Landslide Susceptibility Assessment of Mauritius Island (Indian Ocean)

: This work is focused on the landslide susceptibility assessment, applied to Mauritius Island. The study area is a volcanic island located in the western part of the Indian Ocean and it is characterized by a plateau-like morphology interrupted by three rugged mountain areas. The island is severely a ﬀ ected by geo-hydrological hazards, generally triggered by tropical storms and cyclones. The landslide susceptibility analysis was performed through an integrated approach based on morphometric analysis and preliminary Geographical Information System (GIS)-based techniques, supported by photogeological analysis and geomorphological ﬁeld mapping. The analysis was completed following a mixed heuristic and statistical approach, integrated using GIS technology. This approach led to the identiﬁcation of eight landslide controlling factors. Hence, each factor was evaluated by assigning appropriate expert-based weights and analyzed for the construction of thematic maps. Finally, all the collected data were mapped through a cartographic overlay process in order to realize a new zonation of landslide susceptibility. The resulting map was grouped into four landslide susceptibility classes: low, medium, high, and very high. This work provides a scientiﬁc basis that could be e ﬀ ectively applied in other tropical areas showing similar climatic and geomorphological features, in order to develop sustainable territorial planning, emergency management, and loss-reduction measures.


Introduction
Landslide susceptibility maps provide valuable information for disaster mitigation works and land planning strategies [1]. The evaluation and the identification of sensitive areas where the probability of landslides is high became a very important phase in the process of territorial planning by enabling a timely start of preventive and remedial actions [2]. Landslide susceptibility is defined as a quantitative or qualitative assessment of the classification, volume (or area), and spatial distribution of landslides that exist or potentially may occur in an area [3]. In more detail, it is the likelihood that a landslide phenomenon happens in a specific area based on local terrain conditions, but without any specification about when it could happen. This estimate of "where" landslides are likely to occur involves a degree of interpretation and is based on the correlation of several factors (i.e., topography, geology, geomorphology, geotechnical properties, climate, vegetation, and anthropogenic factors) with the distribution of past events [4][5][6].
Various methods for landslide susceptibility assessment can be encountered in the scientific literature. In general, they are divided into two groups: qualitative and quantitative methods, with the Situated north of the Tropic of Capricorn, Mauritius has a humid tropical climate characterized by cyclones in the summer period (from November to March) and by moderate to low rainfall during the winter period (from April to October). The mean annual temperature ranges between 23 and 28 °C and the mean annual rainfall is ~2000 mm (ranging from ~600 to ~4000 mm) [32][33][34].
Based on the orography of the landscape, the island can be subdivided into different physiographic domains according to [35]: mountain areas (a), central uplands (b), southern highlands (c), lava plains (d), and coastal areas (e) (Figure 2).
(a) The mountain areas are characterized by sectors with moderate elevation (ranging from ∼600 to ∼800 m a.s.l.), dominated by peaks and ridges, with steep slopes and precipitous wall-like escarpments towards the inland areas and more gentle slopes with outer spurs projecting towards Situated north of the Tropic of Capricorn, Mauritius has a humid tropical climate characterized by cyclones in the summer period (from November to March) and by moderate to low rainfall during the winter period (from April to October). The mean annual temperature ranges between 23 and 28 • C and the mean annual rainfall is~2000 mm (ranging from~600 to~4000 mm) [32][33][34].
Based on the orography of the landscape, the island can be subdivided into different physiographic domains according to [35]: mountain areas (a), central uplands (b), southern highlands (c), lava plains (d), and coastal areas (e) (Figure 2).
(a) The mountain areas are characterized by sectors with moderate elevation (ranging from 600 to~800 m a.s.l.), dominated by peaks and ridges, with steep slopes and precipitous wall-like escarpments towards the inland areas and more gentle slopes with outer spurs projecting towards the sea. These domains are located in the north-western sector showing a NE-SW direction dominated by Pieter Both; in the central-eastern sector with an E-W trending ridge dominated by the relief of Mt. Camizard, and in the south-western sector showing a NNE-SSW direction interrupted by the Black River Gorges.
(b) The central uplands are in the Moka District, showing a morphology from planar to gently radial sloping towards the coast. This domain reaches an altitude of about 400 m a.s.l. and includes two of the main urban areas of the island (Quatre Bornes and Plaines Wilhems). It also includes most of the area within the caldera of the main island volcano.
(c) The southern highlands are in the Savanne and Grand Port districts and comprise the southern side of the main caldera showing a planar morphology mostly above 500 m a.s.l.
(d) The lava plains include the surrounding coastal plains and inland gentle slopes, with a wavy morphology, mostly below 300 m a.s.l. These plains are incised by the main rivers of the island, which locally generate deep gorges and valleys.
(e) The coastal areas show a morphology characterized by flat lowlands with bays and straight coasts. This domain is surrounded by a coral reef and the related lagoon, except for two stretches along the western and southern sectors of the island. showing a morphology from planar to gently radial sloping towards the coast. This domain reaches an altitude of about 400 m a.s.l. and includes two of the main urban areas of the island (Quatre Bornes and Plaines Wilhems). It also includes most of the area within the caldera of the main island volcano.
(c) The southern highlands are in the Savanne and Grand Port districts and comprise the southern side of the main caldera showing a planar morphology mostly above 500 m a.s.l.
(d) The lava plains include the surrounding coastal plains and inland gentle slopes, with a wavy morphology, mostly below 300 m a.s.l. These plains are incised by the main rivers of the island, which locally generate deep gorges and valleys.
(e) The coastal areas show a morphology characterized by flat lowlands with bays and straight coasts. This domain is surrounded by a coral reef and the related lagoon, except for two stretches along the western and southern sectors of the island.  The geological history of Mauritius Island and its volcanism are historically well documented [36][37][38][39][40]. The island was built up during three distinct episodes of volcanic activity (from 10 to 0.02 My), which can be described, according to a geological chronology [35], as follows: Ancient Lava Series  including the breccia series (10-7.8 My), which caused the emergence of the island, and the Old Lava Series (7.6-5 My), which consisted of ancient basalts and phonolite building the primitive island shield volcano, whose central part collapsed to form the main Mauritius Caldera; the Intermediate Lava Series (or Early Volcanic Series) (3.5-1.7 My) consisting mostly of pyroclasts and basaltic flows; and the Recent Lava Series (also named Late Lavas, 0.7-0.02 My), the most important and extensive volcanic event in terms of aerial coverage (about 70% of the island), comprising basalts, scoria, tuff, and pyroclasts ( Figure 3). The geological history of Mauritius Island and its volcanism are historically well documented [36][37][38][39][40]. The island was built up during three distinct episodes of volcanic activity (from 10 to 0.02 My), which can be described, according to a geological chronology [35], as follows: Ancient Lava Series  including the breccia series (10-7.8 My), which caused the emergence of the island, and the Old Lava Series (7.6-5 My), which consisted of ancient basalts and phonolite building the primitive island shield volcano, whose central part collapsed to form the main Mauritius Caldera; the Intermediate Lava Series (or Early Volcanic Series) (3.5-1.7 My) consisting mostly of pyroclasts and basaltic flows; and the Recent Lava Series (also named Late Lavas, 0.7-0.02 My), the most important and extensive volcanic event in terms of aerial coverage (about 70% of the island), comprising basalts, scoria, tuff, and pyroclasts ( Figure 3).  and landslide deposits. The alluvial, alluvial fan, and colluvial deposits are present in the main fluvial plains, and debris, landslide, and slope deposits are developed in the mountain environments. The coastal domain is characterized by the presence of coraline sandy beaches and sand dunes bordering approximately 20% of the coastline.

