Landslide is used to describe a wide variety of processes that result in the detectable downward and outward movement of a slope material (soil, rock, and vegetation) under gravitational influence [1]. The materials may move by: falling, toppling, sliding, spreading, or flowing [2]. Landslides can be triggered by both natural and manmade changes in the environment. The main causes of landslide are classified in to two: 1) inherent causes of slope failure such as; weakness in the geological composition or geological structures within the rock or soil and 2) external triggering mechanisms such as; rainfall, earthquake or volcanic activity, snowmelt and change in ground water level and human activity [3] [4]. The study area located in central Ethiopia is frequently affected by landslides of various types, which often lead to eviction of inhabitants and damage to housing, infrastructures and arable land and loss of human lives [5] and [1] [4]. Most of the landslides in this region, including the largest ones, are triggered by heavy precipitation and springs due to groundwater level rise [6], [7]. 3 In addition, earthquake triggering slope failures, which resulted in rock slides, toppling and rock falls are reported by various workers (e.g. [5], [7], [8]. Landslide Susceptibility (LS) is defined as the likelihood of a landslide occurring in an area on the basis of local terrain conditions. LS zonation refers to the division of the land in homogeneous areas or domains according to the degree of actual or potential hazard. In the last years, LS mapping has become a valuable tool for assessing disaster phenomena. The cartographic products of LS modelling can be used for developing mitigation plans and selecting the more suitable locations for construction. Geographic Information Systems (GIS) technology has powerful ability to deal with a large amount of spatial data, LS is analysed through effective spatial analysis tools. Many GIS-based approaches have been developed for LS assessment [9]. Synthetic Aperture Radar (SAR) images contain both the intensity and phase information of the return signals from the earth surface. Whereas SAR Interferometry (InSAR) is a technique in which two SAR images of the same area at the surface of the earth taken from slightly different satellite positions are used to generate an Interferogram representing the phase difference between the return signals in two images [10]; [11]. Over the past 20 years, space-borne Interferometric Synthetic Aperture Radar (InSAR) has developed rapidly and has proven to be a valuable tool for topographic mapping and surface deformation measurements [11], [10], [12] and [13]. Because of its detail spatial coverage and competitive accuracy, InSAR has now become one of the most favored geodetic methods to study surface deformation associated with slow-moving landslides [14] and [15]. The operational space born SAR sensor that makes the availability of SAR data currently for professionals across industries in this study was Sentinel-1. This is due to the fact that Sentinel -1 is relatively good SAR image, which can be used for land deformation analysis including landslide mapping [16]. SARPROZ software is powerful software used to easily process and analyze SAR data and generate products like Digital Elevation Model (DEM) or surface deformation maps, which gives the option to integrate this information with other geospatial products. This unique data analysis 4 capability of SARPROZ converts the data to meaningful and contextual information (https://www.sarproz.com/). In this work, a robust geodetic method of SAR interferometry (InSAR), which has not been used in the Debresina area as well as GIS technique of higher spatial resolution were utilized. Therefore, landslide hazard and vulnerability of the lives, infrastructure and property of the community living in the area calls immediate attention to conduct detail investigation using the above mentioned techniques. The major objectives of this research were to (1), Process multi temporal SAR image using InSAR techniques for landslide mapping, (2) Investigate and delineate landslide susceptibility areas using GIS technique and (3) compare landslide resulted from InSAR and GIS techniques.

The study area is situated between 39° 29' 00'' to 39 ° 55' 00'' longitude and 9 ° 44' 40'' to 10 ° 10' 20'' latitude of Amhara regional State, central Ethiopia and has an area coverage of 1760.1 sq.km (Figure 1). Located at the western escarpment of the MER, Debresina area depicts a rugged topography shaped by easterly facing and moderately steeply sloping mountain chains, rift parallel normal faults and lineaments, and easterly draining deeply dissecting rivers [8]. It covers major landslide localities such as Mafud, Kewet, Lalomeder and Mamamidr and to some extent localities north of Ankobre Sub district. Closest cities including Debresina are Mezezo, Shewarobit, Sela dingay, Armanya, Tarmaber and Gudoberet

Figure 1: Location map of the project area.

