JOURNAL OF DEGRADED AND MINING LANDS MANAGEMENT

Volume 11, Number 2 (January 2024):5169-5179, doi:10.15243/jdmlm.2024.112.5169

ISSN: 2339-076X (p); 2502-2458 (e), www.jdmlm.ub.ac.id

Research Article

Mapping landslide susceptibility in the Debretabor-Alember road sector,

Northwestern Ethiopia through geospatial tools and statistical approaches
Betelhem Tesfaye1, Muralitharan Jothimani2*, Zerihun Dawit2
1
Department of Geology, College of Natural & Computational Sciences, University of Gondar, P O Box:196, Ethiopia
2
Department of Geology, College of Natural & Computational Sciences, Arba Minch University, P O Box:21, Ethiopia
*
corresponding author: muralitharangeo@gmail.com

Abstract
Article history: This study aimed to locate areas along the Debretabor-Alember route
Received 19 June 2023 segment in northern Ethiopia that are susceptible to landslides. Geospatial
Revised 29 July 2023 tools, specifically frequency ratios (FR) and information values (IV), were
Accepted 8 August 2023 used to develop landslide susceptibility maps (LSMs). A comprehensive on-
site investigation and analysis of Google Earth imagery were conducted,
resulting in the detection and analysis of 89 landslides, including current and
Keywords: historical events. The dataset used for validation comprised 78% of the
frequency ratio previously documented landslides, whereas the remaining 22% was used for
geospatial training. Several factors were considered in this study to determine landslide
information value susceptibility, including "slope, aspect, curvature, elevation, lithology,
landslide susceptibility distance from streams, land use and cover, precipitation, normalized
Northwestern Ethiopia difference vegetation index (NDVI)", and the FR and IV models. Based on
the results obtained using the FR approach, specific areas exhibited different
levels of susceptibility, ranging from very low to moderately high, medium,
high, and very high. These areas covered a total of 18.4 km2 (19.9%), 18.9
km2 (20.5%), 19.7 km2 (20.3%), 17.7 km2 (20%), and 17.7 km2 (19%),
respectively. The LSMs generated by the IV model indicated multiple
susceptibility classes in the study area, varying from very low to very high.
These maps revealed that 18.4 km2 (19.8%), 18.8 km2 (20%), 18.9 km2
(19.5%), 18.8 km2 (20.5%), and 18.3 km2 (19.8%) of the area fell into these
susceptibility classes. The landslide density indicator method was employed
to validate the LSMs. The FR and IV models demonstrated that a significant
proportion of confirmed past and current landslide records (72.16% and
73.86%, respectively) occurred in regions with a high or very high
susceptibility to landslides. Overall, the IV model, which utilized latent
variable structural modeling (LSM) in the independent variable model,
outperformed the fixed effects regression model (FR).

To cite this article: Tesfaye, B., Jothimani, M. and Dawit, Z. 2024. Mapping landslide susceptibility in the Debretabor-
Alember road sector, Northwestern Ethiopia through geospatial tools and statistical approaches. Journal of Degraded and
Mining Lands Management 11(2):5169-5179, doi:10.15243/jdmlm.2024.112.5169.

Introduction affected by landslides due to various topographical,

geological, hydrological, and tectonic factors.
Ethiopia suffers significant losses yearly due to Approximately 60% of Ethiopia's population lives in
landslides, a critical geological threat. Landslides these highland areas. These areas are characterized by
cause substantial damage to property, agricultural complex geology and high rainfall. As the population
lands, human lives, and animals (Ayalew, 1999; Dai et increases, previously undeveloped regions have
al., 2002; Abebe et al., 2010). Ethiopia's southern, become more general for settlement, urban
western, and northern highlands are frequently development, and agriculture. This can increase the

