Introduction

Landslides are among the most common geohazard in many parts of the world (Haque et al. 2019; Saadatkhah et al. 2014). They claims and damages human life and infrastructures in many mountainous regions (Corominas et al. 2014; Kanungo et al. 2009; Mohan et al. 2011; Yalcin et al. 2011; Wyllie 2015). The impacts of these events on the local and global economy are huge (Yalcin et al. 2011; AGU 2017), and mountains remain to be the top agenda in terms of geohazard (Debarbieux and Rudaz 2015; United Nations 2016; Price et al. 2019). Many authors and researchers defined landslide as any down slope movement of earth materials due to gravity (Cruden 1991; Highland and Bobrowsky 2008; Hungr et al. 2013; Varnes 1984). They are responsible for the damages of hundreds of billions of properties, transportation networks, buildings and other structures, deaths, and injuries every year. Therefore, identification and mapping of landslide-prone mountain areas is considered important to guarantee the safety of human life and property and to support land planners and decision makers (Zhou et al. 2018).

In Ethiopia, landslide-generated damages are serious problems to the planners and decision makers at various levels of the government (Woldearegay 2013) that cause many social, economical, and geomorphological impacts (Broothaerts et al. 2012). According to Abebe et al. (2010), the most common triggering factor for landslide incidence in Ethiopia is heavy rainfall. Ayenew and Barbieri (2005) also pointed out that gully erosion is one of the triggering factors responsible for generating landslides due to erodible materials and heavy floods. In most rock masses, the large and closely spaced joints are the main causes for rock falls and toppling. In many mountainous areas, rock discontinuities are one of the causative factors for various types of landslides and mass movements (Ayenew and Barbieri 2005; Sun et al. 2018). Most regions of the northern Ethiopia are rugged with steep hillside slopes, steep river banks, and steep road cut slopes that have long histories of landslide activity (Ayenew and Barbieri 2005).

Landslide susceptibility analyses are the initial step in understanding and minimizing landslide damages. Such analysis can be implemented using various methods such as statistical (e.g., frequency ratio (FR)), heuristic (e.g., analytical hierarchy process (AHP)) methods (Corominas et al. 2014; Hong et al. 2016; Nicu 2018; Sun et al. 2018; Pourghasemi et al. 2018; Yalcin et al. 2011), and kinematic analysis (KA). As explained by Broothaerts et al. (2012), landslides are the result of a combination of both intrinsic and triggering factors. A landslide inventory map shows the areal distribution of current and past landslide areas and their characteristics. Analysis and summation of weighted values of landslide conditioning factors such as aspect, elevation, distance to fault, distance to stream, etc. are the basic factors for landslide susceptibility map (LSM) preparation (Demir et al. 2013; Hong et al. 2016; Sun et al. 2018). Analysis of rock slope stability using KA indicates the role of discontinuities for development of landslides (Baharuddin et al. 2016).

To mitigate landslide damage, it is necessary to evaluate and analyze the causative factors, hazard, vulnerability, and risk along with identification and delineation of landslide-susceptible areas (Hong et al. 2016; Pourghasemi and Rahmatib 2018; Saadatkhah et al. 2014). This LSM was carried out once all field evidence-related data and laboratory test results are collected and analyzed. Basically, a LSM subdivides or classifies a given terrain or area into zones with differing likelihoods that landslides or geomorphological processes of different types may occur (Corominas et al. 2014). It predicts “where” landslides are likely to occur (Guzzetti et al. 2005; Reichenbach et al. 2018). Landslide susceptibility assessment is considered an initial step toward a landslide hazard and risk assessment, but according to Corominas et al. (2014), it can also be an end product in itself that can be used in land-use planning and environmental impact assessment. This is especially the case in situations where insufficient information is available on past landslide occurrence to allow the spatial and temporal probabilities of events to be assessed. As landslide susceptibility maps primarily provide a proposed ranking of terrain units in terms of spatial probability of occurrence, they do not explicitly convey information on landslide return periods.

The Mgulat Mountain area is a typical example for its data scarcity. LSMs contain information on the type of landslides that might occur and on their spatial likelihood of occurrence in terms of identifying the most probable initiation areas (based on a combination of geological, geomorphological, and LULC conditions).