Software and Instruments to be utilized

To undertake this study, the following software and instruments were utilized. These are SARPROZ - to easily process and analyze SAR data and generate landslide displacement map using InSAR (PS-InSAR); Arc GIS- to produce landslide map by taking controlling factors; Handheld GPS -to collect landslide inventory data for landslide validation.

Data Sources and Acquisition Techniques

In this study, all pertinent data were collected from primary and secondary sources (Table 1). By using direct and intensive field observations, GPS primary data using hand held GPS were collected whereas secondary data such as soil, DEM and Landsat image for GIS and Sentinel images for InSAR purposes were collected from various sources and are shown in Table1.

Table 1: Types of secondary Data and their Sources.

There are different methods of landslide mapping among which InSAR and GIS techniques are widely used ones [16] and [17]. In this study, InSAR and GIS techniques are employed to investigate deformation triggered by landslide. The complete methodology and flowchart of this study is presented in Figure 2, which depicted all data types and produced maps including software used to accomplish this investigation.

Figure 2: Flow chart showing the whole methodology pattern of the study.

InSAR method

SAR interferometry is used for the detection of Earth´s surface displacements through Differential Interferometry (DInSAR), which has shown to be a tool of great potential over the last decades. Initial single interferogram techniques, commonly referred to as conventional DInSAR has evolved to advanced DInSAR technique which provides information on the temporal evolution of the ground displacement, with a theoretical millimeter precision under favorable conditions. Persistent Scatters (PS-InSAR) is also one of the advanced DInSAR technique methods that work on localized targets. Such technique has been applied to ground displacements such as landslide. The basic concept of DInSAR technique is to monitor an area through time on a regular basis.

SAR images acquired in different dates are combined in pairs to generate a set of differential interferograms that contain information on the interferometric phase. Preferably, differential interferograms should contain only the ground displacement component between the acquisition times of the two SAR images. However, in practice, there are other terms contributing to the interferometric phase that can mask the desired ground displacement information, e.g. phase contributions from atmospheric water vapor. The goal of the different processing techniques is to accurately isolate the displacement term from the remaining components [18].

For this work, 23 Sentinel-1A data of single look complex (SLC) images in interferometry wide swath mode (IW) with VV polarization and descending direction for two acquisition dates of 05 June 2020 and 28 March 2015 were registered as master and slave, respectively, in SARPROZ software. SARPROZ is MATLAB-written software designed for Multi Temporal Synthetic Aperture Radar processing providing modules for SAR and InSAR data analysis. Licensed software which is developed by Dr. Daniele Perissin for running InSAR data was utilized. Although it supports most existing satellite formats and imaging modes, it has no focusing module. As a result, Single Look Complex (SLC) data are a must (Wossila.N, 2019).

GIS Method (Weighted Linear Combination)

Weighted linear combination (WLC) method is extensively used for landslide susceptibility and hazard mapping due to its application flexibility and expertise knowledge dependability [19]. WLC is commonly used among researchers and landslide specialists, but lacks clear refining steps to derive higher degree of accuracy probability maps. WLC is applied based on combining weighted averages of a number of parameters selected by the expert. There are no criteria set but purely depends on expert knowledge and the outcome can significantly vary among experts. Each parameter is classed and multiplied with its assigned weight and within a GIS overlay environment; the weighted averages are added to get the final output map [20]. [21] State that natural hazard prone areas, such as landslide can be mapped by assigning equal importance (weight) for all parameters.

Accordingly for this study, all factors that control the result were given equal importance.

In WLC, the weight of each parameter considered is added by means of overlay as shown below.

Where S is landslide susceptibility, Wi weight factor and Xi criterion score of factor. The different factors are combined by adding their weight to obtain the final output [20], [22] and [23].

Landslide-Influencing Factors

There are no standard guidelines to select important factors for landslide mapping. However, the nature of the study area, the scale of the analysis, the data availability and the general literature guidelines were taken into account [9]. The conditions for landslide to happen are slope, elevation, soil, NDVI, distance from road and stream. These factors used to determine the landslide susceptibility map using overlay analysis in GIS environment.

Slope

The slope angle is an important controlling factor for landslide hazard analysis. The probability of slope failure increases, as the slope becomes steeper and landslide frequency is expected to be minimal in gentle slopes [24]. The slope map of the study area was generated from ASTER data of 30m resolution and is done by producing DEM using GIS environment (Figure3 A).