Methods
The landslide susceptibility assessment was achieved through an integrated approach based on the combination of morphometric analysis and preliminary GIS-based techniques, supported by photogeological analysis and geomorphological field mapping.
Morphometric and slope analysis was carried out with the GIS software (ArcMap ® 10.6.1, ESRI, Redlands, CA, USA). It was performed using topographic maps (1:25,000-1:10,000 scale) and supported by the use of a 40 m Digital Terrain Model (DTM) as a base map, and by the creation of a Digital Elevation Model (10 m DEM) derived from 1:10,000 scale technical maps. This analysis was based on the definition of the orographic features and the drainage network of the island. By means of the SHALSTAB model [41,42], a preliminary map was produced in order to outline the spatial distribution of the landslide susceptibility. This model is able to capture the physics involved in landslides triggered by rainfall in tropical and mountainous terrains and could be implemented successfully to identify slopes prone to failure with only a high-resolution Digital Elevation Model.
The geomorphological analysis was based on stereoscopic air-photo interpretation and field mapping. Aerial photo interpretation and photogeological analysis of high resolution (15 cm) orthophoto images allowed us to map homogeneous areas prone to landslides and slope instability. Field mapping was carried out at an appropriate scale (1:5,000-1:10,000), according to international guidelines [43], Italian geomorphological guidelines [44][45][46], and the thematic literature concerning geomorphological mapping and analysis in different geological and climatic contexts [47][48][49][50][51]. We focused on the definition of lithological features, superficial deposit cover, and the type and distribution of geomorphological landforms with reference to the main landslides affecting the island. Field survey and schematic geological-geomorphological field sketches allowed to analyze landslide geometry and their state of activity, as well as the anthropogenic influence. The analysis of the major lithological features was performed by integrating field data with the geological map of Mauritius Island at 1:50,000 scale [40,52]; while, the land-use analysis was based on a 1:100,000 scale map [53] and information about the soil types have been derived from a soil map at a 1:50,000 scale [54]. Special attention has been devoted to mapping superficial deposits (i.e., colluvial, alluvial, landslide, and slope deposits), which had been poorly mapped in previous geological maps [40,52].
The landslide susceptibility assessment was based on a mixed heuristic and statistical analysis, according to the most relevant literature [55][56][57][58][59][60]. This approach led to the identification and combination of relevant parameters for the mechanism of landslide occurrence. Regarding the study area, eight parameters (slope, aspect, topographic curvature, drainage pattern, vegetation cover, lithology, soil cover, and rainfall) were considered as landslide controlling factors. All the factors were portrayed in thematic maps and divided into different classes. Each class of every single factor was rated following a weighting system based on comparative judgments and synthesis of priorities [61][62][63][64]. This iterative procedure allowed us to rate individual classes in order to provide an immediate measure of the role played by each factor and its related classes on the landslide susceptibility. In detail, this stepwise approach led to compare the factor dataset with the preliminary GIS-based landslide susceptibility and with the distribution of the different types of landslide (i.e., rockfalls, landslides, and rapid earth flows). The analysis was performed through the calculation of the "Landslide index" (Li), expressed by the equation [65,66]: where Al is the surface area of every class of the factors, and At is the whole surface area of the island. Moreover, a parameter reclassification was performed in order to better assign the appropriate weight to each class, using the following equation [65,66]: where Li 0 is the landslide index for a specific class of the considered factor, Li min is the minimum landslide index value of all classes for a given factor, and Li max is the maximum landslide index value of all classes for a given factor. Then, a categorization of the factors according to their role in the weighting system was realized. Each class was normalized to 100 percent so that calibration would have the same scale in all factors. The convenience of such a normalization consists in the possibility to weigh the contribution among classes belonging to factors with different ranges of Li-values. All the weighted factors were summed altogether, using the ArcGIS Raster Calculator tool, and reclassified to obtain the landslide susceptibility for each mass movement type, as follows: where Wf i is the weight of each factor and crf i is the contribution ratio, evaluated through an expert-based approach. This summation resulted in the final susceptibility mapping, where the higher the S value, the higher the susceptibility to different types of mass movements (rockfalls, landslides, or rapid earth flows). Finally, all the collected data, portrayed in three thematic maps, were integrated into the GIS software through a cartographic overlay process [67,68] in order to portray the spatial distribution of the landslide susceptibility.