The main objective of this study was to describe the landslide causative factors and produce LSM using FR and to assess stability of selected rock slopes using KA method. Landslide susceptibility map is the division of land surface into well-defined zones and ranking the zones based on the degree of potential landslide susceptibility. It is used to identify places of landslide occurrences on the basis of a set of causative factors (Hong et al. 2016; Kanungo et al. 2009) in which the value represents the relative susceptibility to future landslide occurrence, as higher values are associated with high landslide susceptibility (Avinash and Ashamanjari 2010).

Description of the study area

Mgulat area is located in the northern part of Ethiopia, Tigray Regional State, Eastern Zone within Ganta-Afeshum Woreda (district). Geographically, it is bounded between 1,570,000 to 1,582,000 mN latitude and 530,000 to 546,000 mE longitude (UTM Zone 37N). The study area covers 88.6 km2, and it is characterized by a dendritic drainage pattern (Fig. 1). The elevation ranges from 2050 to 3250 m a.s.l. The mean monthly rainfall of the area ranges from 4.4 mm in February to 229.7 mm in July with a total annual mean rainfall of 707.6 mm. Nearly 66% of the annual rainfall occurs only in the months of June, July, and August. According to field observation and data from local people, many of the past landslides occurred during these months. The minimum and maximum mean annual temperature of Mgulat area is 10.7 and 22.8 °C, respectively.

Fig. 1
figure 1

Location map of Mgulat area (with 40 m contour interval and DEM) and its surroundings with reference to Ethiopia and Tigray Regional State

Full size image

The present study is located in both rugged and steep topographic terrain with little road cuttings. As a result, active landslides are common and occur every year in the study area, resulting in damages to settlements and farmlands.

Statement of the problem

Landslide triggered by rainfall is one of the common geohazards in the highlands of Ethiopia, especially in northern and central Ethiopian highlands, mainly responsible for the losses of human life and property. As part of the Ethiopian highlands, in Mgulat area (Fig. 1), a number of damages were reported due to landslides. A recent example is the 24th August 2018 landslide event, which killed six people and destroyed farmland, settlement, and property (Fig. 2a). In addition, a rock fall event in August 2017 claimed the life of one woman near Mgulat village (Fig. 2b). The problem is wide and significant. Emanating from these problems, the aim of this research was to assess and understand the causes and failure mechanisms, carry out the mapping of the active and past landslides, and, finally, generate LSM. The ultimate goal was to improve the current spatial land- use plan and guide and support local and regional administrative bodies in managing landslide risk.

Fig. 2
figure 2

Field photographs of a the 24th August 2018 landslide and b a rock fall event in August 2017. Photographs were taken by the second author in September, 2018

Full size image

Methods and materials

For the accomplishment of this research, the FR method and KA approach were selected considering the reliability and availability of data. Generally, the following main steps or approaches were adopted during the research work:

  1. 1.

    Data collection (Wieczorek 1984; Crozier 2005) and construction of spatial database from which the relevant causative factors were extracted and selected,

  2. 2.

    Assessment and generation of LSM using the relationship between the inventoried landslides and selected site-specific causative factors, and

  3. 3.

    Verification and validation of the generated LSM using the inventoried landslides and other approaches.

The slope steepness, slope aspect, drainage, and elevation maps of the study site were extracted from digital elevation model (DEM) with 30 m resolution generated from NASA’s Shuttle Radar Topography Mission (SRTM) in 2000 (http://dds.cr.USGS.gov/SRTM/version2_1/SRTM3/ Africa/). Topographic maps Adigrat (1439 C2), and Nebelet (1439 C4) sheets with a scale of 1:50,000) (EMA 1997) were used as a base map for field data collections. Regional geological map of Adigrat (ND 37-7) (Garland 1980) with a scale of 1:250,000 and fieldwork campaign were used for the compilation and preparation of the geological map of Mgulat area.

The following calculations (equations below) were used to generate the thematic maps and LSM using FR and the spatial analyst tool of ArcGIS (ESRI 2010). Figure 3 shows the simplified overall methods and approaches.

$$ \mathrm{Area}\ \left({\mathrm{km}}^2\right)=\frac{\mathrm{Number}\ \mathrm{of}\ \mathrm{pixel}\times 900\ {\mathrm{m}}^2}{\mathrm{1,000,000}} $$
(1)

where 900 m2 = 30 m × 30 m pixel (from DEM) and 1,000,000 = 1000 × 1000, since the 30 m × 30 m is in meter and when it is converted to km2, 1000 m × 1000 m = 1,000,000 m2.

