Intrinsic Environmental Vulnerability as Shallow Landslide Susceptibility in Environmental Impact Assessment

: This work investigated the susceptibility factors that trigger shallow landslides. In particular, the objective of the research was the implementation of a method to determine the relevant factors that can trigger shallow landslide events. However, with respect to the existing methods, the integration with historical datasets and the inclusion of spatial factors displaying dynamics in the same characteristic timescales were specific features of the developed tool. The study area included the watersheds of the Sessera and Strona rivers in the alpine area of the Province of Biella (Piedmont, NW Italy). The method was developed and tested from two sub-datasets derived from an integrated dataset that referred to an intense event, involving the same area, that occurred in 1968 (2–3 November). This allowed the implementation of an integrated representation of landslides’ predisposing factors and the identiﬁcation and classiﬁcation in di ﬀ erent groups of the areas susceptible to geo-hydrological instability processes. The previously existing databases were veriﬁed and integrated into a geographic information system (GIS) environment, giving a potentially sharable source of information for planning purposes. The obtained maps represent a metric of one of the possible intrinsic environmental vulnerability factors for the area under study. Consequently, this method can represent a future instrument for determining the intrinsic environmental vulnerability dependent on landslides within an environmental impact assessment (EIA), as required by the most recent European regulation on EIA. Moreover, the shared information can be used to implement informed policy and planning processes, based on a bottom-up approach. In particular, the availability online of landslide susceptibility maps could support the generation of augmented information—useful for both local administrators and planners as well as for stakeholders willing to implement speciﬁc projects or infrastructure in vulnerable areas, such as mountains.


Reference Method
The present method was based on a previous approach [19], which is currently the most widely applied due to its performance and possibility of being converted into appropriate spatial maps of landslide susceptibility [20][21][22].
For this method, a set of factors considered to be the triggers of the observed instability are first selected (e.g., [23]). Then, their interplay is considered in assessing the landslide susceptibility in a given area [24]. In particular, for each analyzed factor viewed as a landslide trigger, a set of variability classes is defined, in order to evaluate their relative impact on the overall instability phenomenon. The relative impact factor (IF) is weighted against the abundancy of landslides within each class.
Considering the existence of j different classes, each of them has a relative IF, indicated as IF rel,j . The weight for the j-th class is indicated with w j . Then, the total impact factor, related to landslide susceptibility, indicated as IF tot , is defined as IF rel,j × w j (1) being IF rel,j = IF i IF i(max) × 100 (2) and

Definition of the Case Study
The study area, about 150 km 2 wide, extends from the southern border of the Pennine Alps (NW Italy) to the hills (granitic massif of Biella). The study area is delimited by the drainage basins of the Strona and Sessera rivers, both located in the northern part of the Piemonte region (NW Italy). The Po Valley and the Pennine Alps, in the sub-sector of Biellesi and the Cusiane Alps [25], constitute a further geographical limitation of this area. Such an area exhibits a characteristic central relief, known as the Rubello-Prapiano-Massaro massif, which forms a natural barrier between the alpine and flatland zones. The massif, ranging from west to east between 1400 m and 1200-1300 m, respectively, stretches in parallel along two main fault lines, the Cremosina and the Insubric lines. The latter separates the Pennine Alps (NW) from the Meridional Alps (SE). Figure 1 shows a map of the main geographical and geological features of the study area. The bedrock is prevailingly made of leucocratic gabbro and coarse-grained pyroxene-amphibole gabbro, having a locally parallel texture together with lenses of finely grained ultramafic rocks, often converted to diaftorites, being dark green to blackish-colored. Besides the characteristic granite, biotite-rich diorite also occurs with minor amounts of encrusted muscovite.
An irregularly thick weathered layer covers the granite, into which the middle sector of the Strona river basin is incised, that is densely covered with woodland. Both pedogenic and vegetation products are the basic components for detrital production. In fact, they are found in the main constituents of mass movements, generated over fully saturated slopes in the case of prolonged and intense rainfall.
Previous on-site monitoring activities gave an insight into how, apart from the mother-rock characteristics, the soil cover depth in this area is influenced by the landscape slope and morphology. In particular, in the case of a deeper soil thickness in correspondence to less-inclined slopes, these areas were shown to be subjected less to the erosive power of the running waters as well as to gravitydriven soil losses. Moreover, the soil cover depth decreases with increased elevations and more marked reliefs, where outcrops are more likely to occur.

Analyzed Susceptibility Factors
The historical analysis of instability processes in the study area [26] evidenced that soil slips and debris flows were the most frequent [27,28]. Considering the existing literature [29], the most The bedrock is prevailingly made of leucocratic gabbro and coarse-grained pyroxene-amphibole gabbro, having a locally parallel texture together with lenses of finely grained ultramafic rocks, often converted to diaftorites, being dark green to blackish-colored. Besides the characteristic granite, biotite-rich diorite also occurs with minor amounts of encrusted muscovite.
An irregularly thick weathered layer covers the granite, into which the middle sector of the Strona river basin is incised, that is densely covered with woodland. Both pedogenic and vegetation products are the basic components for detrital production. In fact, they are found in the main constituents of mass movements, generated over fully saturated slopes in the case of prolonged and intense rainfall.
Previous on-site monitoring activities gave an insight into how, apart from the mother-rock characteristics, the soil cover depth in this area is influenced by the landscape slope and morphology. In particular, in the case of a deeper soil thickness in correspondence to less-inclined slopes, these areas were shown to be subjected less to the erosive power of the running waters as well as to gravity-driven soil losses. Moreover, the soil cover depth decreases with increased elevations and more marked reliefs, where outcrops are more likely to occur.