Landslide Controlling Factors
Landslide susceptibility analysis involves data collection and construction of a spatial database from which relevant factors are extracted. Selecting those independent variables with a major role is, however, a difficult task. There are neither universal criteria nor guidelines. Hence, the selection of factors needs to take into account the nature of the study area and data availability [69][70][71].
The eight factors considered for the susceptibility analysis were: slope, aspect, topographic curvature, drainage pattern, vegetation cover, lithology, soil cover, and rainfall. Each factor, described in the following paragraphs, was selected and subdivided into different classes in order to best commensurate the diversity of the data source and difference in the scales and to clearly delineate its role in the mechanism of landslide occurrence in the wider geological and geomorphological context of the island.

Slope
The slope angle is commonly used in landslide susceptibility studies since landsliding is directly related to this factor [58,65,72,73]. In this study, the slope has been divided into ten classes ( Figure 4a). The predominant class is 0-5 • and it is homogeneously present in correspondence of flat areas along coastal and lava plains. Some wide sectors of the lowlands and uplands show slope values ranging from 5-15 • to 25-35 • , with steep slopes affected by landslides and earth flows. The higher slope values are related to peaks and ridges of the mountain areas which present sub-vertical slopes (over 55 • ), affected by rockfalls.

Aspect
Aspect is considered a less important factor in landslide susceptibility studies [55,72]. Nevertheless, aspect-associated parameters, such as exposure to sunlight, drying winds, and rainfall may influence the occurrence of landslides [73][74][75]. The aspect map has been classified into ten classes (Figure 4b).
The overall shape of the island outlines a specific aspect distribution, in which slopes facing north to northwest and south to southwest are slightly predominant over east to southeast facing slopes.

Topographic Curvature
The term curvature is theoretically defined as the rate of change of slope gradient (profile curvature) and/or aspect (planform curvature), usually in a specific direction [76,77]. Positive values of profile curvatures define convexity; negative values of profile curvatures characterize slope concavity [73]. Positive values of topographic curvature characterize ridges, while negative ones define valleys. Values around zero indicate flat surfaces, whatever the slope is. The topographic curvature map (Figure 4c) combines both the profile and planform curvatures. The highest values characterize the three mountain areas with scarps, ridges, deep valleys, and gorges where rockfalls and landslides may occur. Moderate values locally affect the southern highlands and the lava plains along the main fluvial incisions.