Fig. 3
figure 3

Simplified overall research approaches and methods

Full size image

FR is the ratio of occurrence probability to non-occurrence probability for a specific attribute. Several researchers used FR for landslide susceptibility mapping (e.g., Akgun et al. 2008; Donati and Turrini 2002; Lee and Choi 2003; Lee et al. 2002; Luzi et al. 2000; Lee and Talib 2005; Pradhan and Lee 2009; Sun et al. 2018; Vijith and Madhu 2007; Zezere et al. 2004; Zhou et al. 2018). The FR value for each landside conditioning sub-class was determined using Eq. (2). The novelty of this article relies on the use and testing of FR for the first time for Mgulat mountainous area with no existing landslide inventory database.

$$ \mathrm{FR}=\frac{\left(\mathrm{Number}\ \mathrm{of}\ \mathrm{landslide}\ \mathrm{pixels}\ \mathrm{in}\ \mathrm{a}\ \mathrm{sub}-\mathrm{class}/\mathrm{Total}\ \mathrm{number}\ \mathrm{of}\ \mathrm{landslide}\ \mathrm{pixels}\right)\ }{\left(\mathrm{Total}\ \mathrm{number}\ \mathrm{of}\ \mathrm{pixels}\ \mathrm{in}\ \mathrm{a}\ \mathrm{sub}-\mathrm{class}/\mathrm{Total}\ \mathrm{number}\ \mathrm{of}\ \mathrm{pixels}\ \mathrm{in}\ \mathrm{the}\ \mathrm{study}\ \mathrm{area}\right)} $$
(2)

The landslide susceptibility index (LSI) was calculated during the overlay by adding all the FR values of each factor using Eq. (3). The LSI represents the relative susceptibility to landslide occurrence (the greater the value, the higher the susceptibility to landslide occurrence and vice versa) (Chalkias et al. 2014).

$$ \mathrm{LSI}\ \left(\%\right)=\mathrm{FR}\ \left(\mathrm{Lithology}\right)+\mathrm{FR}\ \left(\mathrm{Slope}\right)+\mathrm{FR}\ \left(\mathrm{Distance}\ \mathrm{from}\ \mathrm{stream}\ \right)+\mathrm{FR}\ \left(\mathrm{LULC}\right)+\mathrm{FR}\ \left(\mathrm{Slope}\ \mathrm{aspect}\right) $$
(3)

Geological context

According to Mengesha et al. (1996), the lithological units of Ethiopia fall into three major categories: Precambrian basement, Late Paleozoic to Paleogene sediments, and Cenozoic volcanic and associated sedimentary rocks. The stratigraphic succession of Northern Ethiopia is divided into Precambrian basement (Tsalet group), Enticho sandstones (white sandstone), Edaga Arbi Glacial (tillite), Adigrat sandstones (red sandstone), Antalo limestone, Agula shale, Amba Aradam Formation, Trap series, Mekelle dolerites, and Axum-Adwa plugs (Sembroni et al. 2017). The local geology of the study area was mapped through field investigation (Fig. 4).

Fig. 4
figure 4

Regional geological map of Mgulat area and its surrounding showing the coverage of Mesozoic and Cenozoic sedimentary and volcanic rocks (modified from Garland 1980). The study area boundary is indicated in the center of the map with bold line. Datum: Adindan UTM Zone 37N and projection: transverse Mercator

Full size image

The area consists of five lithological units (Figs. 4 and 5): Quaternary deposits (residual soil and alluvial deposits) (Fig. 5b), basalt (aphanatic, amygdaloidal, agglomeratic, and columnar basalt varieties) (Fig. 5a, c), red sandstone, tillite, and white sandstone. A larger proportion of the study area is covered by basalt (Fig. 4). The tillite unit is weak and impervious, underlain by white sandstone and overlain by red sandstone. The presence of open discontinuities in the red sandstone results in direct percolation of rain water responsible for development of pore water pressure and finally for the failure of slopes in the form of rock fall and topple. The erodibility nature of the tillite unit also contributes to the occurrence of numerous landslides (e.g., debris flow, rock fall, topples, etc.) in the tillite itself (Fig. 5d) and red sandstone unit (Fig. 5a, c).