Analyzed Susceptibility Factors
The historical analysis of instability processes in the study area [26] evidenced that soil slips and debris flows were the most frequent [27,28]. Considering the existing literature [29], the most meaningful triggering factors that should be included in the vulnerability assessment are lithology; morphology, concavity, and convexity of slopes; slope dipping; slope aspect; distance from water bodies; and land use. These factors were selected, depending on the availability of useful databases for their characterization. In particular, the morphological and lithological characteristics, the conformation of the drainage network, and the land use of the area under study served as a basis for this research. In order to define landslide susceptibility as a function of these parameters, an inventory of historical landslides was realized. In parallel, the data, necessary for characterizing the predisposing factors, were searched for and found. Original data were processed through geographical information system (GIS) software, as previously done in another work [30], to create a set of thematic maps, with each map corresponding to a single instability factor to be analyzed. Table 1 shows the list of databases used as a basis for the analysis of instability predisposing factors. The list of derived GIS thematic maps is also given. The obtained thematic maps were re-classified through a comparison with a selected sub-set of landslide data, centered on the same area, in order to maximize the correspondence (LSI corr ) between the landslide density for a single class and the value of the corresponding landslide susceptibility index (IF rel,j ). In particular, a test dataset, i.e., a group of data used for evaluating the susceptibility inside the area of study, was extracted out of the landslide historical inventory. Another sub-set, belonging to the same inventory, was considered as a verification dataset called the test set. This test set was used, after the parameterization of susceptibility, to verify, re-calibrate, and validate the estimation method in use. Finally, a landslide susceptibility map was generated, covering a 25 km 2 area, centered on the upper Strona river basin and a part of the Ponzone river, a tributary of the Sessera Valley in the opposite slope (see Figure 1). Furthermore, the study was divided into the following phases: construction of the landslide inventory, data search for the characterization of the predisposing factors, construction of thematic maps suitable for the characterization of predisposing factors, and intersection of information layers, summary of results, and graphic restitution.
During the first step, historical data were collected to re-construct the past landslide events. Historical analysis is a key tool when studying the problems connected to slope instability and stream flood processes. In fact, it allows (a) the detection of areas that, due to geo-morphological and geo-structural conditions, land use, and climatic regime, appear as the most sensitive to the triggering of processes, thus allowing the possible factors leading to instability to be detected, (b) the identification of the laws, by which the susceptibility factors exert their control in triggering the conditions of the instability processes, through a back-analysis procedure, and (c) the projection into the future of the acquired historical information and, by this way, the possible risk scenarios arising from man-made land modifications can be identified. This occurs according to the principle that the instability processes are liable to occur in areas historically affected by similar processes [26,[31][32][33][34][35]. The reconstruction of the effects induced by historical floods and gravitational processes implied four steps: (1) search of data sources, (2) acquisition of the information and cartographic and/or photographic documents, (3) data selection, analysis, and validation, and (4) GIS transposition of the data.
The data collection was based on the examination of historical sources (e.g., newspaper chronicles, bibliographies, technical manuscripts, and all kinds of inherent documents) preserved in the archives of the Italian National Research Council -Institute for Research on Hydrogeologic Risks (CNR-IRPI) as well as in the archives of some municipalities belonging to the area under study.
The available data were represented in a GIS environment (using the Technical Regional Map, at 1:10,000 scale, as the base map) to draw a geo-referenced database, which was useful for further spatial analyses. Every single identified process was represented either as a punctual or a polygonal element, depending on the detail of the available knowledge. The database records, combined to each geometrical feature, were composed by a group of attributes, which define the typology, geographical location, and triggering period of the process.
The historical analysis allowed the listing of 148 relevant rainfall events, since 1825, which involved 15 municipalities in the eastern part of Biella Province, Piemonte region, NW Italy (Table 2). Single events may have caused the triggering of more than one related process. This is why the number of recorded rainfall events and the number of instability processes does not coincide.  The recorded historical events showed a greater susceptibility to geo-hydrological effects (mainly landslides and floods) in the spring/early summer (mainly in May) and autumn (mainly in October and November), corresponding to the year's rainiest periods for this area.

The Test Event
During the first days of November 1968, a heavy rainfall event was recorded in the area under study. In some sectors, it exhibited exceptional characteristics, which caused the flooding of countryside and several inhabited areas. The consequences were catastrophic, especially in the valleys upstream of the Biella municipality (Strona, Sessera, and Cervo rivers), a rich industrial area in NW Italy for the many textile companies.
The exceptionality and persistence of rainfall in some areas was such that in many localities the historical highs recorded in the thirty years between 1921 and 1950 were exceeded. The average rainfall amounts in the considered basins during the event were double or triple that of the corresponding historical average rainfall for the same month (i.e., November). In particular, on November 2, peaks, with an intensity of 60 mm/h, were recorded in the afternoon. The isohyets map that refers to November 2, 1968 (from 16:00 to 22:00) is representative of the extraordinary nature of this pluviometric event [36]. Daily peaks of 220 mm were also recorded in the upper basin of the Strona and the Sessera streams, as visible from the pluviograms, which recorded significant rainfall intensities for the study area ( Figure 2). The recorded historical events showed a greater susceptibility to geo-hydrological effects (mainly landslides and floods) in the spring/early summer (mainly in May) and autumn (mainly in October and November), corresponding to the year's rainiest periods for this area.