Drainage Pattern
The drainage network, strictly influenced by the underlying lithology, can be used to extract information on the general direction of surface water flows towards individual basin outlets, the angle of intersection between the tributaries and the main channels, and the overall confluences distribution, which can control the drainage discharge and the related instability, particularly in critical hydrological conditions.
The drainage network was divided into nine homogeneous drainage patterns ( Figure 4d). The predominant pattern is sub-dendritic, although in some sectors it is absent due to the high permeability of lavas and pyroclasts. Parallel patterns and parallel with meandering channels characterize the central and southern highlands especially in the southern sector of the island. Radial, centripetal, and rectangular patterns are locally present. The drainage network is totally absent in the northern sector of the island.

Vegetation Cover
The vegetation cover plays an important role in slope stability. In general, sparsely, or poorly vegetated areas are affected by faster soil erosion and greater instability than forested ones [26,58,72]. In the study area, a vegetation map (Figure 4e) was derived from the photogeological analysis and the land-use map [53]. Vegetation cover was classified into eight classes, in which areas with absent vegetation correspond to urban areas. The predominant vegetation type is shrub crops, including sugar canes, which cover most of the lava plains, and the central and southern highlands, whereas forests (heavy tree canopy and sparse trees) cover large parts of the mountain areas.

Lithology and Soil Cover Deposits
Lithology and soil cover are important factors in landslides susceptibility analysis, since different lithological units may be affected by different landslide types with variable susceptibility degrees. Moreover, soil cover deposits, mostly exposed to weathering, may influence land permeability, geotechnical parameters and, therefore, the landslide type, as known from thematic literature [26,73,78,79].
The predominant lithological units are: fresh basaltic flows (Recent Lava Series), mostly outcropping in the lava plains and the central and southern highlands and poorly affected by mass movements; pyroclasts and weathered basaltic flows (Intermediate Lava Series) outcropping in the central and southern highlands, and locally affected by landslides and earth flows; ancient basalts and phonolites (Ancient Lava Series), outcropping in the mountain areas and largely affected by rockfalls; colluvial deposits, widespread at the base of the mountainous slopes, largely affected by earth flows and translational-rotational landslides (Figure 3).
For the study area, the soil types were grouped into eight units depending on their typology and the related thickness (Figure 4f). The most widespread type is the brown to red-brown soil with rocky fragments and basalt outcrops, which usually increases in thickness (from few to >100 cm) in the middle and lower part of the mountain areas, and variably rocky soil of moderate thickness.

Rainfall
Rainfall and heavy rainfall events play a key role in the mechanism of landslide triggering. This mechanism depends on rainfall distribution and it is controlled by the interactions of different factors, such as orography and hydrography, lithology, land use and vegetation, etc. [80][81][82][83]. The analysis of the rainfall and temperature data, evaluated from a 29-year time record , proved certain features of the climate on the island ( Figure 5). The island is characterized by a humid tropical maritime climate and by permanent SE trade winds. The mean annual temperature ranges between 23 and 28 • C. The mean summer (November-March) rainfall is 1300 mm; the mean winter (April-October) rainfall is 660 mm. The mean annual rainfall is greater in the central uplands (4000 mm) and windward slopes (~1400 mm on the eastern coast) and lower in the western slopes (600 mm on the coast). Occasionally, tropical cyclones and storms affect the island inducing geomorphological effects, such as the events occurred in March-April 2013 and in March 2015 [25,[32][33][34]. The average summer rainfall was used in the weighting procedure to best represent the link between rainfall data and landslide susceptibility in this dynamic climatic context. Rainfall and heavy rainfall events play a key role in the mechanism of landslide triggering. This mechanism depends on rainfall distribution and it is controlled by the interactions of different factors, such as orography and hydrography, lithology, land use and vegetation, etc. [80][81][82][83]. The analysis of the rainfall and temperature data, evaluated from a 29-year time record , proved certain features of the climate on the island ( Figure 5). The island is characterized by a humid tropical maritime climate and by permanent SE trade winds. The mean annual temperature ranges between 23 and 28 °C. The mean summer (November-March) rainfall is 1300 mm; the mean winter (April-October) rainfall is 660 mm. The mean annual rainfall is greater in the central uplands (4000 mm) and windward slopes (~1400 mm on the eastern coast) and lower in the western slopes (600 mm on the coast). Occasionally, tropical cyclones and storms affect the island inducing geomorphological effects, such as the events occurred in March-April 2013 and in March 2015 [25,[32][33][34]. The average summer rainfall was used in the weighting procedure to best represent the link between rainfall data and landslide susceptibility in this dynamic climatic context.

