The most important landslide types
and mechanisms can be grouped into three broad classes
in terms of destructive potential:
falls and first-time debris slides;
reactivation of earth slides; and
earth flows.
Figure 4. Pictures showing examples of mass movements affecting the Virginio River basin: (a) earth slide in a vineyard near Baccaiano; (b) first-time
debris slide near Marcialla.
208
Environmental & Engineering Geoscience, Vol. X, No. 3, August 2004, pp. 203–216
Landslide Susceptibility Assessment
Table 3. Results of geotechnical tests on the materials sampled on the
Sant’Appiano landslide (after Caioli, 1993).
Lithology
Ps
Ps-Pag
Pag
/p
34
25
23
/r
22
17
9
Wp
15–18
20–25
25–30
W1
20–25
25–35
50–70
Table 4. Intensity levels associated with landslide types in the Virginio
River basin.
Intensity
Level
I3
/p ¼ Peak friction angle; /r ¼ residual friction angle; Wp ¼ plastic
limit; Wl ¼ liquid limit.
I2
(b) Correct recognition and interpretation of landslide
types allows the realization of an a priori attribution of
intensity levels as a function of observed historical
damages (Table 4). (c) Analysis of the landslide
inventory, through a frequency analysis of factors such
as lithology, slope angle, and land cover of landslide area
(Figure 5), allows us to outline the boundaries of firstorder LPFs associated with each landslide mechanism
(Table 5). (d) During this phase of the analysis of the
inventory it was also possible to select the most relevant
second-order LPF and slope instability indicators or
factors (SIF) for each landslide mechanism (Table 5). A
detailed field survey of a sample of 32 landslides
randomly selected from among the 285 contained in the
inventory was performed. Analysis of this sample
allowed us to identify the most important second-order
LPF for each landslide type, resulting in the following
phase of the susceptibility mapping at local scale. Three
main categories of second-order LPFs were recognized:
lithotechnical, hydrogeological, and human. Table 6
reports some of the results of this field activity. Particular
attention was devoted to the interpretation of lithotechnical features. Following the proposal by Canuti and
others (1982), lithologies, outcropping in correspondence
with the crown of the landslides, have been subdivided on
the basis of lithological assumptions and subsequently
classified following the Unified Soil Classification
System of the U.S. Army Corps of Engineers (Table 7).
This operation enables us to standardize the procedures,
making the adopted methodology eventually applicable
to other geological and geomorphological settings.
Identification of Macroareas
This phase of the investigation, starting from the
analysis of the landslide inventory, has the purpose of
recognizing, for each possible landslide mechanism,
the stable portions of the study area, in order to focus
investigations at slope scale in the rest of the basin area.
Analysis of the inventory allows selection of first-order
preparatory factors that can be associated with the
presence of landslides.
Once identified, the classes of first-order LPFs are
reciprocally intersected by vector GIS overlay techniques
I1
Associated Damages
Rapid or very rapid landslides that
provoke destruction or serious
damage to buildings and
transportation routes. Possible
loss of human lives.
Movements with velocities ranging
from moderate to rapid, capable
of causing the destruction of
permanent installations and
structures. Possible loss
of human lives.
Slow-moving landslides. Usually
those events do not have
consequences on human life and
can be sustainable for the
territory in which they occur.
Effects usually can be controlled
from the realization of
construction work.
Landslide
Mechanism
First-time
debris slides
and earth falls
Reactivated
earth slides
Earth flows
leading to the definition, for the entire basin, of map units
corresponding to the concept of Unique Condition Unit
(UCU) (Carrara et al., 1995). Such units are homogeneous, starting from first-order LPFs, and they result from
the combination of only those classes of factors that can be
associated with landslides. UCUs are then amalgamated to
form macroareas (Figure 6), a sort of spatial enlargement
of a single UCU, which are considered to maintain the
same properties of the original UCU from which they have
been generated. The production of macroarea boundaries
has been created by delimiting, starting from the Digital
Terrain Model (DTM), the catchments with the lowest
order (Carrara et al., 1995), a practice obviously controlled
by the DTM resolution. Then, an intersection of the
catchment layer and the UCU is performed in order to
exclude all the catchments in which the UCUs were
absent. In this phase it is assumed that ‘macroareas’ have
the same properties as the UCU from which they have
been generated. The aim of this operation is to outline for
each landslide type the parts of the territory that may
potentially generate landslides, while the rest of the areas
can be associated with stable conditions with respect to the
considered landslide type. In this way, the extent of the
territory to be subjected to further investigations at slope
scale can be diminished, thereby excluding from the
analysis areas not considered to be prone to landsliding
starting from first-order LPFs.