The Test Event
During the first days of November 1968, a heavy rainfall event was recorded in the area under study. In some sectors, it exhibited exceptional characteristics, which caused the flooding of countryside and several inhabited areas. The consequences were catastrophic, especially in the valleys upstream of the Biella municipality (Strona, Sessera, and Cervo rivers), a rich industrial area in NW Italy for the many textile companies.
The exceptionality and persistence of rainfall in some areas was such that in many localities the historical highs recorded in the thirty years between 1921 and 1950 were exceeded. The average rainfall amounts in the considered basins during the event were double or triple that of the corresponding historical average rainfall for the same month (i.e., November). In particular, on November 2, peaks, with an intensity of 60 mm/h, were recorded in the afternoon. The isohyets map that refers to November 2, 1968 (from 16:00 to 22:00) is representative of the extraordinary nature of this pluviometric event [36]. Daily peaks of 220 mm were also recorded in the upper basin of the Strona and the Sessera streams, as visible from the pluviograms, which recorded significant rainfall intensities for the study area ( Figure 2). The most intense phase of the rain began at 17:00, within the area centered in the villages of Pettinengo and Camandona. A limited area of about 50 km 2 was hit by intense precipitation during just 4 h. Past analyses of both pluviograms and rainfall distribution graphs displayed two distinct As a consequence, a very high number of landslides occurred, mainly centered in the municipality of Trivero (eastern side of Biella Province, Piemonte region, NW Italy). Several erosion processes along the hydrographic network and extensively flooded areas gave the alluvial event a catastrophic dimension ( Figure 3). phases. The first, between 18:00 and 19:30, with precipitation peaks of 29 mm/30 min (Coggiola municipality, time slot from 18:00 to 18:30). The second stage of the event occurred between 20:00 and 21:00. During the last phase, the weather stations in the municipalities of Camandona, Pettinengo, and Trivero recorded precipitation intensities in the order of 60 mm/h. As a consequence, a very high number of landslides occurred, mainly centered in the municipality of Trivero (eastern side of Biella Province, Piemonte region, NW Italy). Several erosion processes along the hydrographic network and extensively flooded areas gave the alluvial event a catastrophic dimension (Figure 3). In the Strona Valley alone, 58 victims were recorded, while in the whole of the Piemonte region, more than 70 victims were recorded. The losses detected in this area reached exceptional levels. In fact, in some municipalities the production assets were 90% destroyed. Overall, in the Province of Vercelli, there were 376 affected buildings, of which 277 were in the Biella area alone, while around 13,000 people were left without work. Substantial damage was also suffered by the homes and the connection infrastructures (bridges, roads), so that for several hours many municipalities were isolated.
The distribution map of the landslides gave a global view of the numerous landslides that occurred in the Strona Valley during these days. By dividing the surface area of the Strona river catchment area into regular meshes of 1 km 2 , the maximum peak of 91 landslides/km 2 was reached (municipality of Valle Mosso). The most common forms of observed soil degradation processes were [37]: • Landslides developed along pre-existing discontinuity surfaces, mostly between the crystal-lined rock substrate and the eluvio-colluvial layer. This type of instability is usually observed on granitic sandy coulters. Landslides of this type have a prevalence of flattened solids, with a smooth sliding surface, represented by a variable inclination plane; • Landslides developed along sliding planes of neo formation. In this case, the failures are promoted by the progressive imbibition of the ground and other factors that lead to overcoming the resistance to the cutting. The phenomenon is typical of pasture or grassland, being favored by local slope variations, building works, or excavations. The movement occurs by the rotation and integral translation of the plate; • Gully erosion due to the flow of surface runoff water either on the bottom of usually dry valleys or on their sides; • Rapid earth flow.

Analysis of Instability Factors
The landslide hazard of an area, defined as the probability of a landslide event of a given intensity, is expressed by a given interval [38]. The mapped results describe the distribution of spatial In the Strona Valley alone, 58 victims were recorded, while in the whole of the Piemonte region, more than 70 victims were recorded. The losses detected in this area reached exceptional levels. In fact, in some municipalities the production assets were 90% destroyed. Overall, in the Province of Vercelli, there were 376 affected buildings, of which 277 were in the Biella area alone, while around 13,000 people were left without work. Substantial damage was also suffered by the homes and the connection infrastructures (bridges, roads), so that for several hours many municipalities were isolated.
The distribution map of the landslides gave a global view of the numerous landslides that occurred in the Strona Valley during these days. By dividing the surface area of the Strona river catchment area into regular meshes of 1 km 2 , the maximum peak of 91 landslides/km 2 was reached (municipality of Valle Mosso). The most common forms of observed soil degradation processes were [37]: • Landslides developed along pre-existing discontinuity surfaces, mostly between the crystal-lined rock substrate and the eluvio-colluvial layer. This type of instability is usually observed on granitic sandy coulters. Landslides of this type have a prevalence of flattened solids, with a smooth sliding surface, represented by a variable inclination plane; • Landslides developed along sliding planes of neo formation. In this case, the failures are promoted by the progressive imbibition of the ground and other factors that lead to overcoming the resistance to the cutting. The phenomenon is typical of pasture or grassland, being favored by local slope variations, building works, or excavations. The movement occurs by the rotation and integral translation of the plate; • Gully erosion due to the flow of surface runoff water either on the bottom of usually dry valleys or on their sides; • Rapid earth flow.

