Landslide Susceptibility Zonation Using Geospatial Technologies and Multi Criteria Evaluation Techniques in Upper Didessa Sub-basin, Southwest Ethiopia

Landslide is a serious geo-hazard that poses destruction and loses of life in different part of the world. The severity of the problem is higher in developing countries like Ethiopia. This study is aimed at assessing the spatial landslide susceptibility in the upper Didessa sub-basin using GIS and multi criteria evaluation (MCE) techniques. In order to reach this objective both primary (eld survey) and secondary data (expert interview, literature, remote sensing data, digital soil map and geological map) were obtained from various source. Eleven landslide causative factor identied in this research are slope, aspect, drainage density, topographic wetness index (TWI), stream power index (SPI), topographic ruggedness index, hypsometric integral, lithology, LULC, soil texture, and distance from road. The analytical hierarchy process (AHP) method was employed to identify the weight of each indicator from the pairwise comparison matrix. The weighted linear combination was then used to generate landslide susceptibility map (LSM). Based on landslide susceptibility, the study area was classied into very high, high, moderate, low, and very low susceptibility zones. Finally, based on the eleven-landslide causative factor analysis, about 24% of the study area is moderately susceptible, while 12% and 6% were classied as high and very high susceptibility to landslides, respectively. The results of this study could help decision makers for future landslide hazardous preventions and mitigation strategies. though, there is a shortage scientic information in the study area, there are various reports which indicates the severity of landslides in the study area. this study makes an attempt at assessing landslide susceptibility zonation by integrating GIS and MCE techniques. This study provides an important information to minimize the potential risks of landslides in the study area.


Background
Landslide is occurred when the bulk of soils separated from its original location and moved to down the hills. Landslides are common natural hazardous occurred worldwide and results in considerable economic loss, life causalities and adverse impacts on infrastructure and environment (Lombardo et  So far, different scholars used various factors for landslide susceptibility mapping, for instance, Youssef and Pourghasemi (2020) used eleven factors "Lithology, lineaments, geomorphology, soil type and depth, slope angle, slope aspect, curvature, altitude, engineering properties of the lithological material, land use patterns, and drainage networks".
Others study by Othman et al. (2012) used ten parameters namely "Slope, lithology, soil properties, geomorphology, land use, aspect, elevation, rainfall, proximity to road, and proximity to the river''. Youssef and Pourghasemi (2020) used a total of twelve parameters for landslide susceptibility mapping in Saudi Arabia recently. Other studies by Chen and Li (2020);  Kamp et al.,2008;Carver, 1991). Even though, there is a shortage scienti c information in the study area, there are various reports which indicates the severity of landslides in the study area. this study makes an attempt at assessing landslide susceptibility zonation by integrating GIS and MCE techniques. This study provides an important information to minimize the potential risks of landslides in the study area.

The Study Area
The study area is located in Oromia National Regional State in three district of Jimma Zone under Didessa Sub basin: Goma, Gumay, and Setema south west Ethiopia. It covers 1500.39 Km 2 ‚within 07 0 40' 0" to 8 0 10' 0"N latitude and 36 0 10' 3" to 36 0 40' 0" E longitude (Fig. 1).The topography of the study area is characterized as lowland plains and plateau with‚ some undulating to steep land forms ‚including depressions and valley oors. The terrain elevation ranges from 1363 m to 2617 m a.s.l. Acrisol, Luvisol, and Cambisol, Nitosol leptosol, lixisol and vertisol are the dominat soil types in the study area. The climate of the study area is cool and humid tropical climate.

Methods And Materials
Five data sources were acquired for landslide susceptibility mapping. These includes: (1) eld survey, which is collected from the eld with the support of Global Positioning Systems(GPS) and experts interview, (2) satellite data for LULC, slope, aspect, topographic wetness index (TWI), stream power index (SPI), topographic ruggedness index (TRI), hypsometric integral (HI) and drainage identi cation, (3) geological data for understanding the existing rocks and its formation in the study area (4) soil data for soil texture and (5) google earth for road network identi cations.
Eleven landslide controlling factor were identi ed by stakeholders and literature review, which was used for landslide susceptibility mapping and zonation: slope, aspect, topographic ruggedness index and hypsometric integral, proximity to road‚ lithology, drainage density, TWI, and SPI, environmental factor (LULC) and soil factors (soil texture). After controlling factors identi cations, new value was assigned (standardized), ranking and scoring landslide susceptibility factor in AHP matrix was performed. AHP was developed by Saaty in the 1970's (Saaty, 1977)