Fig. 5
figure 5

Field photographs showing the varieties of lithological units and landslide types from Mgulat area. Photographs were taken by the first author in June, 2018

Full size image

Results and discussion

Thematic data layers and landslide inventory mapping

The study began with the preparation of a landslide inventory map based on extensive field work and Google Earth image interpretation (Wieczorek 1984; Crozier 2005). Inventory mapping was used to show and locate the type and characteristics of landslides and related features that occurred in the past. The landslide inventory mapping of the present study was used as a basis for the LSM. Landslide inventory map is very necessary for understanding the relationship between the landslide events and the possible controlling parameters of landslides (Ercanoglu and Gokceoglu 2004). A total of 220 landslides were inventoried and, out of these, 205 landslides were used for analysis, while 15 of them were excluded randomly from the analysis for the validation of the LSM. From the total 220 landslides, some of them are giant blocks that affect large areas, while the others cover very small areas. From rasterized area calculation of the landslides, 0.366 km2 is the largest landslide whereas 155 m2 is the smallest one (Fig. 6). The landslides are grouped into four types based on Varnes (1978) classification system, (a) 92 (41.8%) rock fall, (b) 57 (25.9%) complex, (c) 40 (18.2%) topples, and (d) 31 (14.1%) debris flow.

Fig. 6
figure 6

Landslide inventory map of Mgulat area. Dendritic drainage system with overall flow toward the northwest direction is shown. Datum: Adindan UTM Zone 37N and projection: transverse Mercator

Full size image

Causative factors and landslide occurrence

To produce LSM and minimize losses incurred by landslide events, it is indispensable to create a good understanding of the causative factors. In the present study, numerous causative factors play an important role simultaneously and in a complex manner. The right choice or selection of causative factors depends on the modes and type of landslides, conditions and nature of the study site, extent and scope of the analysis, availability of data set, and analysis methods selected (Tseng et al. 2015). Generally, there is no agreement on the universal guidelines for selecting landslide conditioning factors (Pradhan and Lee 2010) applicable everywhere. In the present study, five causative factors were selected, among eight pre-selected factors, for generating LSM based on their contribution toward landslide initiation and field verification and are among the most commonly used in the literature to assess landslide susceptibility (Pradhan and Lee 2010; Pourghasemi et al. 2013). The main causative factors are lithology, slope steepness, distance to stream, LULC, and slope aspect (Figs. 7, 8a, b, and 9a, b). Distance to road, elevation, and distance to fault were excluded from the analysis due to their weak relationship with landslide occurrence, which was verified by field observation and landslide inventory map. In addition, along most of the main road sides, an engineering solution was already provided, which is a retaining wall made of masonry. Hence, it cannot be a major cause for future landslide occurrences. The relationship between causative factors and landslide occurrence was evaluated based on their FR values.

Fig. 7
figure 7

Lithological map of Mgulat area. Lithology is considered as a landslide conditioning factor. Slight variations with Fig. 4 are attributed to the scale and presence of localized lithological units not included in Fig. 4. Kinematic rock slope stability analysis was conducted at three locations (R1, R2, and R3). Datum: Adindan UTM Zone 37N and projection: transverse Mercator

Full size image
Fig. 8
figure 8

Conditioning factors maps. a Slope aspect (degree) (top) and b distance to stream (m) (bottom). Datum: Adindan UTM Zone 37N and projection: transverse Mercator

Full size image
Fig. 9
figure 9

Conditioning factors maps. a LULC (top) and b slope class (degree) (bottom). Datum: Adindan UTM Zone 37N and projection: transverse Mercator

Full size image

Rainfall and manmade activities are the main extrinsic variables or temporary factors contributing to landslide occurrence in the area. Manmade activities include road cut, construction materials extraction or excavation and deforestation, but their relationship with landslide occurrence was found to be insignificant. Most of the roads’ cut sides are stabilized with masonry retaining structures. The predominant causes for landslide occurrence in the study area are lithology and slope steepness (causative factors) and rainfall (triggering factor). The selection of these factors depends mainly on the availability of data for the study area and their significance or contribution to landslide occurrences (Figs. 7, 8a, b, and 9a, b).

All the selected causative factors (lithology, slope steepness, distance to stream, LULC, and slope aspect) were analyzed to generate LSM using the FR method. For each factor, the FR value of each class was determined to understand the most influencing factors as described in Table 1. Accordingly, factors with FR > 1 have strong relationship, whereas those with FR < 1 have weak relationship with landslide occurrence.