Analysis of Instability Factors
The landslide hazard of an area, defined as the probability of a landslide event of a given intensity, is expressed by a given interval [38]. The mapped results describe the distribution of spatial (geographical) landslide occurrence probability. Often, the temporal probabilistic factor is difficult to determine. This is why past studies have expressed the concept of susceptibility for the predisposition to landslides, described either in qualitative or quantitative terms, in a study area [23,39]. This method is based on two main assumptions: (a) the knowledge and mapping of historical events should be integrated in the study of an area and (b) slope phenomena are likely to occur in the future in areas previously affected by analogous events and be characterized by a set of similar parameters (e.g., topographical, lithological, and land use).
The method used for the determination of susceptibility is based on the identification and spatial characterization of a set of control factors, on the quantification of the spatial relations between those factors, and an archive of previous landslides [23]. This method assumes that future landslides will take place under the same conditions as those of the past. Moreover, it represents an application of Bayes theorem, according to which frequency of each map unit is similar to the landslide occurrence probability. Each instability frequency or density represents the probability of the occurrence of a future event.
First, a set of instability predisposing factors was defined (e.g., lithology, land use, and slopes). This set was divided into classes. Then, a relative index was calculated through overlapping the landslides map with each single-factor map. The flowchart related to the calculation process is represented in Figure 4. (geographical) landslide occurrence probability. Often, the temporal probabilistic factor is difficult to determine. This is why past studies have expressed the concept of susceptibility for the predisposition to landslides, described either in qualitative or quantitative terms, in a study area [23,39]. This method is based on two main assumptions: (a) the knowledge and mapping of historical events should be integrated in the study of an area and (b) slope phenomena are likely to occur in the future in areas previously affected by analogous events and be characterized by a set of similar parameters (e.g., topographical, lithological, and land use). The method used for the determination of susceptibility is based on the identification and spatial characterization of a set of control factors, on the quantification of the spatial relations between those factors, and an archive of previous landslides [23]. This method assumes that future landslides will take place under the same conditions as those of the past. Moreover, it represents an application of Bayes theorem, according to which frequency of each map unit is similar to the landslide occurrence probability. Each instability frequency or density represents the probability of the occurrence of a future event.
First, a set of instability predisposing factors was defined (e.g., lithology, land use, and slopes). This set was divided into classes. Then, a relative index was calculated through overlapping the landslides map with each single-factor map. The flowchart related to the calculation process is represented in Figure 4.

Lithology
Lithology is one of the most important factors in the control of slope stability. In fact, the type of bedrock substantially affects the mechanical properties of the materials forming the slopes, e.g., permeability and shear stress.
The basic lithological information was drawn from the technical cartography of the municipalities included in the area under study (map scales: 1:10,000 and 1: 5000). The content of such cartographic documents was checked on-site. The gathered information was geo-referenced,

Lithology
Lithology is one of the most important factors in the control of slope stability. In fact, the type of bedrock substantially affects the mechanical properties of the materials forming the slopes, e.g., permeability and shear stress.
The basic lithological information was drawn from the technical cartography of the municipalities included in the area under study (map scales: 1:10,000 and 1: 5000). The content of such cartographic documents was checked on-site. The gathered information was geo-referenced, digitized in vectorial format, and converted into raster format by adopting 10 × 10 m cells, corresponding to the detail of the base maps.
The acquired data showed that, out of the 8 lithological classes identified in the territory under study, the lithological units most widely determined around the landslide-prone sites were the gabbro, leucogranite (white granite), biotitic gneiss, and kinzigite rocks ( Figure 5). digitized in vectorial format, and converted into raster format by adopting 10 × 10 m cells, corresponding to the detail of the base maps.
The acquired data showed that, out of the 8 lithological classes identified in the territory under study, the lithological units most widely determined around the landslide-prone sites were the gabbro, leucogranite (white granite), biotitic gneiss, and kinzigite rocks ( Figure 5).

Morphology: Concavity and Convexity Slope
The slope curvature with respect to the maximum dipping (angle) is significant for stability and influences the overland flow and drainage of running waters. In particular, shallow landslides seem to occur preferably in topographical convergence zones, where favorable conditions occur for stagnation, an increase of pore pressure, and reduction of shear stress for loose materials [40]. In particular, the drainage in a concave slope leads to an increase in the water pressure within pores and an accumulation in the surface formations. This favors the occurrence of surface movements.
The morphology (distribution of the convex-concave shaped slope surfaces) of the study area was re-constructed through the GIS-based elaboration of the digital elevation model (DEM). The available DEM refers to the present-day slope conditions, even if not representative of the situation prior to the 1968 flood, and allows the determination of the rapidly evolving processes that control the slope morphology (trigger conditions for landsliding, masking or erasing the depositional forms by natural or anthropic agents). A morphology map was obtained, which also identified the watersheds and valley areas. All the watersheds, which can have important ground effects during intense rainfall events, were identified.