Elevation

Elevation is considered to be important causative factor that affects the landslide susceptibility of a slope [25]. The elevation map of this work was produced from 30 m resolution ASTER DEM data set by reclassifying techniques (Figure 3B).

Soil type

It is one of landslide controlling factors to determine which type of soil and the tendency of soil particle to resist sliding across each other. The nature of the movement is, controlled by the earth materials involved. Soil type based on their color can affect landslide. Accordingly, eight type of soil were generated from 1:250,000 digital soil and these are chromic cambisols, chromic vertisols, eutric cambisols, eutric nitisols, eutric regosols, leptosols, pellic vertisols and vertic cambisols. Chromic vertisols and pellic vertisols depict black color; chromic cambisols, eutric cambisols, eutric regosols, leptosols and vertic cambisols show brown color whereas eutric nitisols has red color. The black soils, brown soils and red soils have 0.15, 0.20 and 0.25 erodibility k factors, respectively [26]; [27]; Figure 3C). These imply that in high landslide areas red colors which have 0.25 erodibility are highly affected by erosion whereas black colors with 0.15 erodibility are less affected by erosion obviously showing less landslide prone area. From this argument one can observe that erosion aggravates landslide; clearly indicating that greater erosion can be resulted in greater landslide.

Figure 3: landslide controlling factors.

Normalized difference vegetation Index (NDVI)

Vegetation cover plays an important role in reducing soil erosion. Extensive network of root system provides natural interlocking anchorage of the soil layer along slopes. Highly vegetated slope area generally reduces the effects of soil erosion along the slopes and this plays a major role in minimizing susceptibility of landslides and mass movements in the region whereas barren area with less or no vegetation is highly prone to erosional activities and then to various landslides [23]. In this work, NDVI map was produced from Landsat-8 image (OLI-TIRS sensor) with acquisition date of May 08, 2020 using ArcGIS environment particularly raster calculator (Figure 3D).

Distance from the road and stream

It is argued that the nearest distance to the roads and drainage network show the most susceptibility to landslide occurrence due to the natural slope disruption by roads construction and slope toe erosion by rivers [28]. For this research, proximity to road map (m) was generated by clipping from Ethiopian Road Network and proximity to stream map (m) was also generated according to the distance from stream feature which was produced from aster DEM data set at ArcGIS environment particularly using Euclidean distance tool (Figure 3E & 3F respectively).

Landslide Hazard Zonation

In this study, landslide susceptibility map was generated based on triggering factors such as slope, elevation, soil, NDVI, distance from steam and road features. Thus, spatial relationship between the occurrence of landslides and each landslide causative factor class was derived using overlay analysis (WLC) in GIS environment. The resulted Landslide map was categorized in to four susceptibility zones i.e. low, moderate, high and very high. Such classification supported with various research findings (e.g. Muhammad et al., 2016).According to the landslide result map (Figure 4), the study area is covered with 280.5 sq.km (15.9%), 754.2 sq.km (42.9%), 544.2 sq.km (30.9%) and 181.2 sq.km (10.3%) of low, moderate, high and very high susceptibility, respectively, which needs high attention to protect present and future loss of life, infrastructure and property).

Figure 4: Landslide susceptible zoning map.

Displacement Map (Interferogram)

The landslide deformations in Debresina area is analyzed by using PS-InSAR (advanced DInSAR) technique, which is a powerful technique to observe displacement in the study area especially in very high and high hazard zones depicted in the landslide hazard zonation map (Figure 5).

The analyzed deformation time-series revealed the presence of displacement from 9 up to -27 mm/year in the velocity of landslide movement along slopes. The PS-InSAR results depicted that the process of surface displacement in the area is still active. Unwrapped phase applied to get continuous deformation phase value. The temporal phase difference for each PS pixel are calculated and then unwrapped spatially from reference PS pixels using an iterative least square method. PS pixel is also filter using Goldstein phase filter before performing unwrapping (Figure 5A). Wrapped interferograms from descending orbit data are acquired over Debresina area landslide and the color represents 12 and -12mm of displacement, which can then be estimated using the phase values of the individual PS pixels (Figure 5B)

Figure 5: Interferogram (multi look of 20×20) landslide deformation map.