Preliminary GIS-Based Landslide Susceptibility Map
A preliminary susceptibility map ( Figure 6) was produced using GIS technology. This map allowed us to outline the critical areas in terms of potential spatial distribution of landslide susceptibility. For this kind of analysis, SHALSTAB model provides a DEM modeling calculation and applies a topographic index suggesting that surface topography is a primary indicator of where landslides are most likely to occur. In the current study, different areas marking low, medium, high, and very-high landslide susceptibility were identified. The analysis allowed us to identify areas with high and very high landslide susceptibility in correspondence to the three mountain areas, to some sectors of the southern highlands, and to the incision of the main rivers. Central uplands, southern highlands, and lava plains show the lowest values in accordance with the morphology of the island. A region that has no class values presents no landslide susceptibility and mostly corresponds to coastal areas.

Photogeological Analysis
The photogeological analysis was carried out to support and verify the preliminary GIS mapping. The analysis allowed for outlining the main landslide-prone areas, considering previous studies and technical reports provided by the Mauritius Ministry of Environment and Sustainable Development. According to the orographic (elevation and slope), hydrographic (drainage pattern), vegetational, geological (lithology and soil cover) and geomorphological (slope gravity landforms and processes) features, a detailed analysis was performed in the three mountain areas of the island

Photogeological Analysis
The photogeological analysis was carried out to support and verify the preliminary GIS mapping. The analysis allowed for outlining the main landslide-prone areas, considering previous studies and technical reports provided by the Mauritius Ministry of Environment and Sustainable Development. According to the orographic (elevation and slope), hydrographic (drainage pattern), vegetational, geological (lithology and soil cover) and geomorphological (slope gravity landforms and processes) features, a detailed analysis was performed in the three mountain areas of the island with reference to some specific and mostly representative sites (Chitrakoot, Quatre Soeurs, Chamarel, and Corps de Garde) (Figure 7). Each of the mapped sites is located in a different sector of the island and it is characterized by a specific geomorphological setting highlighting the development of different types of mass movements, such as rockfalls, landslides, and rapid earth flows. Chitrakoot area (Figure 7a) is in the Port Louis District, located at moderate elevation (ranging from 100 to 250 m a.s.l.) with low to moderate slope (25-35°) in the lower part, passing to more gentle slopes in the upper part. The drainage pattern is subdendritic and this site shows sparse tree vegetation, with some patches of heavy three canopy. The bedrock lithology is ancient basalt, generally covered by landslide deposits. Slope gravity processes are referable to large landslides affecting the middle and lower sectors of the area.
Quatre Soeurs area (Figure 7b) is in the Flacq District, located at low elevation (ranging from 0 to 50 m a.s.l.) with low slope values (5-15°). The drainage pattern is almost subdendritic with small Chitrakoot area (Figure 7a) is in the Port Louis District, located at moderate elevation (ranging from 100 to 250 m a.s.l.) with low to moderate slope (25-35 • ) in the lower part, passing to more gentle slopes in the upper part. The drainage pattern is subdendritic and this site shows sparse tree vegetation, with some patches of heavy three canopy. The bedrock lithology is ancient basalt, generally covered by landslide deposits. Slope gravity processes are referable to large landslides affecting the middle and lower sectors of the area.
Quatre Soeurs area (Figure 7b) is in the Flacq District, located at low elevation (ranging from 0 to 50 m a.s.l.) with low slope values (5-15 • ). The drainage pattern is almost subdendritic with small drainage lines perpendicular to the slope. This site shows heavy tree canopy in the very upper part, with dense shrub crops in most of the urban areas along the coast. The bedrock lithology is ancient basalt, covered by colluvial deposits. The area is affected by slope gravity processes referable to a complex landslide located in the lower part of the slope near the coastal area.
Chamarel area (Figure 7c) is in the Black River District, located at intermediate elevation (ranging from 250 to 400 m a.s.l.); slope values are around 45-50 • with some vertical scarps in the lower sector. The drainage pattern is subdendritic and the vegetation is characterized by heavy tree canopy. The bedrock lithologies are pyroclastic rocks and interbedded weathered lavas, covered by thick to moderately thick brown soil. The area shows evidence of rapid earth flows along the main drainage lines.
Corps de Garde area (Figure 7d) is in the Plaines Wilhems District and it is represented by a vertical scarp at an altitude >500 m a.s.l. with high slope values (>73 • ). The hydrography is absent, and the vegetation is absent as well. The main geomorphological feature is represented by a wide scarp representing the detachment area of rockfalls.

