Main Ethiopian Rift landslides formed in contrasting geological 1 settings and climatic conditions 2 3

The Main Ethiopian Rift (MER), where active continental rifting creates specific conditions for landslide 16 formation, provides a prospective area to study the influence of tectonics, lithology, geomorphology, and climate on 17 landslide formation. New structural and morphotectonic data from CMER and SMER support a model of 18 progressive change in the regional extension from NW – SE to the recent E(ENE) – W(WSW) direction driven by 19 the African and Somalian plates moving apart with the presumed contribution of the NNE(NE) – SSW(SW) 20 extension controlled by the Arabic Plate. The formation and polyphase reactivation of faults in the changing regional 21 stress-field significantly increase the rocks’ tectonic anisotropy and the risk of slope instabilities forming. 22 According to geostatistical analysis landslides in the central and southern MER occur on steep slopes, almost 23 exclusively formed on active normal fault escarpments. Landslides are also influenced by higher annual 24 precipitation, precipitation seasonality, vegetation density and seasonality. 25 A detailed study on active rift escarpment in the Arba Minch area revealed similar affinities as in regional study of 26 MER. Landslides here are closely associated with steep, mostly faulted, slopes and a higher density of vegetation. 27 Active tectonics and seismicity are the main triggers. The Mejo area situated on the uplifting Ethiopian Plateau 60 28 km east of the Rift Valley shows that landslide occurrence is strongly influenced by steep erosional slopes and 29 deeply weathered Proterozoic metamorphic basement. Rapid headward erosion, unfavourable lithological conditions 30 and more intense precipitation and higher precipitation seasonality are the main triggers here. 31 32


144
In addition to the seismic tremors, volcanism is also of apparent risk. Among the recent events are the Nabro

163
The precipitation data were obtained from the national database that was set up by the Centre for Development and The main approach for the morphotectonic analysis followed that used by Dhont and Chorowicz (2006 and 180 references therein). The main aim was to use DEM imagery to interpret the largest neotectonic structures in the 181 central and southern MER regions. Single-directional and multi-directional shaded reliefs and an elevation coloured 182 ASTER DEM image (Fig. 3) was generated using ArcMap 10.6 (www.esri.com). This DEM constitutes the basis for 183 morphotectonic analysis at the regional scale. The faults mapped can be considered as the main neotectonic faults 184 because they have a prominent expression in the morphology. In some cases they form asymmetric ranges with one 185 side corresponding to breaks in slope or scarps; by the displacement of Pleistocene and Neogene lithological 186 boundaries; by the occurrence of straight lines of kilometres to several tens of kilometres in length. The images were 187 compared with field geological mapping data to distinguish the scarps formed by active faults from those formed by 188 differential erosion of contrasted lithology.

189
The emplacement of volcanoes, which are abundant in study area, can also be related to tectonic structures such as

199
In addition to a visual interpretation of linear indices, a quantitative technique -morphometry -was also employed 200 to analyse landforms in a quantitative manner. This technique uses numerical parameters such as slope, surface 201 curvature and convexity to extract morphological and hydrological objects (e.g., stream networks, landforms) from 202 DEM (Fisher et al., 2004;Pike, 2000;Wood, 1996). Landforms and lithological units reflect also different 203 geotechnical properties (e.g. rock strength, degree of weathering) so they can be identified by these numerical minimum convexity values were derived at a pixel by pixel basis. The variation in these parameters was quantified 209 for each pixel with respect to neighbouring pixels (in orthogonal directions). Secondly, based on a set of tolerance 210 rules, morphometric classes were defined for each pixel: ridge, channel, plane, peak, pit and pass (Wood, 1996).

211
Wood's algorithm allows the relief to be parametrized by setting different values for the tolerance of the topographic 212 slope and convexity. In this study the slope tolerance of 3.0 and convexity tolerance of 0.02 were used for the best 213 fit.   (Fig. 9) 238 and Arba Minch (Fig. 11) detailed study areas are also shown (see section 4.5).

240
A combination of a visual morphotectonic interpretation based on DEMs (Fig. 2) and an interpretation on 241 morphometric landforms (Fig. 3) was used to map lineaments. The study area is characterised by a predominance of    The geomorphology is highly variable across the MER and is mainly the result of volcanic and tectonic events with 296 the associated erosional and depositional processes (Billi, 2015). The principal feature of the MER is the graben 297 bounded by normal faults. The drainage network is largely controlled by tectonic activity and lithological variation.