Coherence map is generated to clearly show the correlation between master and slave images, and it determines the quality of the resulted interferogram [29]. In this study, 0.3 coherence level was used to generate coherence map shown in Figure 6. Multi look process was applied for unwrapped, wrapped and coherence map to reduce the speckle appearance and to improve the image interpretability.

Figure 6: Coherence (multi look of 20×20) map of the study area.

Validation Landslide Susceptibility Map

The produced landslide map was validated using Relative landslide density index (R-index, shown in equation 2) method, which helped to assess the relationship between the landslide susceptibility and landslide inventory map (GCP). The sample points were collected in field investigation using handheld GPS. This was carried out by overlay GCP points on susceptible map.in this techniques susceptibility zones correlated with inventory data.

(Where ni and Ni are the number of landslides occurrences and the number of pixels respectively in the i sensitivity class [30].

Among 11 GCP points that showed landslide occurrences, 10(n) points coincide on very high and high susceptibility zones(N=805,993 pixel numbers),and 1 point overlapped on low susceptibility (=311652 number of pixels).based on R index formula validation value gives 79.45% [((10/805,993)/((10/805,993) +(1/311,652))] *100.

In addition, these points also overlapped on Interferogram which depicted landslide occurrences in the study area. Based on these facts, landslide susceptibility could be mapped using InSAR and GIS techniques, but for better landslide mapping, InSAR is more powerful than GIS. Even though it is not very powerful to quantify the amount of landslide displacement, GIS technique can preferably be utilized for landslide zonation. That is why slide susceptibility mapping at GIS environment can be validated by InSAR methods [18]. [8] Revealed that landslide in Debresina area were very large and affected human activity starting 13 September 2005. In addition, it is debated that slopes of the Ethiopian Highlands are frequently affected by landslides of various types, which often lead to eviction of inhabitants, damage to housing, infrastructure and arable land and even loss of human lives [5]. Thus, research results of these authors and this work clearly show that they are in a good agreement.

Figure 7: validation of InSAR and GIS result.

Landslide susceptibility mapping provides fundamental information for hazard assessment and monitoring strategies. In mountainous area like Deberesina, direct ground based landslide triggering factor mapping and evaluation is expensive and virtually impossible within a short period of time. To resolve such limitations, InSAR and GIS techniques provide powerful alternatives for detecting, identifying and monitoring landslides and related triggering factors.

Six control factors (slope, elevation, soil, NDVI, distance from stream feature and road feature) were used to map landslide zoning. Aster DEM data was used to produce slope, elevation and proximity map from stream and Landsat 8 OLI image was used to generate NDVI map. Eth_trs_roads_osm website and digital soil data from the Minister of Water and Resource (MoWR) were used to get proximity from road and soil map, respectively. Consequently, vector data were converted to raster data (30 m resolution).

All causative factors overlaid produced landslide susceptibility zone by weighting linear combination techniques at GIS environment. Nonetheless, the weight of each factors were given according to literatures. Each factors were classified in to 9 class by natural breaks (Jenks) classify method. Among the classified classes, highly slopes and elevations, red color soils which show high erosion, less vegetation (bare lands), very close proximity to the road and stream were the most susceptible to landslide. The final landslide susceptibility map clearly depicts 4 classes of very high, high, moderate and low landslide vulnerable areas. Since Debresina area is characterized by steep slope and high elevation, approximate 41 % of the study area is strongly affected with landslide geohazard.

PS-InSAR (advanced D –InSAR) is also the most important techniques for landslide displacement mapping.in this project. SARPROZ software (which is installed by permission) is utilized for analyses of SAR images. As a result, -9 mm /yr. to 27 mm/yr. clearly showing significant displacement was observed.

The resulting landslide map was validated using R index method which is simple, and this was carried out using GPS points as inventory data. Unfortunately most of selected sample points were put on places where very high or highly prone to landslide. In most case, researchers argue that InSAR and GIS could be used for landslide mapping together or separately. From this work, we strongly suggest that for real/quantifying mapping, InSAR method is more robust whereas for simple susceptibility zoning, GIS is dependable technique.