Geomorphological Field Survey
The field survey was performed to verify the presence, geometry, typology, and state of activity of mass movements in specific and most representative sites (Chitrakoot, Quatre Soeurs, Chamarel, and Corps de Garde). The main landslides-prone areas were mapped in the field by collecting data on specific field sheets, including geological-geomorphological field sketches ( Figure 8) and photo documentation (Figure 9), showing the relationship between landslides, bedrock lithology, and superficial deposits.
Chitrakoot area is characterized by the presence of the main landslide scarp on the ancient basalts lithologies. In the middle part of the slope, moving towards the urban area, the wide gently-undulating sector is referable to a landslide terrace, with small counterslopes, that could be connected to a large landslide involving the whole slope, with a possible deep slip surface (>20-30 m). This area is also affected by shallow to moderately-slow landslides involving the colluvial deposits and inducing secondary landslide scarps, tilting, severe cracks, and damages to residential houses. This geomorphological setting is supposed to be linked to the superimposition of small slides on a large complex landslide, probably triggered by heavy rainfall events (Figure 8a).
Quatre Soeurs area is also dominated by the main landslide scarp on the ancient basalts lithologies, partly covered by slope deposits. In the middle part of the slope a wide counterslope, with small swamps and small scarps, is present, affecting colluvial and slope deposits. In the lower part of the slope, the houses are affected by wide cracks and tilting both down-slope and counterslope. At the base of the slope, finally, small springs are present and strong damages affect the main road. This geomorphological setting is supposed to be linked to the superimposition of shallow small rotational or complex landslides on a wide deep translational-rotational landslide, sliding on the main slip surface (Figure 8b).
Chamarel area presents a geomorphological setting variable from the upper to the lower sector of the slope, according to the complex morphological setting. The upper sector of the slope, developed on pyroclast rocks and weathered basaltic flows, is characterized by small valleys with V-and U-shape, arranged in subdendritic drainage pattern. These valleys are affected by gully erosion and earth flows during heavy rainfall events. In the middle and lower part of the slope, the alternation of thick layers of basalts and pyroclastic rocks induced a step-like morphology affected by small landslides and rockfalls; the latter affects mainly the lower layer of ancient fractured basalts, with the formation of rock blocks up to >10 m in size (Figure 8c).

Landslide Distribution
The spatial distribution of landslide susceptibility resulted from the combination of photogeological analysis and field survey. In detail, it was produced by merging data from aerial photo interpretation of high-resolution orthophoto images in the three mountain areas of the island and data from geomorphological field activity in specific-site investigations, defined considering technical reports provided by the Mauritius Ministry of Environment and Sustainable Development. This analysis allowed us to characterize different geomorphological homogeneous areas showing evidence of slope gravity processes such as rockfalls, landslides (complex landslides, translational

Landslide Distribution
The spatial distribution of landslide susceptibility resulted from the combination of photogeological analysis and field survey. In detail, it was produced by merging data from aerial photo interpretation of high-resolution orthophoto images in the three mountain areas of the island and data from geomorphological field activity in specific-site investigations, defined considering technical reports provided by the Mauritius Ministry of Environment and Sustainable Development. This analysis allowed us to characterize different geomorphological homogeneous areas showing evidence of slope gravity processes such as rockfalls, landslides (complex landslides, translational and rotational slides), and rapid earth flows. This arrangement is summarized and graphically shown in Figure 10. and rotational slides), and rapid earth flows. This arrangement is summarized and graphically shown in Figure 10. Rockfalls occur on steep or vertical wall-like escarpments in the mountain areas, at elevations ranging from about 500 to 800 m a.s.l., showing the highest slopes values (from 50 to >73°). The escarpments are on the bare rock (mostly ancient basalts lithologies with no vegetation and no soil cover). Surface hydrography is generally absent, only related to surface water runoff and infiltration due to rock fracturation. The main landslide scarps represent the detachment areas of rockfalls and at the base large talus slopes deposits are present, made up of large blocks embedded in coarse debris resulting from repeated rock falls, and mostly covered by thick shrub and tree vegetation. Large complex landslides affect the middle and lower parts of the slopes and are located at a The escarpments are on the bare rock (mostly ancient basalts lithologies with no vegetation and no soil cover). Surface hydrography is generally absent, only related to surface water runoff and infiltration due to rock fracturation. The main landslide scarps represent the detachment areas of rockfalls and at the base large talus slopes deposits are present, made up of large blocks embedded in coarse debris resulting from repeated rock falls, and mostly covered by thick shrub and tree vegetation. Large complex landslides affect the middle and lower parts of the slopes and are located at a moderate elevation ranging from 150 to 500 m a.s.l. on concave or convex-concave slopes surrounding the mountain areas, with moderate slope values (25-35 • ); locally gentle landslides terraces occur. The bedrock lithology is represented by ancient basalts or pyroclasts, generally covered by superficial deposits (colluvial and slope deposits). The drainage pattern is mostly subdendritic or parallel (locally radial). The main sites show sparse tree vegetation, with some patches of heavy tree canopy in the middle and upper part.
Translational and rotational landslides mostly occur at low elevations ranging from 0 to 150 m a.s.l. showing low slopes values (5-15 • ), changing to steeper slopes in the upper sectors (15-25 • ); they affect the basal slopes surrounding the mountain areas and in some cases the slopes of the central and southern highlands. The bedrock lithology is represented by intermediate basalts or pyroclasts, covered by colluvial deposits. The drainage pattern is almost parallel with small drainage lines perpendicular to the slope. Most of the sites show a heavy tree canopy in the very upper part, and dense shrub crops and shrubs in most of the area. Locally, this type of slide occurs also along the fluvial scarps of the main rivers.
Rapid earth flows are located at intermediate elevations ranging from 150 to 500 m a.s.l. in the southern highlands and locally around the mountain areas. The slope values are usually around 15-25 • , locally increasing in the correspondence of vertical scarps. The bedrock lithologies are mostly pyroclastic rocks and interbedded weathered lavas, covered by colluvial deposits. The drainage pattern is mostly subdendritic; the vegetation is characterized by sparse trees and in many cases by shrub crops (sugar cane).