Open Access 5169

 B. Tesfaye et al. / Journal of Degraded and Mining Lands Management 11(2):5169-5179 (2024)

likelihood of landslides during the rainy season investigate the various causal factors contributing to
(Abebe et al., 2010; Woldearegay, 2013). Landslides landslide susceptibility in the present study area. A
have damaged property and infrastructure in Ethiopia, series of thematic layers of landslide causal factors
resulting in human casualties. Between 1993 and 1998, were mapped, including "slope, aspect, curvature,
Ethiopia saw approximately 300 deaths, 200 homes elevation, lithology, distance from streams, land use
destroyed, more than 100 km of roads damaged, and and cover, precipitation, normalized difference
500 ha of land lost. Between 2018 and 2019, rainfall- vegetation index (NDVI)," and FR and IV models for
triggered landslides in different parts of Ethiopia mapping landslide susceptibility.
resulted in 60 deaths, 30 injured households, 5,091
families displaced, homes destroyed, and cultivated
and uncultivated land damaged (Mersha and Meten, Materials and Methods
2020). Description of the study area
Landslides have been devastating in the
northwestern region of Ethiopia in recent years The research was carried out along a segment of the
(Chimidi et al., 2017; Mersha and Meten, 2020). Debra Tabor-Alember Road located in the Amhara
Debretabor-to-Almeber section is the study area in National Regional State, South Gondar Zone, in the
Ethiopia's northwestern highlands, characterized by northwestern Ethiopian Highlands. In the north, the
hilly terrain. A translational slide on the right damaged region extends from N 1311050 to N 1321492, and in
the road's Debretabor-to-Almeber section, and the east, from E378514 to 392381. The projection
farmland was affected near the stream by a rotating system of the study area was the WGS_1984, UTM 37
slide. The study area is susceptible to landslides, which N zone. It is located 100 km from Bahirdar, the capital
cause severe damage to infrastructure, including roads of the Amhara Regional State. The study covers an
and agriculture. A systematic investigation of area of 92.3 square kilometers. Located between 2717
landslide susceptibility must be conducted to prevent and 1958 m above mean sea level, the study site has a
the loss of life and damage to infrastructure, maximum elevation near Debre Tabor and a minimum
settlements, and agricultural land. These studies are elevation in Alember. There is a difference in
crucial for planning land use and landslide mitigation. elevation of less than 759 m between the study region
Geo-environmental factors determine the causes of and Tana Basin. The region's tributaries are empty in
landslides. These factors can be classified as initiating, the Gumara and Ribe Rivers. Several important rivers
causing primary, or secondary effects (Wang et al., are under study in the area, including Ambawa Wiha
2019). Rainfall, volcanic activity, and human activity (Gebere Matebya), Layiku, Gumara, Klbe, Arbi, and
can cause landslides (Ozioko and Igwe, 2020). Geo- Melo. Gebere Matebya, flowing in the northwest and
environmental factors also influence landslide southwest, eventually joins the Gumara River at the
frequency (Youssef et al., 2016). GIS and remote confluence of the Gumar and Klbe rivers. As the
sensing are crucial for determining landslide remaining rivers travel toward the southeast and
susceptibility zones and minimizing damage and loss northeast, they eventually converge with the Rib River
(Ozioko and Igwe, 2020; Tang et al., 2020). (Figure 1).
GIS can map different trigger factors for Data collection
landslides and identify past and current landslides to
create an inventory map (Zhao, 2022). Landslide risk This study used a digital elevation model (DEM)
assessment techniques can be divided into three obtained from an advanced space-based thermal
primary categories: expert, heuristic, and statistical. emission and reflectance radiometer (ASTER) to
Researchers worldwide have used these techniques to calculate dip and aspect attributes within the study
map landslide susceptibility in various regions. area. The ASTER DEM had a spatial resolution of 30
(Mondal and Maiti, 2013; Nohani et al., 2019; Silalahi m. Sentinel-2 satellite imagery was used to generate
et al., 2019; Yang et al., 2019). GIS can be used to map land use and land cover maps (LULC) and normalized
landslide susceptibility using several statistical differential vegetation index (NDVI) maps. The
methods. These methods can provide reliable and Sentinel-2 images, available for download at
accurate results in complex topographic settings http://gdex.cr.usgs.gov/gdex/, had a resolution of 20
(Youssef et al., 2016; Ba et al., 2017; Yang et al., meters. In addition to the Ethiopian Meteorological
2019). The subsequent studies used GIS, FR, and IV Agency Center, the Tropical Rainfall Measuring
models to map landslide vulnerability. Many Mission (TRMM) also provided precipitation data. A
researchers continue to use these models and have field investigation was conducted to determine
accepted them (Ba et al. 2017; Nohani et al. 2019; landslides' location, size, nature, and severity. As part
Silalahi et al. 2019; Wang et al. 2019). of the fieldwork database, landslide inventory records
This study aimed to identify landslide-prone were included (both for training and validation
areas along the Debretabor-Alember road segment in purposes), as well as causative factors, including slope,
northern Ethiopia using geospatial frequency ratios curvature, height, lithology, distance to roads,
(FR) and information values (IVs). To date, no effort precipitation, LULC, river distance, and NDVI. A
has been made to apply FR and IV models or digital elevation model (DEM) was used to identify the

Open Access 5170

 B. Tesfaye et al. / Journal of Degraded and Mining Lands Management 11(2):5169-5179 (2024)

topographical elements that influence the occurrence geological map of Ethiopia. Sentinel-2 satellite
of landslides in the research area. Lithologies were imagery and Google Earth data were used to extract
determined using fieldwork observations and a the road networks.

Figure 1. Location map of the study area.