Slope
The dipping angle plays an important role for slopes. In fact, it is the control factor determining the velocity and impact energy of the sliding masses. The dip value defines, for a given exposed material along the slope, the 'distance' from the natural equilibrium conditions, expressed through the internal friction angle.
The dip-slope map was drawn from a GIS-based DEM elaboration. In the map, 6 slope angle intervals were defined as <15°, 15-20°, 20-25°, 25-30°, 30-35°, and >35°. In order to assess the areal landslide density inside each class and for calculating the relevant landslide susceptibility index, a

Morphology: Concavity and Convexity Slope
The slope curvature with respect to the maximum dipping (angle) is significant for stability and influences the overland flow and drainage of running waters. In particular, shallow landslides seem to occur preferably in topographical convergence zones, where favorable conditions occur for stagnation, an increase of pore pressure, and reduction of shear stress for loose materials [40]. In particular, the drainage in a concave slope leads to an increase in the water pressure within pores and an accumulation in the surface formations. This favors the occurrence of surface movements.
The morphology (distribution of the convex-concave shaped slope surfaces) of the study area was re-constructed through the GIS-based elaboration of the digital elevation model (DEM). The available DEM refers to the present-day slope conditions, even if not representative of the situation prior to the 1968 flood, and allows the determination of the rapidly evolving processes that control the slope morphology (trigger conditions for landsliding, masking or erasing the depositional forms by natural or anthropic agents). A morphology map was obtained, which also identified the watersheds and valley areas. All the watersheds, which can have important ground effects during intense rainfall events, were identified.

Slope
The dipping angle plays an important role for slopes. In fact, it is the control factor determining the velocity and impact energy of the sliding masses. The dip value defines, for a given exposed material along the slope, the 'distance' from the natural equilibrium conditions, expressed through the internal friction angle.
The dip-slope map was drawn from a GIS-based DEM elaboration. In the map, 6 slope angle intervals were defined as <15 • , 15-20 • , 20-25 • , 25-30 • , 30-35 • , and >35 • . In order to assess the areal landslide density inside each class and for calculating the relevant landslide susceptibility index, a slope angle was assigned corresponding to the position of the detachment niche for every landslide reported in the instability map.

Aspect
The slope's aspect influences the micro-climatic conditions and vegetation cover along the slope. Moreover, the slope aspect affects the weathering of bedrock and controls the pedogenesis [41]. The aspect map was obtained by processing the DEM of the area and sub-dividing it into 4 spatial classes (North, South, East, and West).

Distance from Watercourses
Streambank erosion processes can be important conditioning agents for instability. Moreover, it is supposed that the distance from the impluvia may be another key factor to characterize potential instability. The 'distance from the impluvia (watercourses)' map was derived from the hydrographic network map in vector format, which was produced by the Regione Piemonte (1:10,000 scale), and processed through the ArcGIS®software's (ESRI, Redlands, California, USA) 'path distance' function.
The map was sub-divided into 4 classes, grouping the areas as a function of distance from the impluvia and stream incisions. The fixed class thresholds were (1) 50 m, (2) 50-150 m, (3) 150-300 m, and (4) above 300 m.
The class of maximum landslide density included the buffer area between 50 and 150 m from the watercourses. The Landslide Susceptibility Index (LSI) was computed by considering the frequency of the various classes inside the study area, and the highest LSI was assigned to the areas located at a distance from watercourses farther than 300 m. This class, rather small with respect to the whole area under study (7.5% of the entire area), appears to be particularly extended in the surroundings of the Rovella mountain, which were severely hit by the instability processes that occurred during the 1968 event.

Land Use
Land use is an important factor in the definition of landslide susceptibility, because it influences the relationship between the permeable and impermeable surfaces exposed over the slopes. Moreover, it affects the rainwater flow behavior and the soil saturation. Land use data were obtained from the available aerial photographs relative to 1968. Photo-interpreted data were transposed on basic cartography and validated through land surveys in detail. Then, data were digitized in order to draw a land use map in raster format.
The map classified the areas according to 6 different uses: (1) wood, (2) thin wood, (3) grassland/ pastureland, (4) urban fabric, (5) sparse urbanization, and (6) industrial plants. Data showed that the majority (82%) of landslides that occurred during the 1968 event affected wood-covered slopes ( Figure 6). However, the current state of knowledge did not allow inference of the impact the wood cover had on the slope stability. Thus, the land use factor was not included in the LSI map elaboration.

Lithology
First, the relationship between stability and the lithological classes ( Figure 4)

Lithology
First, the relationship between stability and the lithological classes ( Figure 4) was considered, in order to determine the relevance of each layer class with respect to its areal extension. In particular, this relationship is expressed through the Landslide Susceptibility Index relative to each of the j-th factors being considered (LSI rel,j ) [19]. Then, the obtained values were re-classified according to a scale from 1 to 100 (100 maximum value) to define the susceptibility in relation to each class for the considered case study (Table 3). Then, the data were synthesized in the corresponding thematic map (Figure 7).

Morphology
Slope morphology, according to three categories (convex, concave, and plane), was based on the areal ratio between the observed unstable and stable surfaces ( Table 4). The relative landslide indexes were identified and mapped (Figure 8).

Morphology
Slope morphology, according to three categories (convex, concave, and plane), was based on the areal ratio between the observed unstable and stable surfaces ( Table 4). The relative landslide indexes were identified and mapped (Figure 8).