We greatly thank to Ethiopian space science and technology institute for financing this work and Deresina area municipality offices for their helping during reconnaissance and taking sample GCP points.

No conflict between authors

1. Woldearegay K. Review of the occurrences and influencing factors of landslides in the highlands of Ethiopia: With implications for infrastructural development. Momona Ethiopian Journal of Science (MEJS). 2013; 5: 3-31.
2. Kanungo DP. Arora MK. Sarkar S. Gupta RP. Landslide Susceptibility Zonation (LSZ) Mapping – a review. Journal of south Asia disaster studies. 2009; 2: 81-105.
3. Varnes DJ. International Association of Engineering Geology Commission on Landslides and Other Mass Movements on Slopes: Landslide hazard zonation: a review of principles and practice. UNESCO, Paris. 1984; 63.
4. Meten M. Bhandary NP. Yatebe R. GIS-based frequency and logistic regression modelling for landslide susceptibility mapping at Debresina area. Journal of Mountain Science. 2015; 12: 1355-1372.
5. Abebe B. Dramis F. Fubelli G. Umer M. Asrat A. Landslides in the Ethiopian highlands and the Rift margins. J. Afr. Earth Sci. 2010; 56: 131-138.
6. Ayalew L. The effect of seasonal rainfall on landslides in the highlands of Ethiopia. Bull. Eng. Geol. Environ. 1999; 58: 9-19.
7. Abay A. Barbieri G. Landslide Susceptibility and Causative Factors Evaluation of the Landslide Area of Debresina, in the Southwestern Afar Escarpment, Ethiopia. J. Earth Sci Eng. 2012; 2: 133-144.
8. Kropácek J. Varilová Z. Baro I. Bhattacharya A. Eberle J. Hochschild V. Remote Sensing for Characterisation and Kinematic Analysis of Large Slope Failures: Debre Sina Landslide, Main Ethiopian Rift Escarpment. Remote Sens. 2015; 7: 16183–16203.
9. Chalkias C, Ferentinou M, Polykretis C. Karymbalis E. Integrating Expert Knowledge with Statistical Analysis for Landslide Susceptibility Assessment at Regional Scale. 2016; 6: 1-14.
10. Hanssen RF. Radar Interferometry: Data interpretation and error analysis, Kluwer Academic Publishers, Dordrecht. 2001; 2: 308.
11. Massonnet D. Feigl KL. Radar interferometry and its application to changes in the Earth's surface. Rev. Geophys. 1998; 36: 441-500.
12. Singh LP. Application of SAR Interferometry in land slide studies with spatial reference to generation of input data for statistical susceptibility assessment. 2003.
13. Sudhaus H. Jonsson S. Improved source modelling through combined use of InSAR and GPS under consideration of correlated data errors: application to the June 2000 Kleifarvatn earthquake, Iceland Geophys. J. Int. 2009; 176: 389-404.
14. Strozzi T. Farina P. Corsini A. Ambrosi C. Thüring M. Zilger J. et.al. Survey and monitoring of landslide displacements by means of L-band satellite SAR interferometry. Landslides. 2005; 2: 193-201.
15. Tantianuparp P. Shi X. Zhang L. Balz T. Liao M. Characterization of landslide deformations in three gorges area using multiple InSAR data stacks. Remote Sens. 2013; 5: 2704-2719.
16. Kyriou A. Nikolakopoulos K. Assessing the suitability of Sentinel-1 data for landslide mapping, European Journal of Remote Sensing. 2018; 51: 402-411.
17. Vöge M. Frauenfelder R. Ekseth K. Arora MK. Bhattacharya A. Bhasin RK. The use of SAR interferometry for landslide mapping in the Indian Himalayas. 2015.
18. K?ncal C. Li Z. Drummond J. Liu P. Hoey T. Muller JP. Landslide susceptibility mapping using GIS-based vector grid file (VGF) validating with insar techniques: three gorges, Yangtze River (china). 2017; 3: 116-141.
19. Tomás R. Romero R. Mulas J. Marturià JJ. Mallorquí JJ. Lopez-Sanchez JM. Et.al. Radar interferometry techniques for the study of ground subsidence phenomena: a review of practical issues through cases in Spain. 2014; 71: 163-181.