Table 1 FR values of the landslide conditioning or causative parameters (factors)
Full size table

For each of the thematic maps, the FR value for each class was assigned using the ArcGIS spatial analyst tool (Fig. 10). The LSI values of the present study were classified into very low, low, moderate, high, and very high susceptibility classes based on the degree of susceptibility by using the natural breaks provided in ArcGIS. The areal percentage of landslides (Table 2) for very low, low, and moderate susceptibility classes are 1.32, 8.35, and 20.39%, respectively, while percentage distribution for the high and very high classes are higher (24.5 and 45.44%, respectively). Hence, it can be fairly inferred that the landslide susceptibility zones reflect the existing or observed field instability conditions as a function of the landslide inventory map.

Fig. 10
figure 10

Generated landslide susceptibility map of Mgulat area. Datum: Adindan UTM Zone 37N and projection: transverse Mercator

Full size image
Table 2 Statistical summary of the inventoried landslides in the Mgulat Mountain
Full size table

A schematic representation of the mechanisms of the landslides, particularly rock falls and topples, for the study area is depicted in Fig. 11. The schematic representation shows rainfall infiltration through vertical joints and interconnected pores. The infiltrated water produces pore water pressure due to the presence of impervious units below (tillite) and resulted in the occurrence of a variety of landslides (e.g., rock fall, topples). The vertical joints in the red sandstone were developed due to stress relief toward the cliffs.

Fig. 11
figure 11

Conceptual model showing rock fall and topple occurrence and mechanisms (not to scale, approximately in E-W direction)

Full size image

Verification and validation of the LSM

Validation is an essential component in landslide susceptibility mapping to confirm the efficiency, applicability and scientific significance of the methods used (Frattini et al. 2010; Zhou et al. 2018). To verify and validate the generated LSM, four approaches were used as depicted below.

Landslide density or frequency

The LSM obtained from FR was verified through a comparison of the past landslide distribution and the generated LSM, which is sometimes called the accuracy statistic method (Zhou et al. 2018) or landside density or frequency (Khan et al. 2019; Reichenbach et al. 2018). In this approach, the landslide distribution in different susceptibility class was statistically analyzed. From the landslide distribution, 45.5% of the inventoried landslide areas (14.35 km2) fall in the very high susceptibility class, 24.5% fall in the high, 20.4% fall in the moderate, 8.4% fall in the low, and 1.32% fall in the very low susceptibility class. From this verification or comparison of the LSM with inventoried landslides, 70% of the existing landslide areas fall in the high and very high susceptible classes, indicating good performance of FR model (Fig. 12). This shows a gradual increase in percentage distribution of landslide occurrences (in terms of area) from very low susceptible class to very high susceptible class. This result was found to be similar and in agreement with different researchers elsewhere (e.g., Frattini et al. 2010; Sarkar and Kanungo 2004; Yalcin et al. 2011; Yan et al. 2018; Zhou et al. 2018). Moreover, 43.68% of the whole study area is marked as belonging to the high or very high susceptibility classes. Field observations confirmed that areas of high susceptible zones showed significant indications of slope instabilities marked by old landslides, intensive erosion, and steep cliffs with presence of discontinuities liable for future landslide events.

Fig. 12
figure 12

Statistical summary of LSM zones with percentage distribution of past landslides occurrence and total study area

Full size image

Validation using a landslide subset

As an independent verification approach, 15 inventoried landslides randomly selected were excluded from the FR analysis for verification purposes (validating data sets). The 15 inventoried landslides were superimposed over the generated LSM to check its landslide predictability power. The result show that 86.6% (13 of the 15 landslides) fall in the very high and high susceptibility classes and 13.3% (2 of the 15 landslides) fall in the moderate susceptibility zone, whereas none of the landslides fall in the very low and low susceptibility classes (Table 3, Fig. 13). This is a good sign of the reliability and predictability power of the generated LSM.

Table 3 Areal percentage of the 15 validating landslides for the susceptibility classes
Full size table
Fig. 13
figure 13

Superimposed 15 validating landslide areas over the generated LSM

Full size image

Success rate curve

Success rate curve is shown in Fig. 14 which was used to validate the developed landslide susceptibility map (Fig. 10). The success rate curve is generated from the LSI (Khan et al. 2019; Reichenbach et al. 2018). It is a quantitative measure of the predictive performance of the statistical model. It is used to predict the total area of landslide distribution (inventoried map) by comparing the landslide susceptibility index with the cumulative landslide occurrence in terms of areal coverage. Pradhan et al. (2014) and other researchers suggest that prediction power values > 70% in the very high and high susceptibility classes are generally considered satisfactory for prediction of the ground truth. The curve of the present study predicted that 10% of the pixels with the highest LSI values account for 72% of the total landslide pixels (Fig. 14).