Slope
The majority (83%) of recorded landslides occurred within the slope class of 15-35°. This fact is in agreement with what was previously known in the literature [42,43]. Among the 4 slope classes, the distribution of landslides was almost uniform. The frequency of instability processes sharply

Slope
The majority (83%) of recorded landslides occurred within the slope class of 15-35 • . This fact is in agreement with what was previously known in the literature [42,43]. Among the 4 slope classes, the distribution of landslides was almost uniform. The frequency of instability processes sharply diminished for slopes lower than a 15 • dip (10% of the processes recognized). This depends on the general equilibrium conditions, which characterize low-angle slopes. Similarly, a marked reduction of the landslide occurrence (8%) emerged for slope angles above 35 • (Figure 9). As confirmed by on-site observations, this might be caused by the reduction of soil thickness, which is a function of slope-dip growth. In particular, colluvial cover decreases until the bare rock becomes exposed. Coherently, the amount of movable materials during heavy, erosive rainfall events declines until it is minimized. diminished for slopes lower than a 15° dip (10% of the processes recognized). This depends on the general equilibrium conditions, which characterize low-angle slopes. Similarly, a marked reduction of the landslide occurrence (8%) emerged for slope angles above 35° ( Figure 9). As confirmed by onsite observations, this might be caused by the reduction of soil thickness, which is a function of slopedip growth. In particular, colluvial cover decreases until the bare rock becomes exposed. Coherently, the amount of movable materials during heavy, erosive rainfall events declines until it is minimized. Based on the relationship between landslide and stable areas with respect to the fixed slope-dip classes, the relative 'Landslide Indexes' were computed. The inferred values showed that the majority of the landslide-prone areas (almost 50% of the identified processes) occurred in the dip interval between 20 and 30° ( Figure 10). Based on the relationship between landslide and stable areas with respect to the fixed slope-dip classes, the relative 'Landslide Indexes' were computed. The inferred values showed that the majority of the landslide-prone areas (almost 50% of the identified processes) occurred in the dip interval between 20 and 30 • (Figure 10). Based on the relationship between landslide and stable areas with respect to the fixed slope-dip classes, the relative 'Landslide Indexes' were computed. The inferred values showed that the majority of the landslide-prone areas (almost 50% of the identified processes) occurred in the dip interval between 20 and 30° ( Figure 10).

Aspect
An aspect value was assigned corresponding to the number of landslides detected in relation to its detected detachment niche. Data showed that the north-facing slopes were scarcely affected (5% of the cataloged processes) by landslides. Conversely, south-facing slopes contained the highest incidence of gravitational processes (48%).
North-facing slopes only accounted for 14% of the total surface in the study area. Instead, southand east-facing slopes were predominant. Considering the relationship between 'landslide' and 'no landslide' pixels, the greatest susceptibility was observed in relation to the west-and south-facing classes ( Figure 11).

Aspect
An aspect value was assigned corresponding to the number of landslides detected in relation to its detected detachment niche. Data showed that the north-facing slopes were scarcely affected (5% of the cataloged processes) by landslides. Conversely, south-facing slopes contained the highest incidence of gravitational processes (48%).
North-facing slopes only accounted for 14% of the total surface in the study area. Instead, southand east-facing slopes were predominant. Considering the relationship between 'landslide' and 'no landslide' pixels, the greatest susceptibility was observed in relation to the west-and south-facing classes ( Figure 11).  Figure 11. Relative Landslide Susceptibility Index (LSIrel,j) as a function of the aspect factor (n.d. is used in the map for the areas where the aspect is not definable, due to the poor level of detail of the available raster)

Distance from Watercourses
The estimated LSI values in relation to the distance from watercourses, as mapped in Figure 12, appeared to be contradictory both with respect to the expected results and the experimental evidence. It was hypothesized that the most prone class could be the one closest to watercourses (i.e., distance shorter than 50 m). The discrepancy between the expected and the real results could depend upon the stream network map scale and the scarce level of detail in the available cartography, where loworder stream segments were omitted. This could explain the singular attribution of the Rovella mountain landslides to a high distance class. In fact, this sub-area, noteworthy because it is located far from the main watercourses, was affected by several low-order incisions. Figure 11. Relative Landslide Susceptibility Index (LSI rel,j ) as a function of the aspect factor (n.d. is used in the map for the areas where the aspect is not definable, due to the poor level of detail of the available raster)

Distance from Watercourses
The estimated LSI values in relation to the distance from watercourses, as mapped in Figure 12, appeared to be contradictory both with respect to the expected results and the experimental evidence. It was hypothesized that the most prone class could be the one closest to watercourses (i.e., distance shorter than 50 m). The discrepancy between the expected and the real results could depend upon the stream network map scale and the scarce level of detail in the available cartography, where low-order stream segments were omitted. This could explain the singular attribution of the Rovella mountain landslides to a high distance class. In fact, this sub-area, noteworthy because it is located far from the main watercourses, was affected by several low-order incisions.

Distance from Watercourses
The estimated LSI values in relation to the distance from watercourses, as mapped in Figure 12, appeared to be contradictory both with respect to the expected results and the experimental evidence. It was hypothesized that the most prone class could be the one closest to watercourses (i.e., distance shorter than 50 m). The discrepancy between the expected and the real results could depend upon the stream network map scale and the scarce level of detail in the available cartography, where loworder stream segments were omitted. This could explain the singular attribution of the Rovella mountain landslides to a high distance class. In fact, this sub-area, noteworthy because it is located far from the main watercourses, was affected by several low-order incisions.