Methods FR= (A/B)/(C/D) (1)

Landslide susceptibility mapping Based on the given equation, A represents the number
of landslides used as training sets for each factor, while
During the process of mapping landslide B represents the total number of landslides within the
susceptibility, it is postulated that the factors study area. The number of pixels in the class area
responsible for triggering landslides affect their spatial corresponding to the specific factor is denoted by C. In
arrangement, and it is expected that future landslides contrast, the total number of pixels in the study area is
will manifest under conditions comparable to those represented by D.
observed in the past. This study mapped landslide The frequency ratio derived from Eq 1 is added
susceptibility using the Frequency Ratio (FR) and to generate a landslide susceptibility index (LSI) map
Information Value (IV) methods. by using Equation 2.
Frequency ratio LSI = ∑ Fr1 + Ff2r + Fr3+. . . . . . +Frn (2)
A thorough understanding of the mechanisms and In assessing the relationship between landslide
environmental factors that cause landslides in this occurrence and selected contributing variables, the
research area is essential. It has proven quite frequency ratio (Fr) is used. Here, 'n' represents the
successful in mapping landslide susceptibility total number of these contributing variables.
extensively using the FR approach and GIS In this approach, the ratio in question represents
technologies. It has been demonstrated that there is a the relative extent of the areas affected by landslides,
quantitative relationship between the inventory of with a value of one indicating the average occurrence.
landslides and their causes (Lee and Sambath, 2006; A correlation value above 1 indicates that the
Wu et al., 2016; Ding et al., 2017; Wang and Li, 2017). percentage of landslides exceeded the total area,
Using the factor map and landslide inventory, FRs indicating a higher likelihood of landslide activity.
were calculated for each category of landslide-causing Conversely, a correlation value below 1 indicates a
factor (Mondal and Maiti, 2013). The frequency ratio reduced correlation between the landslides and the
is shown in Equation 1.

Open Access 5171

 B. Tesfaye et al. / Journal of Degraded and Mining Lands Management 11(2):5169-5179 (2024)

area under consideration (Akgun et al., 2007; Wu et distinct classes:0-6o, 6-11o, 11-16o, 16-23o, 23-32o, and
al., 2016). The Landslide Susceptibility Index (LSI) 32-65.5o (Figure 2). Majorly, 20.34% of landslides
map was categorized into different classes to create occurred between 11° and 16°.
landslide susceptibility maps.
Aspect
Information value
In numerous studies (Lee and Sambath, 2006; Jamal
The contribution of each aspect and its interactions can and Mandal, 2016; Hamid et al., 2023), aspects have
be considered when determining the risk of a landslide been generally recognized as significant factors
in a specific area (Sarkar et al., 2013; Mandal et al., influencing landslides. As shown in Figure 1, the study
2018; Sharma et al., 2019)—determined the weight of area was divided into eight aspect classes: Northeast,
each layer of factors using the information value Northeast, Southeast, Southwest, West, Northwest,
technique by comparing the landslide densities of each West, and East (Figure 2). The distribution of past
class to those of the overall region. This method landslides concerning these aspect classes showed a
considers the association between landslides and descending order as follows: Southwest (18.63%),
landslide factor classes and is a probabilistic approach West (17.28%), Northeast (15.66%), Northwest
for bivariate statistical analysis. These numbers should (14.31%), North (11.25%), South (8.55%), East
be used to assess the contribution of each causal factor (8.28%) and Southeast (6.03%).
to various landslide types (Sharma et al., 2019). This
Elevation
method has been adopted by several researchers
(Sarkar et al., 2013; Mandal et al., 2018; Sharma et al., As previously reported (Mandal et al., 2018), altitude
2019). Using Equation 3, the information weight or is essential in landslides, particularly in harsh
value (wi) can be calculated. mountainous regions. Topographic relief's maximum
( )/ ( ) and minimum heights were used to classify the relief
IV = ln ∑ (3) (Nohani et al., 2019). This determines the overall
( )/ ∑ ( )
geomorphological processes. Using the digital
where: w i = weight of a factor class; Npix (Si) = a elevation model (DEM), the study area was divided
number of the pixel of the landslide within class i; Npix into six classes according to elevation:1958-2104,
(Ni) = a number of the pixel of class i; ∑Npix(Si) = a 2104-2216, 2216-2312, 2312-2412, 2412-2536, 2536-
number of the pixel of the landslide within the whole 2216, and 2536-2712 m (Figure 2). According to the
study area; ∑Npix(Ni) = a number of the pixel of the landslide inventory information evaluation, 43.02% of
entire study area. the landslides occurred between 2312 m and 2440 m
The natural logarithm determined negative and above sea level. In addition, about 19.98% of past
positive weights for landslide density deviations from landslides occurred in the 2412-2536 m class, 15.57%
the normal range. Weights were computed for cases in the 2216-2312 m class, 14.13% in the 2104-2206 m
involving landslides that exceeded the normal range. class, 5.94% in the 2536-2717m class and 1.35% in the
The positive weights indicated a direct correlation altitude class of 1958-2104m.
between the predictor factor and landslides, whereas Curvature
negative weights indicated an inverse relationship. We
utilized a class-by-class calculation approach for The curvature of a slope can be characterized by the
determining landslide susceptibility maps, where rate at which the slope changes in a given direction
weighted numbers were incorporated into the (Youssef et al., 2016; Nohani et al., 2019). Slope
calculations. curvature is crucial for assessing slope stability,
The landslide susceptibility index (LSI) is regulating water flow within slopes, and identifying
computed by summing the weight values of each factor landslides in a particular direction (Ayalew et al.,
using Equation 4. 2004; Youssef et al., 2016; Meten et al., 2015). A
positive rake curvature indicates a convex upward
LSI = IV1+IV2+IV3……. +IVn (4) shape, while a negative curvature indicates a concave
IV1 represents the information value, and n represents shape. DEM was used in this study to calculate the
the total number of selected contributing variables. curvature of the study area, which was then classified
into flat, concave, and convex (Figure 2). According to
Preparation of causative factors the analysis of landslide data, the highest percentage of
The following sections discuss the preparation of landslides (34.47%) occurred in flat areas. In addition,
various causative factors used in the present study. 33.39% of the landslides were observed in convex
areas, whereas 32.13% occurred in concave areas.
Slope
Lithology
The stability of a slope depends on the interaction
between the slope angle and material characteristics The importance of comprehensive lithological analysis
such as friction angle or permeability. The slope angle in landslide analysis cannot be overstated, as variations
for the study area was calculated using a digital in lithology lead to differences in soil strength and
elevation model (DEM) and classified it into six permeability (Meten et al., 2015; Bahrami et al., 2020).