At the end of the process, different sets of macroareas,
corresponding to the number of landslide types, are
produced. In summary, the macroareas are characterized
by the following properties:
Environmental & Engineering Geoscience, Vol. X, No. 3, August 2004, pp. 203–216
209
Casagli, Catani, Puglisi, Delmonaco, Ermini, and Margottini
Figure 5. Input data: (a) land cover map, (b) lithological map, (c) slope angle map, and (d) landslide inventory map.
all the inventoried landslides are located inside
macroareas;
macroareas without landslides may exist: in these areas
first-time landslides may be triggered by the presence
210
of second-order LPFs; in addition, they can contain
landslide areas not recognizable or obliterated because
of intense human activity;
the landslide susceptibility outside macroareas is zero;
Environmental & Engineering Geoscience, Vol. X, No. 3, August 2004, pp. 203–216
Landslide Susceptibility Assessment
Table 5. Classes of first-order landslide instability factors associated
with landslides in the Virginio River basin.
Landslide
Types
Slope Angle
Classes
Falls and
first-time
debris slides
Reactivated
earth slides
308–908
Earth flows
Land
Cover
Lithology
Pcg, Pcg-Ps, Ps
158–308
Forest, vineyard,
bush, meadows
Pcg-Ps, Ps, Ps-Pag Vineyard, bush,
meadows,
cultivated land
Ps, Ps-Pag
Vineyard, bush,
meadows,
cultivated land
108–208
complete, especially in areas with intense anthropogenic activity, macroareas are simply zones in which
the combination of preparatory factors makes it
possible that undiscovered or obliterated landslides
(capable of reactivation) may be present.
for first-time failures, macroareas represent zones with
a combination of LPFs that favor the possible
formation of new slides; and
for reactivated landslides, macroareas have a different
meaning since, if the inventory were complete,
macroareas would exactly coincide with landslide
boundaries. Since no inventory can be considered
Investigations at Hillslope Scale
With reference to Figure 3, the central part of the application of the methodology is represented by investigations at local hillslope scale (1:10,000–1:2,000). Each
macroarea is the object of a field survey aimed at
mapping second-order LPFs (Table 2) and SIIs (Table 8).
Then landslide factors are transformed into vector-based
Boolean variables of presence (1) and absence (0),
following the same approach proposed by Chung and
others (1995) in a multivariate statistical analysis. The
intersection of second-order LPFs by GIS techniques,
allows us to single out second-order UCUs that represent
the terrain units for mapping the landslide susceptibility
(within each set of macroareas).
The transformation of second-order LPFs into Boolean
variables represents a crucial step for application of the
Table 6. Results of the investigation at basin scale regarding second-order preparatory factors for reactivation of rotational and translational slides.
See Table 7 for the meaning of abbreviations.