Final Landslide Susceptibility Map
First, the relevance of each chosen control factor for instability was determined, as reported in Figure 13. Lithology appeared to be the most relevant factor in controlling the slope stability (23%). As already shown, shallow landslides only involved the soil layer (0.5 m thick, with a maximum depth of between 2 and 3 m). The mother rock, only emerging in restricted land portions (narrow torrent incisions above 1300 m above sea level), was hardly affected. The relationship with the Relative Landslide Susceptibility Index (LSI rel,j ) as a function of the distance from watercourses.

Final Landslide Susceptibility Map
First, the relevance of each chosen control factor for instability was determined, as reported in Figure 13. Lithology appeared to be the most relevant factor in controlling the slope stability (23%). As already shown, shallow landslides only involved the soil layer (0.5 m thick, with a maximum depth of between 2 and 3 m). The mother rock, only emerging in restricted land portions (narrow torrent incisions above 1300 m above sea level), was hardly affected. The relationship with the bedrock emerges through the eluvium-related soils' granulometry, which is related to different geo-mechanical characteristics. In particular, in the area under study, the gabbro predominates (40%), followed by biotite-bearing leucogranite (31%) and biotitic gneiss (14%). bedrock emerges through the eluvium-related soils' granulometry, which is related to different geomechanical characteristics. In particular, in the area under study, the gabbro predominates (40%), followed by biotite-bearing leucogranite (31%) and biotitic gneiss (14%). Slope inclination is another relevant factor [42,43]. Nearly 50% of the observed processes occurred for dipping angles between 20 and 30°. A lower landslide susceptibility was calculated for the sectors with slope angles between 0 and 20° and those above 35°. On-site surveys confirmed these results, depending on the decline of the eluvial and colluvial cover depth, above a given slope steepness and up to the bare soil outcrop. Finally, a susceptibility map in raster format was produced, representing the landslide susceptibility for the area under study. The data were classified according to different susceptibility classes (Low, Middle, High, and Very High). This classification was obtained through the natural breaks (algorithm of Jenks) classification method [44]. An extract of the susceptibility map is shown in Figure 14. Slope inclination is another relevant factor [42,43]. Nearly 50% of the observed processes occurred for dipping angles between 20 and 30 • . A lower landslide susceptibility was calculated for the sectors with slope angles between 0 and 20 • and those above 35 • . On-site surveys confirmed these results, depending on the decline of the eluvial and colluvial cover depth, above a given slope steepness and up to the bare soil outcrop.
Finally, a susceptibility map in raster format was produced, representing the landslide susceptibility for the area under study. The data were classified according to different susceptibility classes (Low, Middle, High, and Very High). This classification was obtained through the natural breaks (algorithm of Jenks) classification method [44]. An extract of the susceptibility map is shown in Figure 14. Finally, a susceptibility map in raster format was produced, representing the landslide susceptibility for the area under study. The data were classified according to different susceptibility classes (Low, Middle, High, and Very High). This classification was obtained through the natural breaks (algorithm of Jenks) classification method [44]. An extract of the susceptibility map is shown in Figure 14.

Results Validation
In order to test the predictive ability of the susceptibility map, a verification dataset of observed landslides was used as a parallel input. The data, derived from direct field observations, belong to two different events that occurred in the same area in October 2000 and November 1994. The cartographic bases for the identification of landslides (scale 1:5000) were the geo-morphological map and the geo-hydrological instability processes map of the municipality of Valle Mosso, belonging to the same area mapped in the susceptibility map. The landslide positions were digitized as a punctual shapefile called 'Landslide Validation'. Subsequently, in coincidence with each landslide, using the ArcGIS function 'Extract Values to Point', the local susceptibility values were extracted from the raster of the Total Landslide Susceptibility Index (in pixels). Thus, it was possible to find the susceptibility category in relation to each observed instability phenomenon. Figure 15 shows the obtained results of validation process.
In particular, 67% of landslides belonging to the test dataset (used for the validation) fell into the 'Very High' susceptibility class, 24% fell into the 'High' susceptibility class, while only 9% belonged to the 'Middle' or 'Low' classes. the same area mapped in the susceptibility map. The landslide positions were digitized as a punctual shapefile called 'Landslide Validation'. Subsequently, in coincidence with each landslide, using the ArcGIS function 'Extract Values to Point', the local susceptibility values were extracted from the raster of the Total Landslide Susceptibility Index (in pixels). Thus, it was possible to find the susceptibility category in relation to each observed instability phenomenon. Figure 15 shows the obtained results of validation process. In particular, 67% of landslides belonging to the test dataset (used for the validation) fell into the 'Very High' susceptibility class, 24% fell into the 'High' susceptibility class, while only 9% belonged to the 'Middle' or 'Low' classes.

Discussion
The final landslide susceptibility map, based on factors with similar characteristic timescale dynamics, allowed the evaluation of the potential slope instability of the area under study through different sensitivity factors. These analyses were originally used for the management of landslide hazards and risks [45,46]. In particular, qualitative and quantitative approaches were developed to realize inventory maps, relating to landslide predisposing and triggering factors [28]. This was especially possible thanks to the use of GIS platforms [47][48][49][50]. In order to apply a sensitivity assessment, the literature proposed different methods, including probabilistic approaches [51], statistical analyses [52][53][54][55][56][57], physical methods [58,59], or machine learning [60][61][62].
Implementing the original susceptibility calculation method, past extreme geo-hydrological events and susceptibility factors were combined in the vulnerability estimation of the area under study. In particular, through the integration of historical and geo-morphological data, a database with 148 flood/landslide events that affected the area under study was created. Then, a classification of factors was implemented, based on the technical map of the region of interest (scale: 1:10,000). Owing to the potentialities offered by the use of GIS platforms, these maps can be used and shared for land planning and civil protection activities.