Open Access 5172

 B. Tesfaye et al. / Journal of Degraded and Mining Lands Management 11(2):5169-5179 (2024)

Open Access 5173

 B. Tesfaye et al. / Journal of Degraded and Mining Lands Management 11(2):5169-5179 (2024)

Figure 2. Landslide causative factors.

Extensive fieldwork was conducted at a scale of stream: 0-88 m, 88-186 m, 186-294 m, 294-429 m, and
1:50,000 to create a detailed lithological map of the 429-832 m. The distance to the stream in the past
study area (Figure 2), which revealed six primary landslide decreased by 36.54%, 26.46%, 17.55%,
lithological units within the region: weathered basalt, 15.66%, and 3.79% in the first, third, fourth, and fifth
rhyolite, tuff, trachyte, residual soil, and alluvial classes, respectively.
deposits. According to an analysis of past landslide
Distance to road
distribution based on lithology, the tuff had the highest
proportion of landslides (43.83%). It was estimated According to (Meten et al., 2015; Bahrami et al.,
that 34.38% of the landslides consisted of weathered 2020), road distance is one of the principal causes of
basalt, 15.66% trachyte, 5.76% residual soil, and landslides in mountainous terrain. Road construction
34.75% alluvial deposits. that alters topography or reduces slope stability can cut
slopes (Wang et al., 2019). To map the road network
Distance to stream
in this area, Google Earth data was used, and it was
Streams can adversely affect slope stability due to divided into five groups: 0-4885 m, 4855-1055 m,
erosion and saturation of materials in lower areas (Dai 1055-1657 m, 2310-3081 m, and 3081-4269 m (Figure
et al., 2002; Mondal and Maiti, 2013; Wang et al., 2). The landslide distribution decreased with
2019). Distance from the stream is crucial for slope increasing road distance by 47.888%, 24.30%,
stability (Wang and Li, 2017). This study delineated a 13.77%, 7.47%, and 6.57%, respectively.
stream network using digital elevation models (DEMs)
Rainfall
and topography maps to evaluate its influence on slope
stability in the study area. As shown in Figure 2, five Landslides are frequently caused by rainfall
categories were used to classify the distance from the (Woldaregay, 2013). As a result of rainfall, landslides