First-Time Debris Slide—Earth Falls
Frequency
Reactivated Earth Slides
Frequency Earth Flows
Frequency
Geomorphological
Concurrence with other
mass movements
Longitudinal convex-concave profile
Fluvial erosion at the slope toe
10
3
3
Creep
10
Concurrence with other
mass movements
Concave profile
Longitudinal
concave-convex profile
Fluvial erosion at the
slope toe
Erosion along the
landslide flanks
5
5
7
4
Concurrence with other
mass movements
Creep
10
5
Longitudinal concave-convex profile
Variation in the longitudinal profile of
the river cuts
Fluvial erosion at the slope toe
7
7
4
4
Hydrogeological
Contrast of permeability
Lithotypes subject to liquefaction
2
2
Presence of aquifers
Seepage and spring levels
2
2
Contrast of permeability
Water ponds on the
landslide surface
Seepage and spring levels
Anomalous drainage pattern
4
4
Watering-dewatering
Contrast in permeability
5
4
3
3
3
3
Softening
Variation in water
drainage capacity
3
3
Seepage and spring levels
Temporaneous superficial
flow accumulation
Water ponds on the landslide surface
5
Artificial cuts
2
4
5
3
1
SC-CL
CL
2
6
3
3
Human
Artificial cuts
4
Artificial cuts
SC-CL
GC-CL
GW
4
3
8
SC-CL
CL
GC-CL
GW
Geotechnical
Environmental & Engineering Geoscience, Vol. X, No. 3, August 2004, pp. 203–216
211
Casagli, Catani, Puglisi, Delmonaco, Ermini, and Margottini
Table 7. Classification of the lithologies from the Virginio River basin
on the basis of the USCS.
Soil Classes
Description
USCS
Pcg
Pcg-Ps
Ps-Pag
Pag-Ps
Pag
Weakly indurated conglomerates
Gravels with sands
Sands with silt and clay
Silty clay and sands
Silty clay
GW
GC-CL
SC-CL
CL
CH
USCS ¼ Unified Soil Classification System
presented methodology, allowing the subdivision of the
territory as a function of the number of LPFs and SII that
are present in each second-order UCU. This is the basic
requirement for the successive setting up of the landslide
susceptibility function. For this reason, particular attention has been focused on detecting and mapping local
instability indicators that have been represented as point
features on the GIS. Table 8 shows a checklist of local
SIIs slightly modified from what was proposed by
Crozier (1984) for the reconnaissance of the state of
activity of landslides.
Susceptibility Function
The last part of the analysis is aimed at setting up a
function for the landslide susceptibility assessment.
Taking into account the geomorphological settings of
the study area, a logic function was chosen, based on the
number of second-order LPFs and SIIs that are present in
each second-order UCU.
The first step is constituted by the a priori definition
of the number of classes for performing the landslide
susceptibility ranking (relative hazard). In this application, we decided to utilize four classes, following the
requirements of the law of the Italian Republic 267/98.
Different methods for combining second-order LPFs and
SIIs on the basis of different susceptibility classes could
be adopted as well; they are almost the same as those
proposed in the literature for the landslide susceptibility
analysis at basin scale (Carrara and Guzzetti, 1995;
Aleotti and Chowdhury, 1999). However, data gathered
by the analysis at slope scale are usually not sufficient for
carrying out deterministic approaches, whereas statistical
methods are subject to the previously mentioned problems associated with the coupling in the same analysis of
categorical, discrete, numerical variables. For this reason,
a logic function is preferred and was based on binary
switches of a checklist of second-order LPFs and SIIs.
Table 9 describes in detail the utilized landslide
susceptibility function:
(S3): areas at very high susceptibility containing landslides classified as active or dormant following Cruden
and Varnes (1996). As previously mentioned, landslides affecting the Virginio basin essentially move as
212
Figure 6. Map showing the envelopment of UCU to form ‘macroareas’
for the three different types of landslides; the gray dotted line
highlights the basin portion to which the three ‘macroareas’ selected
for the methodology description belong.
reactivations. It was assumed that areas containing
active or dormant landslides are more hazardous
because of their higher frequency of activation (in
time). Furthermore, it was also decided to include in
this class the areas not strictly involved in landslides,
but in which at least three SIIs were recognized. In this
way, SIIs are managed as elements precursory to the
movement;
(S2): areas in which second-order LPFs coexist with at
least one SII can be considered as moderately susceptible because they are a measure of how the slope is
prone to landslides;
(S1): areas classified as low susceptibility are characterized by the presence of only second-order LPFs; and
(S0): areas placed outside the borders of the macroareas can be considered to have no susceptibility to
landsliding.
It should be noted that there are high subjectivity levels
associated with the application of an experience-driven
geomorphological approach to landslide susceptibility
analysis, which may be misleading if the final goal is the
physical comprehension of the rules that control the
triggering and development of landslides. The choice of
a logic function was preferred in order to minimize the
uncertainties derived from the application of the methodology. As a general rule, great emphasis was given to
Environmental & Engineering Geoscience, Vol. X, No. 3, August 2004, pp. 203–216
Landslide Susceptibility Assessment
Table 8. Checklist of local slope instability indicators (SIIs) considered
in the analysis (modified after Crozier, 1984).