Fig. 14
figure 14

Cumulative percentage of study area classified as susceptible (x-axis) in cumulative percent of landslide occurrence (y-axis)

Full size image

Kinematic rock slope stability analysis

Kinematic rock slope stability analysis is used to determine or check whether the slope is kinematically stable or not without consideration of associated forces. The analysis helps to distinguish which discontinuities have favorable conditions that could actually lead to failure with respect to specific slope geometry. Classification and characterization (mainly the discontinuity parameters, Table 4) of rocks are input parameters for the determination of the type of failures in the rock slopes. ROCKPACK III software (Rockware 2003) was used for kinematic slope analysis. The analysis was carried out on three different lithological units (see Fig. 7 for R1, R2, and R3 locations): basalt, red sandstone, and white sandstone (Table 4). These lithologic units are the dominant in the study site. The first rock slope analysis was carried out on the basalt in the southeastern part of the study area (R1 in Fig. 7). The basalt is characterized by moderate to high degree of weathering with some discontinuities of varied orientations. From the analysis, potentially planar, potentially toppling, and marginally wedge types of failures were identified (Fig. 15).

Table 4 Field data for the rock discontinuities and slope characteristics of the selected rock slopes
Full size table
Fig. 15
figure 15

a Outcrop view of rock slope failure surface in basalts. b Stereoplots of three discontinuities. Field photograph was taken by the first author on the 13th of August 2018

Full size image

The second rock slope analysis was performed on the red sandstone in the eastern part of the study area (R2 in Fig. 7). The red sandstone is characterized by closely-spaced and vertical master joints and moderate to high degree of weathering. From the analysis, potentially planar, potentially wedge and marginally topple and planar types of failures were identified (Fig. 16).

Fig. 16
figure 16

a Outcrop view of the rock slope failure in red sandstone and (b) stereoplots of three discontinuities. Field photograph was taken by the second author on the 9th of August 2018

Full size image

The third rock slope analysis was carried out on the white sandstone in the eastern part of the study area (R3 in Fig. 7). The white sandstone is characterized by widely spaced vertical joints and moderate degree of weathering, and it is generally massive with rare occurrence of landslides. From the kinematic analysis, potentially topple and marginally planar types of failures were identified (Fig. 17). Field observation confirmed the presence of potentially unstable blocks of rocks at the selected rock slopes.

Fig. 17
figure 17

a Outcrop view of rock slope failure in white sandstone and b stereoplots of two discontinuities. Field photograph was taken by the second author on the 3rd of August 2018

Full size image

Conclusion and recommendation

Landslide susceptibility assessment and mapping is crucial for sustainable land-use planning and risk reduction in landslide-prone regions. The factors influencing landslide occurrence are many, very complex, and many times act simultaneously. Lithology (red sandstone and tillite), distance to stream (0 to 300 m), LULC (bare land and settlement), and slope steepness (> 25°) resulted in FR > 1, indicating strong relationship with landslide occurrence. It was determined that 19.53 and 24.15% of the study area corresponds to very high and high landslide susceptibility zones in the LSM generated with FR, respectively. Low and moderate susceptible zones make up 22.57 and 18.74% of the total area, respectively. The very low susceptible zone value is 15.01%. Based on the different validation approaches, the FR method was found to be suitable for landslide susceptibility mapping for the area and may be applied to other areas with similar topographic and geological settings. Selected rock slope stability analysis using kinematic approach resulted in potential and marginal failures of planar, wedge, and topple types. From the landslide distribution map, rock fall is the predominant (42% of 220 landslides) type of failure.

LSM is an appropriate mean for the evaluation of the degree of landslide susceptibility of an area before administrative and engineering measures are taken. Areas with high and very high landslide susceptibility should be given priority for solutions and need proper land-use planning. Based on the LSM, the high and very high susceptibility zones with high population density and fertile land should be given emphasis in order to reduce future losses and damages. To implement remedial measures, detailed and site-specific multidisciplinary investigations are recommended for hazard and risk estimations.