Open Access 5174

 B. Tesfaye et al. / Journal of Degraded and Mining Lands Management 11(2):5169-5179 (2024)

are caused by saturated soil and an increase in the pore roads and streams. The landslide types in the current
water pressure (Wang et al., 2019). As precipitation location were mainly transitional slides, rotational
increases, the strength of the water-filled particles slides, flows, falls, and lateral spreads, in decreasing
decreases. The TRMM data were used to generate a order. Landslides are initiated by steep slopes,
rainfall map of the study area, which was categorized weathering, erosion, and undercutting by streams and
into five distinct classes (1809-2060 mm; 2060-2176 human activities such as road construction, quarry
mm; 2176-2289 mm; 2434-2653 mm) (Figure 2). The activity, and farm practices. Slides occur in tuff and
past landslides occurred predominantly within Class 2 residual soils along the study area's roads, rivers, and
(77.31%), followed by Class 3 (17.55%) and Class 1 hillsides. Translational and rotational slides were
(5.04%). observed during fieldwork and in Google Earth image
Flow: This type of flow can occur as earth and debris
Land use and land cover
flows and commonly occurs in tuff and soils along a
Land use/land cover (LU/LC) in hilly terrain road section on steep slopes and streams in the present
contributes to slope failure because of anthropogenic study area. Earth and debris flows were observed
activities. Previous studies have demonstrated that during the fieldwork (Figure 3a). Falls occurred in
land use and cover (LULC) are essential in landslide moderately weathered basalt and tuff lithological units
occurrence (Meten et al., 2015; Bahrami et al., 2020). along the road section and hillsides of the study area.
Sentinel-2 satellite imagery was used as the input data In the study area, 89 landslide locations were identified
for supervised classification in ArcGIS to generate the using Google Earth images during fieldwork. Forty-six
LULC map within the study area. The LU/LC landslides were directly observed during field surveys,
categories in the study area included wetlands, while the remaining 43 were identified through
agriculture, settlements, grasslands, lakes, rivers, and analysis of Google Earth images. The landslides were
barren land (Figure 2). Among the LULC types, the categorized based on their spatial distribution in the
incidence of past landslides on agricultural land was training and validation sets. The model was
the highest (62.02%). Landslide prevalence increased constructed using 64 trained landslides (78%). In
in dense forests (13.95%), barren land (10.35%), contrast, the validation landslides, which totaled 25
grasslands (9.81%), settlements (3.33%), wetlands points (22%), were used to validate the model output.
(0.45%), and bodies of water (0.09%).
Landslide susceptibility zone map
Normalized Difference Vegetation Index
Both models revealed clear and identifiable patterns in
An important component for assessing landslide the correlation between landslide occurrence and the
vulnerability is the use of vegetation indices (Wang et factors contributing to their causes. There was a low
al., 2019). The NDVI (Normalized Difference probability of landslides at slope angles below 16°.
Vegetation Index) is an important index derived from The ratio was less than one and approached zero when
reflectance measurements in the red and near-infrared the slope angle was less than 16°. The ratio was closer
regions of the electromagnetic spectrum. NDVI can be to 1 for slope angles of 16° or more significantly,
used to determine whether photosynthetically active suggesting a higher likelihood of landslides. Debris
biomass is present during imaging (Ercanoglu, 2005). flows, and translational slides were included in this
As part of this study, the NDVI was utilized to validate category.
land cover maps and land use. NDVI values generally Additionally, the possibility of landslides
range from -1 to 1, with higher values indicating a increased with increasing slope angle. Regarding
greater vegetation density. Therefore, landslides aspect class, slopes facing southwest, northeast, or
exhibit an inverse relationship with vegetation density west have a high probability of landslides occurring
(Rasyid et al., 2016). The NDVI values were divided when their ratio is close to one. The ratios for the other
into six classes: -0.49 to -0.02, -0.02 to 0.10, 0.10 to classes - northeast, southwest, southeast, south, and
0.27, 0.27 to 0.49, 0.49 to 0.77, and 0.77 to 1 (Figure southeast–were close to one with a negative value,
2). In the analysis of historical landslide data, it was indicating a low likelihood of landslides in those areas.
found that 29.25 percent of landslides occurred within The relationship between the occurrence of
the NDVI class in the range of 0.02 to 0.10. landslides and elevation showed variations at different
elevation ranges within the study area. Landslides
were more likely in the altitude ranges of 2312-2412
Results and Discussion m, 2412-2536 m, and 2104-2216 m. Conversely, 1958-
Landslide inventory in the study area 2104, 2216-2312, and 2536-2717 m classes showed
near-zero or negative ratios, suggesting that landslides
A detailed field survey has been conducted, and were unlikely in these specific altitude ranges. In
Google Earth images have been used to create a addition, landslides showed a higher probability of
landslide inventory map. Extensive field studies were slopes characterized by concave or convex curvatures.
conducted using GPS to identify the types of The proximity between the first and second distance
movement and materials. Figure 3 shows landslides in classes (0-485 m, 485-1055 m) also showed a positive
the study area were predominantly distributed along value, indicating an increased probability of