ID
Local Slope
Instability Indicator
1
2
3
4
5
6
7
8
Crevices
Trenches
Local failures
Bulges
Counterslopes
Settlements
Fissures on main infrastructures
Tilting of vegetation
the interpretation of local SIIs that represent the most
important, and maybe unequivocal (when physical terrain
properties are not enough for adopting deterministic
models), tool for understanding how a certain area is
potentially prone to landslides. The proposed function
can be considered a conceptual scheme exportable as well
to other geographical contexts that show similar characteristics, starting from the processes that are responsible
for landscape evolution. For instance, it can be assumed
that its direct application, in the present form, is
recommended only for Neogene basins belonging to the
Northern Apennines, where landslides that move as
reactivations on deep-seated surfaces prevail, while it
should be adapted in the case of areas mostly affected by
first-generation, shallow-seated landslides.
RESULTS
An outline of the results gathered by the application of
the previously mentioned landslide susceptibility function
is shown in Figure 7, which refers to the same area located
in the western middle portion of the Virginio River basin.
These results emphasize that the areas susceptible to falls
and first-time debris slides are the most dangerous in
terms of potential risk, since they are of high intensity and
overlay most of the urban areas and part of the main road
system located in the basin. In contrast, areas susceptible
to reactivation of rotational slides are sited along the
middle portion of the slopes, involving zones devoted to
agricultural practices and high-value typical crops and, in
some cases, isolated buildings and secondary road
networks are affected; the highest susceptibility for
reactivation of earth flows is concentrated in the lowest
portion of the slopes, sometimes affecting the main road
system located along the flood plain.
CONCLUSIONS
The assessment of landslide susceptibility represents
a first level of knowledge that is crucial and precursory
Table 9. Logic landslide susceptibility function.
Class
Susceptibility
Active or
Dormant
Landslide
S3
S3
S2
S1
S0
High
High
Moderate
Low
Negligible
Yes
No
No
No
No
Second-Order
Preparatory
Factors
Slope
Instability
Indicators
–
1
1
1
No
–
3
1
No
No
for the whole process that leads to landslide hazard
identification. In a highly urbanized country such as Italy,
a hazard analysis can be performed with adequate
accuracy only at a local scale. As a consequence, the
proposed methodology addresses a microzoning of susceptibility on the basis of detailed (1:10,000) standard
geomorphological investigations.
Such an approach, although resulting in very intense
and time-consuming collection of data through fieldwork,
permits an accurate identification of the different physical
and anthropogenic variables characterizing unstable
areas. Together with the adoption of first-order LPFs
and macroareas, it allows us detailed scale surveys, not
only on unstable areas but also on regions potentially
prone to first-generation phenomena, or in zones in which
active movements are obliterated or difficult to recognize
because of human activities.
The GIS mapping and the following transformation of
both first- and second-order LPFs into GIS vector-based
layers enables the upgrading of the various themes and,
by means of improvement of the susceptibility functions,
a better reliability of final maps. This feature is very important since many preparatory factors, especially those
associated with first-generation landslides, are variable in
time, determining a change in landslide susceptibility and
associated hazard.
In summary, some features resulting from the application of this kind of approach can be highlighted: (a)
ease of use and reapplication for end-users (local administrations, urban planners) without assistance of academic
landslide ‘experts’; (b) investigations and techniques are
differentiated according to the different landslide types;
(c) consistency with the principle that landslides should
be correctly interpreted and mapped at the scale at which
they occur. One of the most important problems
connected with the application of indirect mapping
methods comes from the scales at which the survey is
carried out. Several phenomena can be correctly
interpreted only at the slope scale and, in this case,
hazard levels are difficult to map correctly by indirect
mapping techniques when the scale is rather large
(1:10,000); (d) intensity of the events considered as
a function of the different landslide types; (e) different
Environmental & Engineering Geoscience, Vol. X, No. 3, August 2004, pp. 203–216
213
Casagli, Catani, Puglisi, Delmonaco, Ermini, and Margottini
Figure 7. (A) Susceptibility map for falls and first-time debris slides, (B) susceptibility map for reactivated earth slides, and (C) susceptibility map for
reactivated earth flows.