298
Parts of grabens form endorheic depressions filled by temporal lakes. The area is climatically highly variable; the 299 average amounts of annual rainfall vary from 500 in the Gibe and Omo Gorges to 2,600 mm on the escarpments and 300 the adjacent highlands. The mean annual temperature is about 20°C. High RMS = 6, Medium RMS = 5, Low RMS = 4, Very Low RMS = 3, Soils = 2, Lacustrine deposits = 1. A 332 significant correlation between RMS and slope and most precipitation parameters was found (see Table 1). More    Table 2). Proximity to tectonic features is expressed in terms of the percentage area of a particular 354 geohazard within a particular buffer zone (500 m and 1 km buffer).

361
Most landslides and rockfalls form on steeper slopes close to faults and in areas with higher lineament density.

362
Rockfalls are formed on steeper slopes than landslides (Table 2) but slope factor has higher importance for the 363 formation of landslides (in comparison to other factors, see Fig. 8). Rockfalls typically occur on areas receiving 364 lower precipitation. Most of them occupy areas with grassland and, to a lesser extent, also on cultivated land and 365 bush land cover. Higher vegetation seasonality is also found to coincide well with rockfall occurrences. A low, very 366 low and high rock mass strength class probably influence the occurrence of rockfalls (see Table 2 and Fig. 8). While   difference between the average wet (July + August) and dry season (December + January) is 310 mm (CDE, 1999).

396
Vegetation cover is dense (NDVI values almost double comparing the Arba Minch area) and moderately seasonal 397 (see Table 3). Due to intense weathering the area is dominated by rocks with low and medium mass strengths. The     Table 2 was used for most parameters, but faults and lineaments data were adopted from more detailed studies at a study area (see Fig. 10 and Table 3). They are also formed in areas with a higher vegetation density and medium and

453
The Arba Minch study area is located directly in the main rift valley on the western normal fault escarpment. The

454
total displacement of the syn-and post-rift normal faults is more than 1500 metres. The average slope in the area is 455 less than 10 degrees because a large part of the area is covered by Abaya Lake (see Fig. 11   half of Arba Minch area) and moderately seasonal (see Table 3). Rocks with low and medium mass strengths and

492
The mean values of the same factors as for the Mejo site were also calculated for each landslide and rockfall 493 polygon area at the Arba Minch site. Here the landslides and rockfalls are situated in areas with much higher slopes,

494
compared to the overall study area (see Fig. 10 and Table 3), there is a much higher density of faults and lineaments 495 close to faults. They are also formed in areas with much higher vegetation density and medium and low RMS.

496
Landslide and rockfall areas are also dominantly covered by cultivated areas with woodlands taking a minor role.  (Table 2), is more important for the formation of rockfalls than landslides. Rockfalls also 514 show a much higher affinity to the proximity of faults.

515
Rockfalls occur in areas with lower precipitation, while for landslides high precipitation and high precipitation 516 seasonality is typical. It correlates well with high vegetation density and low vegetation seasonality, which are found 517 to have strong affinity with landslide occurrences. Thus, precipitation does not seem to be an important factor for 518 rockfall formation but is important for landslides.

519
Rockfalls and landslides occur on areas with bushland, grassland and cultivated landcover. It leaves deforestation as 520 one of the possible triggering factors. They also occur in areas with a wide range of rock mass strength classes (very 521 low, low, medium and high) so lithology and intensity of weathering do not seem to be an important triggering 522 factor.

523
In the large area of the MER the vast majority of slope instabilities is located on active normal fault escarpments 524 (Fig. 12). This is a major natural triggering factor for rockfalls. While for landslides there is also the important 525 influence of higher precipitation, precipitation seasonality and vegetation density and seasonality. fracturing along faults, these zones are often altered, which lowers the slope stability of the rock environment.

540
Alteration is also enhanced by more intense water-rock interactionsmost springs are located on fault zones (Arba

541
Minch means "Forty Springs"). Precipitation was not confirmed as an important factor.

542
The Arba Minch area is seismically active, according to the catalogue of earthquakes of the United States

604
Competing interests. The authors declare that they have no conflict of interest.