Discussion
Landslide susceptibility assessment was performed through the analysis and the cartographic overlay of susceptibility to different types of mass movements: rockfalls, landslides (including rotational and translational slides, complex landslides, general slow landslides) and rapid earth flow. The analysis was based on the interaction between controlling factors, which provides certain clues for landslide susceptibility, such as orographic features (slope, aspect, profile, and planar curvature), hydrographic features (drainage pattern), geological features (lithology and soil cover), vegetation features, and rainfall data. Original data were converted to digital form and individual factors were processed through digitization in a GIS environment. All the factors were portrayed in thematic maps and divided into different classes. Each class of every single factor was analyzed through a weighting procedure to provide an immediate measure of the role played on the landslide susceptibility. In detail, this stepwise approach allowed us to compare the factor dataset with the preliminary GIS-based landslide susceptibility and with the distribution of the different types of landslide (i.e., rockfalls, landslides, and rapid earth flows). The correlation between controlling factors and landslides has been computed by evaluating the "Landslide index" (Li) for the individual classes into which each factor has been divided. Then, a categorization of the factors according to their role in the weighting system was realized. Each class was normalized to 100 percent to best weigh the contribution among classes belonging to factors with different ranges of Li-values and work with the same scale in all factors. This weighting procedure is summarized in Table 1. Moreover, with reference to the physiographic context of the island, the slope factor was considered as a primary indicator of slope instability and it was processed to define its significant role in the occurrence of different types of landslides ( Table 2). The slope was classified into ten classes from 0 to >73 • , and each class was weighted defining Li and W [65,66], separately for different landslide types. Class 1 of the slope layer (slope ranging 0-5 • ) has almost no observed rockfalls area, resulting in zero weight; while, for example, class 9 (slope ranging 65-73 • ) has the maximum Li, resulting in the maximum weight (defined as 100), since sub-vertical slopes are largely affected by rockfalls. As for landslide distribution, the highest Li values are for class 2 (slope ranging 5-15 • ) and class 3 (slope ranging 15-25 • ), resulting in a 100 and 23 weight, respectively. As for rapid earth flow distribution, the highest Li values are again for class 2 (5-15 • ), resulting in a 100 weight, and for classes 3 to 7 (15-55 • ). To better calibrate and testify the role played by landslide controlling factors, an overall contribution ratio (crfi) was assigned to each factor (Table 3). This process of calibration was performed, through an expert-based approach, considering the influence of factors and related classes in the mechanism of occurrence of different types of mass movements. In detail, the contribution ratio was defined by combining data obtained from photogeological analysis and geomorphological field surveys in specific-site investigations. The analysis revealed that the contribution of each influencing factor varies across the island. The spatial variation showed that morphological factors are important driving parameters for slope failures and have multiple influences on landslide susceptibility. The slope factor directly affects shear stress and, generally, gentle slopes are expected to have lower susceptibility to landsliding than steep ones. Profile and planar curvature express the shape of the slope and were rated in order to reflect the topography of the island. Lithological features also influence the landslide susceptibility with strong effects on hydrological and mechanical characteristics of rock mass. Hence, lithology was considered as to the different role played in the mechanism of occurrence of different types of mass movements. Even though Ancient Lava Series are considered as hard and stiff rocks, due to the high fragmentation they are favorable to rock falls. Loose and high erodible formations, like superficial deposits, tuff, pyroclasts, and weathered basaltic flows, are prone to generate rotational and translational slides, and general slow landslides, since they are unconsolidated formations. Finally, all the weighted factors were summed altogether, using the ArcGIS Raster Calculator tool, and reclassified to portray the spatial distribution of landslide susceptibility in thematic maps ( Figure 11). This analysis allowed us to produce three different susceptibility maps for rockfalls, landslides (including rotational and translational slides, complex landslides, general slow landslides), and rapid earth flows.  Landslide susceptibility map (Figure 11b) shows very-high levels in the lower part of the main slopes of the mountainous-hilly areas of the northern, eastern, and south-western parts of the island. These areas are mostly on ancient basalt rocks in the upper part and on basalt or pyroclastic rock in the lower part, covered by superficial deposits. They are characterized by sparse trees and shrubs vegetation cover and show the presence of several scarps and landslides terraces outlining an irregular convex-concave morphology.
Rapid earth flow susceptibility map (Figure 11c) shows very high values from the upper to the lower part of the main slopes of the mountainous and hilly areas of the northern, eastern, and south-western sides of the island, even though these are covered by sparse trees and tree canopy. High values are also present in some hilly or gently undulated landscapes that developed on pyroclastic rocks and recent lavas, covered by sugar cane crops, on the southern side of the island (Savanne District; i.e., Chamarel area).
These maps were summed altogether and reclassified to define, through a cartographic overlay approach into the GIS software [64,67,68], the spatial distribution of landslide susceptibility (Figure 11d). This analysis depicts four main classes (low, medium, high, and very high); thus, the maximum values, resulting from the combination of the three mass movements susceptibility maps, are classified as the maximum susceptibility value.
The spatial distribution of susceptibility related to the different types of mass movement outlines that high/very high landslide susceptibility is strictly consistent with the distribution of the overall "Landslide index" (Li) derived from the statistical approach and confirmed by photogeological analysis and field survey. This is mostly related to the steep slopes on basalt rocks (mostly affected by rockfalls) and to the steep to moderately steep slopes at the base of the basalt rock slopes (potentially affected by landslides and rapid earth flows). Concerning the medium class, the large areas are particularly significant, since they are mostly related to moderate to gentle slopes characterized by superficial deposits (i.e., colluvial and slope deposits), outlined by the photogeological and field survey, potentially affected by large complex landslides.