Results And Discussions
In the present study, landslides susceptibility areas were identi ed and collected with the support of GPS along the periphery, and wherever, inaccessibility condition we supported with Google Earth image, then point and polygon data for were prepared as KML format which was later converted into vector (Fig. 3). Landslides near to crop land is very serious and common in the study area (Fig. 4).
The major soil texture in the study area is sandy loam, clay and clay loam accounting 37%, 19.84% and 17.2%, respectively. Whereas, sandy clay loam and loam share about 15.54% and 10.22% of the total cover area, respectively (Table, 1 Regarding landslide occurrence in each category of soil texture in the study area sandy loam accounts the greater share and is mostly distributed in the west to south western part of the study area (Fig. 5). Textural classes are distributed where past landslide is occurred i.e. sandy loam has greater in uence on landslide. Moreover, sandy loam has manifested the greater share (52%) of total landslide occurrences in 37.2% of the total study area. Study by (Maidment,1992) indicates that sandy loam has maximum value in terms of angle of friction and porosity which has a direct relation with slope instability.
Assigning rank for LULC From the LULC, the vegetative areas are less susceptible to landslides in the areas. Vegetation can adversely in uence soil stability (Subramani and Krishnan, 2015). Woldaregay (2013) also con rmed that deforestation and land degradation are the major contributing factors for landslides hazardous in Ethiopia. The maximum probability of landslide occurrence was observed in cultivated land with 9 landslide incidences out of 25, built-up is the second land use that have manifested 7 landslides, whereas, dense forest, sparse forest and grazing land showed a low probability of landslide hazard index.
Thus, the incidence of landslide is inversely related to the vegetation density. Hence, built-up and barren slopes are more prone to landslide activity.
Landslide frequency is mostly found at mountainous and hilly regions (Sharma et al., 2012), where forests have been severely destroyed due to the deforestation (Mersha and Meten, 2020). About 57% of the basin is dominated by low to high density of vegetation. The cultivated land covers 15%, grassland, built-up accounts 13% and 12%, respectively. The remaining 4% are bare land ( Table 2). In summer season, the hilly and mountainous area that have less covered vegetation and utilized for agriculture and barren land are more susceptible to landslide.  In AHP, all factors are compared pairwise in terms of the intensity of their importance using a continuous 1-9-point scale (Table, 3). This scale enables the decision-maker to incorporate experience and knowledge intuitively (Ladas et al., 2007).
In AHP the importance matrix needs to be analyzed by Eigen value for normalization. To normalize the value, we divide the value by its total column and to calculate weight we consider the mean value of the rows. The most critical threat to landslides is slope, as they are subjected to gravitational force. The second is drainage density as they accelerate erosion process. As the value of the stream density increases, the susceptibility of landslide increase.
Soil texture also plays a vital role in this area for landslide activity. Hence, these are given higher rating, whereas loam and clay loam are less prone to landslide as observed in the eld. The LU/LC considered in this study area are dense forest, sparse forest, grass land, bare land, settlement and cultivated area in steep slope area. The landslide occurrence probability value is higher in Barren land, sparse forest, and cultivated land area and lower in dense forest. TWI was classi ed into ve classes and ratings were given in increasing order as TWI value increases. The road construction most often causes slope instability therefore, distance of 50 m from the road side is considered the most prone to landslide activity than the rest of area. The nal step is calculating consistency ratio (CR) to consider how the judgment is relatively correct or not. If CR > 0.1 our judgment is not accepted and if it is less than 0.1 it is accepted according to Saaty (1980). The CR is calculated as follows: Where µmax is the principal Eigen value and n is the number of parameters employed in our case 10. µmax is the Σ of each weight to multiplied by column total in our case (Table, 4 Due to the criterion weights being summed to one, the nal scores of the combined solution are expressed on the same scale (Feizizadeh and Blaschke 2013). In this case, the higher the factor weight. the more in uence on the nal landslide susceptibility map (Saaty, 1977). The LSZ map of upper Didessa sub basin is graded into ve classes as vary low, low, moderate, high and very high using natural break method of Jenks available in ArcGIS (Figure, 6). ArcMap identi es break points by picking the class breaks that best group similar values and maximize the differences between classes.

Analysis of output LHZ Map
The landslide distribution results show that 12% and 6% of the total land area are high and very high susceptibility to landslide ( Table 5). The high hazard zone is found surrounding the areas of moderate and encompass, whereas moderate hazard zones is occurring in the north western, central and north eastern parts of the study area major portions of low hazard zone and very low hazard zones exist in the east to NE and south west parts of the study area part of the study area have clearly indicated moderate to very high susceptibility zones (Figure, 7).

Conclusions
Landslide is one of the natural environmental disasters that affects environmental health socioeconomic development of all nations in spite of level of economic development. For this study, remote sensing data were obtained from Landsat TM image and DEM of 30 m resolution, which were used for landslides mapping, geomorphological, hydrological structure and land use change studies. The study evaluated and estimated the landslide hazard based on eleven landslide causative factor i.e., slope, aspect, drainage density, topographic wetness index (TWI), stream power index (SPI), topographic ruggedness index, hypsometric integral, lithology, LULC, soil texture, and distance from road. The research generated a series of landslide susceptibility maps using AHP method. Several causal factors were determined, including: rainfall, slope degree, geology, geomorphology, weathering crust, land cover, lineament density, drainage density, and elevation.

Declarations
Availability of data and materials: Included in the manuscript.