20. Alemayehu, L. Gerra S. Zvelebil J. Šíma J. Landslide Investigations in Tarmaber, Debre Sina, North Shewa Zone; Amhara Regional State; AQUATEST a.s.: Prague, Czech Republic. 2012.
21. Michael EA. Samanta S. Landslide vulnerability mapping (LVM) using weighted linear combination (WLC) model through remote sensing and GIS techniques. 2016; 88.
22. Karlsson CSJ. Kalantari Z. Mörtberg U. Olofsson B. Lyon SW. Natural Hazard Susceptibility Assessment for Road Planning Using Spatial Multi-Criteria Analysis. 2016; 60: 823-851.
23. Feizizadeh B. Blaschke T. GIS-multicriteria decision analysis for landslide susceptibility mapping: comparing three methods for the Urmia lake basin, Iran. 2013; 65: 2105-2128.
24. He H. Hu D. Sun Q. Zhu L. Liu Y. A Landslide Susceptibility Assessment Method Based on GIS Technology and an AHP-Weighted Information Content Method: A Case Study of Southern Anhui, China. 2019; 8: 266.
25. Demissie S. Landslide hazard evaluation and zonation in Dilbe and its surrounding area Northwestern central Ethiopia; an integrated GIS based grid overlay and statistical approach. 2018.
26. Ahmed S. Slope Stability Analysis Using GIS and Numerical Modeling Techniques.Unpublished MSc Thesis. VrijeUniversiteit, Brussel, 2009; 120.
27. Hurni, H. Soil conservation manual for Ethiopia. Addis Ababa: Ministry of Agriculture. 1985.
28. Helden U. An assessment of woody biomass, community forests, land use and soil erosion in Ethiopia. Lund: Lund University Press. 1987; 75.
29. Vakhshoori V. Zare M. Landslide susceptibility mapping by comparing weight of evidence, fuzzy logic, and frequency ratio methods. 2016; 1731-1752.
30. Hammad M. Mucsi L. Leeuwen BV. Landslide Investigation Using Differential Synthetic Aperture Radar Interferometry: A Case Study of Balloran Dam Area in Syria. 2019; 133-138.
31. Shahabi H. Hashim M. Landslide susceptibility mapping using GIS-based statistical models and Remote sensing data in tropical environment. 2015; 9899.
32. Carrara A. Cardinali M. Guzzetti F. Reichenbach P. 1995. GIS technology in mapping landslide hazard. Geographical information systems in assessing natural hazards. Kluwer Academic Publishers, Dordrecht, the Netherlands: pp 150-180. Earth's surface. Reviews of Geophysics36, 441-500.
33. Basharat M. Shah HR. Hameed N. Landslide susceptibility mapping using GIS and weighted overlay method: a case study from NW Himalayas, Pakistan. 2016; 9.
34. Svigkas N. Papoutsis I. Loupasakis C. Tsangaratos P. Kiratzi A. Charalampos HK. InSAR time-series monitoring of ground displacement trends in an industrial area (Oreokastro—Thessaloniki, Greece): detection of natural surface rebound and new tectonic insights. 2017; 76: 1-18.
35. Schneider J. Woldearegay K. Atsbah G. Reactivated large-scale landslides in Tarmaber district, central Ethiopian Highlands at the western rim of afar triangle. In Proceedings of the International Geological Congress, Oslo, Sweeden, 2008; 6-14.
36. Van Westen, CJ. GIS in landslide hazard zonation: A review, with examples from the Andes of Colombia, In: M. F. Price & D. I. Heywood (eds), Mountain Environments and Geographic Information Systems, Taylor and Francis Publishers. 1994; 135-165.
37. Woldearegay K. Characteristics of a large-scale landslide triggered by heavy rainfall in Tarmaber area, central highlands of Ethiopia. Geophys. Res. Abstr. 2008; 10.
38. Woldearegay K. Review of the occurrences and influencing factors of landslides in the highlands of Ethiopia: With implications for infrastructural development. Momona Ethiop. J. Sci. 2013; 5: 3-31.
39. Nassir W. Structural Deformation Monitoring in Urban Area (Addis Ababa city) Using INSAR. 2019.