data management for first-time landslides and reactivations (which also have different intensities); (f) in order to
overcome limitations presented in item c, the methodology requires that investigations should be carried out at
two different scales. The basin scale analyses aimed at
selecting the portion of the territory that may be prone to
landsliding and analyses at the slope scale that allows us
the detail inside each macroarea of different landslide
susceptibility levels (inside each landslide level); and (g)
214
an important need for the correct use of this approach is
the availability of spatial terrain data. For most of Italy,
geological maps, land cover, and landslide inventories are
commonly available at scales ranging between 1:25,000
and 1:10,000, as well as a Digital Terrain Models, with
a 20 m/pixel resolution.
The analysis carried out for the case study of the
Virginio River basin divided the landslide susceptibility
into four classes (S0, S1, S2, S3). The adopted sus-
Environmental & Engineering Geoscience, Vol. X, No. 3, August 2004, pp. 203–216
Landslide Susceptibility Assessment
ceptibility function is based on logic operators and results
from the assessment of the presence/absence of actual
geological and geomorphological evidence, such as active
or dormant landslides and local SIIs. For such reasons, by
adequately varying the input parameters, the susceptibility
function can be adapted to other situations, close to the
Virginio River basin, starting from the geological and geomorphological point of view.
ACKNOWLEDGMENTS
The authors want to thank Maia L. Ibsen for the careful
review of the original manuscript and the referees Gerald
F. Wieczorek, Robert J. Watters, and Gary E. Christenson
for their valuable comments. The authors are grateful to
all the colleagues who were involved in the joint project
between the Earth Sciences Department, University of
Florence, and ENEA C.R. Casaccia–Rome: Luca Falconi,
Gennaro D’Agostino, Lorenzo Codebò (geological and
geomorphological survey); Vladimiro Verrubbi and
Stefano Amato; Alessandra Bevilacqua (geomorphological survey and GIS elaboration); Daniele Spizzichino
(geotechnical survey and GIS elaboration); Salvatore
Martino (geotechnical survey); Manuela Ruisi (hydrogeological survey); Rosanna Foraci and Gabriele Leoni
(GIS elaboration); and Alberto Iotti and Ugo Tarchiani
(landslide inventory). The project was supported by the
Basin Authority of Arno River, to whom the authors are
grateful for their collaboration and data accessibility.
REFERENCES
ALEOTTI, P. AND CHOWDHURY, R., 1999, Landslide hazard assessment:
Summary review and new perspectives: Bulletin Engineering
Geology and Environment, Vol. 58, pp. 21–44.
AMANTI, M.; CASAGLI, N.; CATANI, F.; D’OREFICE, M.; AND MOTTERAN,
G., 1996, Guida al Censimento Dei Fenomeni Franosi e Alla
Loro Archiviazione (versione 1.0): Miscellanea 7, a cura della
Presidenza del consiglio dei ministri, Dipartimento dei Servizi
Tecnici Nazionali, Servizio Geologico, CNR-GNDCI, Linea 2,
Ist. Pol. Zecca dello Stato, Roma, 109 pp.
AMATO, S., 2001, Caratterizzazione Idrologico Erosiva Dei Suoli del
Bacino del Torrente Virginio: Università degli Studi di Firenze.
Corso di Laurea in Scienze Geologiche. Anno accademico 1999–
2000. Tesi di laurea inedita, 153 pp.
ARDIZZONE, F.; CARDINALI, M.; CARRARA, A.; GUZZETTI, F.; AND
REICHENBACH, P., 2002, Impact of mapping errors on the
reliability of landslide hazard maps: Natural Hazards and Earth
System Sciences, Vol. 2, pp. 3–14.
BRABB, E. E., 1984, Innovative approaches to landslides hazard
mapping: Proceedings of the Fourth International Symposium on
Landslides, Toronto, Vol. 1, pp. 307–324.