Conclusions
This paper presents a landslide susceptibility assessment realized following a multidisciplinary approach, which involves morphometric analysis, GIS-based analysis, photogeological analysis, and geomorphological field mapping. This detailed investigation shows clearly that the geomorphological response of the island is highly variable and dependent on a range of eight parameters including slope, aspect, topographic curvature, drainage pattern, vegetation cover, lithological features, soil cover, and rainfall, which can be considered as landslide controlling factors. The landslide susceptibility maps created are intended for use as a general guide to depicting areas of relative susceptibility to slope failure and as a predictor of landslide hazards at specific sites. Areas of high and very high landslide susceptibility depict the potential for slope failure to occur but do not depict the time frame of the failure, the type of failure nor the volume of the mass involved. The study was performed using input data from different scale (ranging from 1:5000-1:100,000) in order to best summarize all the bibliographic information and to describe landslides in the wider context of their geomorphological situation with their role in landscape development assessed so as to understand the likelihood and potential magnitude of any hazards [84]. In order to avoid inaccuracy in the results, it was verified by detailed geological-geomorphological field mapping and photogeological analysis in specific and most representative sites. The type and degree of landslide susceptibility were derived from mixed heuristic and statistical analysis, based on either direct (photogeological analysis and field mapping) or indirect (GIS data processing) analysis. The overall results of this work allowed us to realize a new zonation of landslide susceptibility for Mauritius Island, identifying four landslide susceptibility classes (low, medium, high, and very high). This work gave a better comprehension of the relations between morphological, lithological, and geomorphological conditions, and the rainfall events triggering mass movements, such as rockfalls, landslides, and rapid earth flows. Therefore, this methodological approach was useful to identify areas that are susceptible to mass movements. Moreover, the comparison between preliminary GIS-based susceptibility map ( Figure 6) and final susceptibility maps (Figure 11d) clearly outlines the significant contribution of photogeology analysis and geomorphological field survey, together with mixed heuristic-statistical procedure, to develop a technique for landslide susceptibility assessment. This approach allowed us to reduce and refine the spatial distribution of high and very-high landslide susceptibility and made it possible to map a large distribution of medium susceptibility that mostly pertains to the occurrence of landslides on moderate to gentle slopes not outlined by the preliminary maps and poorly known in the previous thematic studies. In conclusion, this approach allowed for a better calibration of landslide susceptibility analysis, for the improvement of results in terms of consistency with mass movements distribution, and for the definition of the main critical areas all over the Mauritius Island. Hence, landslide susceptibility assessment provides a scientific basis for the implementation of land use, emergency management, climate change resilience actions, and loss-reduction measures. This approach could be effectively applied in other tropical areas showing similar climatic and geomorphological features in order to identify landslide-prone areas and give information for disaster mitigation works and land planning strategies.