Open Access 5175

 B. Tesfaye et al. / Journal of Degraded and Mining Lands Management 11(2):5169-5179 (2024)

landslides. This result suggests that the presence of and Gebere Matebya rivers, is also dominated by very
unstable lithology in the study area is a significant low and low susceptibility zones. Several factors
contributor to the occurrence of landslides. However, contribute to the significant occurrence of landslides
other classifications of distance to streams indicated a in these zones, including the lithology of the tuff, slope
lower likelihood of landslides. Notably, the first and angles, and human activities such as quarrying and
fourth distance classes had positive values close to road cutting.
one, implying a significant risk of landslides,
Validation of the models
especially for areas associated with third-order
streams. Landslide Density Index (LDI)
On the other hand, the other class has a value
Validation is crucial when evaluating the reliability,
close to zero with a negative value, indicating that
landslides are unlikely to occur. Among the lithology accuracy, and effectiveness of a Landslide
types, tuff and weathered basalt units are highly Susceptibility Zone (LSZ) map (Jamal and Mandal,
2016). They play a fundamental role in understanding
susceptible to landslides. Landslides are strongly
map quality, refining the map production process,
associated with weathered basalt because of their steep
producing hazard maps (probability maps), and
slope angle. By contrast, tuff tends to be less resistant
evaluating models derived from different
to lithology because it is exposed along roads and in
small quarries that overlie basalt. Trachyte, alluvial, methodologies (Mandal et al., 2018). The Landslide
and residual deposits had a low likelihood of causing Density Index (LDI) method was used to validate the
LSZ map in this study. Both models were rated using
landslides. Landslides are more likely to occur in
the LDI. For validation, landslide-affected areas not
grasslands and wetlands due to weathering and
considered in the modeling were flagged as potential
erosion. However, other land use and land cover types
exhibited a lower probability of landslides. Based on future landslide areas. The LD index was used to
NDVI values between 0.10-0.27, 0.27-0.49, and 0.49- validate the model, and the percentage of landslide
pixels was compared to the percentage of pixels in
0.77, the ratio is close to zero and positive. In these
each class on the landslide susceptibility map (Mandal
NDVI ranges, there was a high probability of landslide
et al., 2018). Other researchers have widely adopted
occurrence. In contrast, the other class exhibited a
this method (Mersha and Meten, 2020; Nachappa et
lower likelihood of landslides. Landslides are
generally attributed to slope, lithology, rainfall, al., 2020). The success of validating the landslide
curvature, and elevation, among the 11 causative susceptibility models was indicated by an increasing
value of the landslide density index, which ranged
factors.
from very low to very high. According to Dai et al.
Frequency ratio and landslide susceptibility index (2002), landslide data should predominantly fall into
high or very high classes to successfully validate
By reclassifying the landslide susceptibility index in
landslide susceptibility models.
the range of 0.4 to 3.7, a map was created that
The validity of the landslide susceptibility maps
illustrates the varying degrees of landslide
was confirmed by observing a gradual decrease in the
susceptibility in the study area (Figure. 3b). This study
landslide density values from zones of high
employs a quantile classification method. Five
susceptibility to areas of very low susceptibility. These
susceptibility zones were identified: very low (0.4-
values were significantly higher than those of the other
1.2), low (1.2-1.7), moderate (1.7-2.2), high (2.2-2.8),
classes, highlighting their importance. In summary, the
and very high (2.8-3.7). The respective coverage areas
zones with high landslide susceptibility had the highest
for the classes of very low, low, moderate, high, and
percentage or density of landslides, further confirming
very high susceptibility were 19.9% (18.4 km2), 20.5%
the reliability of the landslide susceptibility maps
(18.9 km2), 20.6% (18.7 km2), 20% (18.5 km2), and
produced by both models. The modified frequency
19% (17.7 km2). As shown in Figure 3b, the study
ratio model showed that 50.57% of the landslides fell
area's central, northwest, and southeast regions are
in the high susceptibility zone, followed by 21.59% in
primarily characterized by very high and high
the moderate susceptibility zone, 14.77% in the low
vulnerability zones. It is important to note that these
susceptibility zone, and 8.52% in the very low
zones include significant areas, such as the Arbi River
susceptibility zone. By contrast, the information value
and its tributaries and the villages of Dangors and
model showed that 55.11% of the validated landslides
Gundiba that lie along the roads and hillsides in the
occurred in the very high vulnerability zone, 18.75%
study area. However, most zones of moderate
in the high vulnerability zone, 15.91% in the
susceptibility were found in the southeast and
intermediate vulnerability zone, and 10.23% in the low
southwest regions of the study area. In this region are
vulnerability zone. Combining the high- and very-
the Layiku River and its tributaries, as well as Kolay,
high-susceptibility zones, the information value model
Boda, Chalmut, and the Dangors and Gundiba sectors
accounted for 73.86% of the validated landslides. This
and the surroundings of the Amoragrdel villages. The
indicates that the information value model is better
southeast portion of the study area, including
suited than the modified abundance ratio model to
Teraroch, Duy, Chalmut, Sostu Tsegur, Wayibla, and
accurately identify areas prone to landslides.
Amoragedel Villages, as well as the banks of the Melo

Open Access 5176

 B. Tesfaye et al. / Journal of Degraded and Mining Lands Management 11(2):5169-5179 (2024)

b c

Figure. 3 (a). Landslide inventory, (b). Frequency ratio method's landslide susceptibility map, (c). Information
value method's landslide susceptibility map

Conclusion significant factors include slope, aspect, relative relief,

land use and land cover (LULC), and geological
Evaluation of landslide susceptibility is crucial for structures. A substantial positive correlation exists
conducting risk and hazard analyses. In this study, the between slope, susceptibility map, relative relief, and
spatial distribution of landslides was predicted using precipitation. Constructed reinforced concrete walls
two GIS techniques: the information value and are recommended in regions with a high likelihood of
frequency ratio. The investigation revealed that certain landslides. For slope stabilization, additional measures
factors contributed substantially to the landslide should include roadside drainage systems, dewatering
occurrence. Excessive rainfall was the most influential techniques, vegetative cover, buttressing/riprap cover,
factor in landslides in the present study area. Other retaining walls, and soil riveting. It is essential to