Discussion
The final landslide susceptibility map, based on factors with similar characteristic timescale dynamics, allowed the evaluation of the potential slope instability of the area under study through different sensitivity factors. These analyses were originally used for the management of landslide hazards and risks [45,46]. In particular, qualitative and quantitative approaches were developed to realize inventory maps, relating to landslide predisposing and triggering factors [28]. This was especially possible thanks to the use of GIS platforms [47][48][49][50]. In order to apply a sensitivity assessment, the literature proposed different methods, including probabilistic approaches [51], statistical analyses [52][53][54][55][56][57], physical methods [58,59], or machine learning [60][61][62].
Implementing the original susceptibility calculation method, past extreme geo-hydrological events and susceptibility factors were combined in the vulnerability estimation of the area under study. In particular, through the integration of historical and geo-morphological data, a database with 148 flood/landslide events that affected the area under study was created. Then, a classification of factors was implemented, based on the technical map of the region of interest (scale: 1:10,000). Owing to the potentialities offered by the use of GIS platforms, these maps can be used and shared for land planning and civil protection activities.
Applying this approach, it was possible to assess the geographical distribution of the events and the relevance of each triggering factor, also known as a predisposing factor, which refers to the slope's internal conditions [63]. The results confirm the relevance of these factors, as shown by a recent literature review on this topic [64]. Lithology appeared to be the most relevant landslide susceptibility factor. The literature confirms that, as observed in the case study, volcaniclastic rocks, containing highly weatherable minerals, are more prone to the development of shallow instability phenomena than intrusive rocks [65]. Morphology, slope, and aspect were the other relevant factors, in order of magnitude identified by this study. In this respect, considering that the data were extracted from DEM map elaborations, a previous study showed that 10-m spatial resolution DEMs, jointly with the identification of landslide source areas, represent the best possible choice for the preparation of susceptibility maps [66].
Several works included the distance from watercourses as a key parameter in the evaluation of susceptibility [64]. None of them reported the contradictory results that we obtained. This confirms the need to improve the amount and quality of data with respect to this parameter in our case study. With respect to the relevance of watercourses, their artificialization should be mentioned. In fact, as shown in several works, stemming from a historical analysis of relevant flooding and landslide events in different areas, the engineered transformation of river courses enables the significant alteration of the occurrence of instability phenomena [67][68][69][70][71].
In this case study, the poor quality of land use data, which referred to the period prior to the analyzed event, did not allow the inclusion of this parameter in the susceptibility final estimation. Previous studies have shown that land use, slope, and distance-to-river are less significant shallow landslide triggers [72]. The lower relevance of slope as a shallow landslide trigger cannot be confirmed by this study, since it ranked third. Considering the influence of the artificialization of rivers on landslide occurrence, the results with respect to this parameter might be contrasting. Finally, with respect to previous research, the relevance of land use and land use change as a relevant susceptibility factor was proved [73,74].
Following the approach used in this study might be a good option, thus excluding the factors characterized by short timescale dynamics. In fact, these parameters are strongly influenced by anthropogenic alterations of landscape, such as land use and distance-to-river, which might be altered by engineered works and new infrastructures. Moreover, this agrees with the approach that separates predisposing (internal) factors from external factors, such as rainfall [63]. This work excluded a discussion on the climate change impact on weather patterns, such as precipitation intensity, since it cannot represent a factor of intrinsic landscape vulnerability.
The method for landslide susceptibility mapping could be applied as one of the metrics of intrinsic environmental vulnerability of mountain areas in environmental impact assessment. In particular, the development of an integrated mapping tool can complement other methods already available from the literature and be applied to the sustainable development of projects and planning, due to its ability in evidencing the vulnerability of a territory with a sufficiently high spatial resolution and accuracy [75,76]. A further integration with other online tools could also serve as a basis for information sharing along the process of policy implementation by applying a bottom-up approach, as already suggested by the literature [77]. In fact, such representations can condense the meaningful indicators, converting them into augmented information, enabling them to become effective planning instruments [78][79][80].

Conclusions
The applied integrated method allowed the implementation of a zonation technique to assess the intrinsic environmental vulnerability of a test area with respect to landslide susceptibility. Such an approach allowed the determination of areas that were more prone to instability processes. By integrating different databases in a GIS environment and detailing the different factors that trigger the observed instability phenomena, it was possible to obtain an integrated digital representation, with the potential of it being continuously updated. The results, validated through an independent set referring to the same area under study, allowed the identification of different susceptibility classes. A very high percentage of observed slope processes fell under the categories of 'High' (24%) and 'Very High' (70%) susceptibility, showing the efficiency of the method. The validity of prior limiting the analyzed factors to the ones with similar dynamics timescales was also implicitly confirmed.
This GIS-based method can serve as a useful basis for planning and environmental impact assessment, thanks to the possibility of sharing the results among different users and stakeholders. Integrated and homogenized with other intrinsic vulnerability factors, it could also better support the implementation of local and regional informed policies for the sustainable development of vulnerable territories, such as mountains, and for infrastructures and projects to be developed in these areas.