CAIOLI, C., 1993, Studio Geomorfologico e Geologico Tecnico di
un’area di Barberino Val d’Elsa: Tesi di laurea inedita,
Università degli Studi di Firenze, 135 pp.
CANUTI, P. AND CASAGLI, N., 1996, Considerazioni sulla valutazione
del rischio di frana. Estratto da ‘‘Fenomeni Franosi e Centri
Abitati’’: Atti del Convegno di Bologna del 27 Maggio 1994.
CNR-GNDCI, Linea 2, Previsione e Prevenzione di Eventi
Franosi a Grande Rischio. Pubblicazione no. 846, 57 pp.
CANUTI, P.; FOCARDI, P.; GARZONIO, C. A.; RODOLFI, G.; VANNOCCI, P.;
BOCCACCI, L.; FRASCATI, F.; MORETTI, S.; AND SALVI, R., 1982,
Stabilità dei Versanti Nell’area Rapprentativa di Montespertoli:
C.N.R., P.F. Conservazione del Suolo, S.P. Fenomeni Franosi
U.O. no. 18. Università degli Studi di Firenze, Istituto di Geologia
e Paleontologia. Tipografia S. EL. CA. Firenze.
CANUTI, P.; GARZONIO, C. A.; AND RODOLFI, G., 1979a, Dinamica
morfologica di un ambiente soggetto a fenomeni franosi e ad
intensa attività agricola. Area rappresentativa di Montespertoli,
Firenze: Annali Istituto Sperimentale Studio e Difesa del Suolo,
Vol. 10, pp. 81–102.
CANUTI, P.; GARZONIO, C. A.; AND RODOLFI, G., 1979b, The influence of
agricultural activity on slope stability: An example from
Montespertoli (Tuscany, Italy) representative area: Proceedings
of the Symposium of the International Association Engineering
Geology, Warsavia, pp. 195–203.
CANUTI, P.; PRANZINI, G.; AND SESTINI, G., 1966, Provenienza e
ambiente di sedimentazione dei ciottolami del Pliocene di San
Casciano (Firenze): Mem. Society Geol. It., Vol. 5, pp. 340–364.
CARRARA, A., 1983, Multivariate methods for landslide hazard
evaluation: Math. Geology, Vol. 15, No. 3 pp. 403–426.
CARRARA, A., 1984, Landslide hazard mapping: Aims and methods:
Association Francaise de Géographie Physique. Colloque de
CAEN, pp. 141–151.
CARRARA, A., 1989, Landslide hazard mapping by statistical methods:
A ‘‘black-box’’ model approach: Proceedings of the Workshop
on Natural Disasters in European-Mediterranean Countries,
Perugia, CNR-US NFS, pp. 427–445.
CARRARA, A.; CARDINALI, M.; DETTI, R.; GUZZETTI, F.; PASQUI, V.; AND
REICHENBACH, P., 1990, Geographical Information System and
multivariate models in landslide hazard evaluation, in Proceedings of the Sixth International Conference Field Workshop on
Landslides, Switzerland-Austria-Italy, Vol. 1, pp. 17–28.
CARRARA, A.; CARDINALI, M.; DETTI, R.; GUZZETTI, F.; PASQUI, V.; AND
REICHENBACH, P., 1991, GIS techniques and statistical models in
evaluating landslide hazard: Earth Surface Processes and Landforms, Vol. 16, pp. 427–445.
CARRARA, A.; CARDINALI, M.; GUZZETTI, F.; AND REICHENBACH, P., 1995,
GIS techniques in mapping landslide hazard. In Carrara, A. and
Guzzetti, F. (Editors), Geographical Information Systems in
Assessing Natural Hazards: Kluwer Academic Publishers,
Dordrecht, The Netherlands, pp. 135–175.
CARRARA, A.; CATALANO, E.; SORRISO VALVO, M.; REALI, C.; AND OSSO,
I., 1978, Digital terrain analysis for land evaluation: Geologia
Applicata e Idrogeologia, Vol. 13, pp. 69–127.