Open Access 5177

 B. Tesfaye et al. / Journal of Degraded and Mining Lands Management 11(2):5169-5179 (2024)

maintain adequate ground cover to avoid erosion of Ercanoglu, M.U.R.A.T. 2005. Landslide susceptibility
sand-covered slopes. Unforeseen construction assessment of SE Bartin (West Black Sea region,
activities, including road cutting and housing Turkey) by artificial neural networks. Natural Hazards
development, can cause slope instability, which must and Earth System Sciences 5(6):979-992.
Hamid, B., Massinissa, B. and Nabila, G. 2023. Landslide
be avoided. Deforestation and conversion of forested susceptibility mapping using GIS-based statistical and
slopes into urbanized areas should also cease. To machine learning modeling in the city of Sidi Abdellah,
mitigate frequent slope failures, it is preferable to Northern Algeria. Modeling Earth Systems and
regularly maintain rainfall outlets on upper slopes, Environment 9:2477-2500, doi:10.1007/s40808-022-
develop roadside plantations, and employ proper 01633-x.
planning for road construction. Jamal, M. and Mandal, S. 2016. Monitoring forest dynamics
and landslide susceptibility in Mechi–Balason
interfluves of Darjiling Himalaya, West Bengal using
Acknowledgments forest canopy density model (FCDM) and Landslide
Susceptibility Index model (LSIM). Modeling Earth
The authors would like to acknowledge the United States Systems and Environment 2:1-17, doi:10.1007/s40808-
Geological Survey (USGS), Center for Hydrometeorology 016-0243-2.
and Remote Sensing (CHRS), University of California, Lee, S. and Sambath, T. 2006. Landslide susceptibility
Irvine (UCI), and Ethiopian Soil Information System mapping in the Damrei Romel area, Cambodia using
(EthioSIS) for providing optical and DEM satellite products, frequency ratio and logistic regression models.
rainfall, and geology data conduct this present study. Environmental Geology 50:847-855,
doi:10.1007/s00254-006-0256-7.
Mandal, S.P., Chakrabarty, A. and Maity, P. 2018.
References Comparative evaluation of information value and
Abebe, B., Dramis, F., Fubelli, G., Umar, M. and Asrat, A. frequency ratio in landslide susceptibility analysis along
2010. Landslides in the Ethiopian highlands and the rift national highways of Sikkim Himalaya. Spatial
margins. African Earth Science 56:131-138, Information Research 26:127-141, doi:10.1007/s41324-
doi:10.1016/j.jafrearsci.2009.06.006. 017-0160-0.
Akgun, A., Dag, S. and Bulut, F. 2007. Landslide Mersha, T. and Meten, M. 2020. GIS-based landslide
susceptibility mapping for a landslide-prone area susceptibility mapping and assessment using bivariate
(Findikli, NE of Turkey) by likelihood-frequency ratio statistical methods in Simada area, northwestern
and weighted linear combination models. Engineering Ethiopia. Geoenvironmental Disasters 7:20,
Geology 54:1127-1143, doi:10.1007/s00254-007-0882- doi:10.1186/s40677-020-00155-x.
8. Meten, M., Bhandary, N.P. and Yatabe, R. 2015. Effect of
Ayalew, L. 1999. The effect of seasonal rainfall on landslide factor combinations on the prediction accuracy
landslides in the highlands of Ethiopia. Bulletin of of landslide susceptibility maps in the Blue Nile Gorge
Engineering Geology and the Environment 58:9-19. of Central Ethiopia. Geoenvironmental Disasters 2:9,
Ayalew, L., Yamagishi, H. and Ugawa, N. 2004. Landslide doi:10.1186/s40677-015-0016-7.
susceptibility mapping using GIS-based weighted linear Mondal, S. and Maiti, R. 2013. Integrating the analytical
combination, the case in Tsugawa area of Agano River, hierarchy process (AHP) and the frequency ratio (FR)
Niigata Prefecture, Japan. Landslides 1:73-81, model in landslide susceptibility mapping of Shiv-khola
doi:10.1007/s10346-003-0006-9. watershed. International Journal of Disaster Risk
Ba, Q., Chen, Y., Deng, S., Wu, Q., Yang, J. and Zhang, J. Science 4(4):200-212, doi:10.1007/s13753-013-0021-y.
2017. An improved information value model based on Nachappa, T.G., Ghorbanzadeh, O., Gholamnia, K. and
gray clustering for landslide susceptibility mapping. Blaschke, T. 2020. Multi-hazard exposure mapping
ISPRS International Journal of Geo-Information, using machine learning for the State of Salzburg,
doi:10.3390/ijgi6010018. Austria. Remote Sensing 12(17):.2757.
Bahrami, S., Rahimzadeh, B. and Khaleghi, S. 2020. Nohani, E., Moharrami, M., Sharafi, S., Khosravi, K.,
Analyzing the effects of tectonic and lithology on the Pradhan, B., Pham, B.T., Lee, S. and Melesse, A.M.
occurrence of landslide along Zagros ophiolitic suture: a 2019. Landslide susceptibility mapping using different
case study of Sarv-Abad, Kurdistan, Iran. Bulletin of GIS-based bivariate models. Water 11(1402):1-22,
Engineering Geology and the Environment 79:1619- doi:10.3390/w11071402.
1637, doi:10.1007/s10064-019-01639-3. Ozioko, O.H. and Igwe, O. 2020. GIS-based landslide
Chimidi, G., Raghuvanshi, T.K. and Suryabhagavan, K.V. susceptibility mapping using heuristic and bivariate
2017. Landslide hazard evaluation and zonation in and statistical methods for Iva Valley and environs Southeast
around Gimbi town, western Ethiopia - A GIS-based Nigeria. Environmental Monitoring and Assessment
statistical approach. Applied Geomatics 9:219-236, 192:119, doi:10.1007/s10661-019-7951-9.
doi:10.1007/s12518-017-0195-x. Rasyid, A.R., Bhandary, N.P. and Yatabe, R. 2016.
Dai F.C., Lee, C.F. and Ngai, Y.Y. 2002. Landslide risk Performance of frequency ratio and logistic regression
assessment and management: an overview. Engineering model in creating GIS based landslides susceptibility
Geology 64:65-87, doi:10.1016/S0013-7952(01)00093- map at Lompobattang Mountain, Indonesia.
X. Geoenvironmental Disasters 3(19):1-16,
Ding, Q., Chen, W. and Hong, H. 2017. Application of doi:10.1186/s40677-016-0053-x.
frequency ratio, weights of evidence and evidential Sarkar, S., Roy, A.K. and Martha, T.R. 2013. Landslide
belief function models in landslide susceptibility susceptibility assessment using Information Value
mapping. Geocarto International 32(6):619-639, Method in parts of the Darjeeling Himalayas. Journal of
doi:10.1080/10106049.2016.1165294.

Open Access 5178

 B. Tesfaye et al. / Journal of Degraded and Mining Lands Management 11(2):5169-5179 (2024)

the Geological Society of India 82:351-362, Woldearegay, K. 2013. Review of the occurrences and
doi:10.1007/s12594-013-0162-z. influencing factors of landslides in the highlands of
Sharma, S. and Mahajan, A.K. 2019. Information value- Ethiopia: With implications for infrastructural
based landslide susceptibility zonation of Dharamshala development. Momona Ethiopian Journal of Science
region, northwestern Himalaya, India. Spatial 5(1):3-31.
Information Research 27:553-564, doi:10.1007/s41324- Wu, Y., Li, W., Wang, Q., Liu, Q., Yang, D., Xing, M., Pei,
019-00259-z. Y. and Yan, S. 2016. Landslide susceptibility assessment
Silalahi, F.E.S., Arifianti, Y.P. and Hidayat, F. 2019. using frequency ratio, statistical index and certainty
Landslide susceptibility assessment using frequency factor models for Gangu County, China. Arabian
ratio model in Bogor, West Java, Indonesia. Geoscience Journal of Geosciences 9(2):1-16, doi:10.1007/s12517-
Letters 6(10):1-17, doi:10.1186/s40562-019-0140-4. 015-2112-0.
Tang, R.X., Kulatilake, P.H., Yan, E. and Cai, J.-S. 2020. Yang, Y., Yang, J., Xu, C., Xu, C. and Song, C. 2019. Local-
Evaluating landslide susceptibility based on cluster scale landslide susceptibility mapping using the B-
analysis, probabilistic methods, and artificial neural GeoSVC model. Landslides 16:1301-1312,
networks. Bulletin of Engineering Geology and the doi:10.1007/s10346-019-01174-y.
Environment 79:2235-2254, doi:10.1007/s10064-019- Youssef, A.M., Pourghasemi, H.R., Pourtaghi, Z.S. and Al-
01684-y. Katheeri, M.M. 2016. Landslide susceptibility mapping
Wang, Q. and Li, W. 2017. A GIS-based comparative using random forest, boosted regression tree,
evaluation of analytical hierarchy process and frequency classification and regression tree, and general linear
ratio models for landslide susceptibility mapping. models and comparison of their performance at Wadi
Physical Geography 38(4):318-337, Tayyah Basin, Asir Region, Saudi Arabia. Landslides
doi:10.1080/02723646.2017.1294522. 13:839-856, doi:10.1007/s10346-015-0614-1.
Wang, Q., Guo, Y., Li, W., He, J. and Wu, Z. 2019. Zhao, P., Masoumi, Z., Kalantari, M., Aflaki, M. and
Predictive modeling of landslide hazards in Wen County, Mansourian, A. 2022. A GIS-based landslide
northwestern China based on information value, susceptibility mapping and variable importance analysis
weights-of-evidence, and certainty factor. Geomatics using artificial intelligent training-based methods.
Natural Hazards and Risk 10(1):820-835, Remote Sensing 14:211, doi:10.3390/rs14010211.
doi:10.1080/19475705.2018.1549111.

Open Access 5179