CARRARA, A. AND GUZZETTI, F., 1995, Geographical Information Systems in Assessing Natural Hazards: Kluwer Academic Publishers,
Dordrecht, The Netherlands, 353 pp.
CHUNG, Ch. F.; FABBRI, A. G.; AND VAN WESTERN, C. J., 1995, Multivariate regression analysis for landslide hazard zonation. In
Carrara, A. and Guzzetti, F. (Editors), Geographical Information
System in Assessing Natural Hazards: Kluwer Academic
Publishers, Dordrecht, The Netherlands, pp. 107–142.
CROZIER, M. J., 1984, Field assessment of slope instability. In
Brunsden, D. and Prior, D. (Editors), Slope Instability: John
Wiley and Sons, Chichester, pp. 103–140.
CRUDEN, D. M. AND VARNES, D. J., 1996, Landslide types and processes. In Turner, A. K. and Schuster, R. L. (Editors), Landslides:
Investigation and Mitigation: Sp. Rep. 247, Transportation
Research Board, National Research Council, National Academy
Press, Washington, DC, pp. 36–75.
EINSTEIN, H. H., 1988, Special lecture: Landslide risk assessment
Environmental & Engineering Geoscience, Vol. X, No. 3, August 2004, pp. 203–216
215
Casagli, Catani, Puglisi, Delmonaco, Ermini, and Margottini
procedure, in Proceedings of the Fifth International Symposium
on Landslides, Lausanne, Vol. 2, pp. 1075–1090.
GUZZETTI, F.; CARRARA, A.; CARDINALI, M.; AND REICHENBACH, P.,
1999, Landslide hazard evaluation: A review of current
techniques and their application in a multi-scale study, Central
Italy: Geomorphology, Vol. 31, pp. 181–216.
MARTINI, P. I.; SAGRI, M.; AND COLELLA, A., 2001, Neogene-quaternary
basins of the Inner Apennines and Calabrian arc, Italy. In Vai,
G. B. and Martini, I. P. (Editors), Anatomy of an Orogen—The
Apennines and Adjacent Mediterranean Basins, Kluwer Academic Publishers, United Kingdom, pp. 375–399.
MERLA, G., 1952, Geologia dell’Appennino settentrionale: Bollettino
Società Geologica Italiana, Vol. 70, pp. 95–382.
MONTGOMERY, D. R. AND DIETRICH, W. E., 1994, A physically based
model for the topographic control on shallow landsliding: Water
Resources Research, Vol. 30, No. 4 pp. 1153–1171.
SCHUSTER, R. L., 1996, Socio-economic significance of landslides.
In Turner, A. K. and Schuster, R. L. (Editors), Landslides:
Investigation and Mitigation: Sp. Rep. 247, Transportation
Research Board, National Research Council, National Academy
Press, Washington, DC, pp. 12–31.
216
SOETERS, R. AND VAN WESTEN, C. J., 1996, Slope stability recognition,
analysis, and zonation: Application of Geographical Information
System to landslide hazard zonation. In Turner, A. K. and
Schuster, R. L. (Editors), Landslides: Investigation and Mitigation: Sp. Rep. 247, Transportation Research Board, National
Research Council, National Academy Press, Washington, DC,
pp. 129–177.
SORRISO VALVO, M., 2002, Landslides: From inventory to risk. In
Rybář, J.; Stemnerk, J.; and Wagner, P., Landslides, Proceedings
of the IECL, Prague, Czech Republic, June 24–26, 2002,
Balkema, The Netherlands, pp. 79–93.
TERLIEN, M. T. J.; VAN WESTEN, C. J.; AND VAN ASCH, T. W. J.,
1995, Deterministic modelling in GIS-based landslide hazard
assessment. In Carrara, A. and Guzzetti, F. (Editors), Geographical Information System in Assessing Natural Hazards:
Kluwer Academic Publishers, Dordrecht, The Netherlands, pp.
57–77.
VARNES, D. J. IAEG COMMISSION ON LANDSLIDES, 1984, Landslide
Hazard Zonation—A review of Principles and Practice:
UNESCO, Paris, 63 pp.
Environmental & Engineering Geoscience, Vol. X, No. 3, August 2004, pp. 203–216