Catena 97 (2012) 127–136

Contents lists available at SciVerse ScienceDirect

Catena

j ourna l homepage: www.e lsev ie r .com/ locate /catena

Spatial patterns, causes and consequences of landslides in the Gilgel Gibe catchment,SW Ethiopia

N. Broothaerts a,⁎, E. Kissi b, J. Poesen a, A. Van Rompaey a, K. Getahun c, E. Van Ranst d, J. Diels a

a Department of Earth and Environmental Sciences, K.U. Leuven, Celestijnenlaan 200E, 3001 Heverlee, Belgiumb Department of Natural Resources Management, Jimma University, Jimma, Ethiopiac Department of Geography and Environmental Studies, Jimma University, Jimma, Ethiopiad Department of Geology and Soil Science (WE13), Ghent University, Krijgslaan 281/S8, 9000 Gent, Belgium

⁎ Corresponding author. Tel.: +32 16 327800; fax: +E-mail address: nils.broothaerts@ees.kuleuven.be (N

0341-8162/$ – see front matter © 2012 Elsevier B.V. Alldoi:10.1016/j.catena.2012.05.011

a b s t r a c t

a r t i c l e i n f o

Article history:Received 25 November 2011Received in revised form 15 May 2012Accepted 30 May 2012Available online xxxx

Keywords:LandslidesLandslide inventoryCausal factorsSoil lossEast AfricaEthiopia

The Gilgel Gibe catchment in SW Ethiopia is one of the areas in East Africa affected by landslides. To betterunderstand the patterns and the causal factors of these landslides, all landslides in a small study area(14 km²) in the hilly parts of the Gilgel Gibe catchment were mapped and analyzed. In total, 60 landslideswere mapped. These landslides caused a displacement of 1 million m³ slope material, which correspondsto a mean displaced volume of 50 ton ha−1 y−1 in the last 20 years. Moreover many landslides deliver direct-ly sediment to the rivers and hence increase the sediment load in the rivers. This soil loss to the rivers wasestimated at 11 ton ha−1 y−1 during the same period.High annual rainfall (ca. 2000 mm y−1), lithological and pedological properties and to a lesser extent steep(>16°) slopes turn the area into an inherent unstable situation and can be indicated as preconditions forthe landslides in the study area. Distance to rivers is significantly the most important precondition, as slopesnear rivers are less stable than slopes further away from the rivers. This is mainly caused by river incision andbank erosion which often occur in the area and which can be attributed to increased runoff due to defores-tation over the past 20 years. Therefore recent deforestation caused more shallow landslides but also indi-rectly more deep-seated landslides close to the rivers. Heavy rainfall is indicated as the main triggeringfactor for almost all landslides.

© 2012 Elsevier B.V. All rights reserved.

1. Introduction

During the second half of the 20th century the number of damag-ing landslides have been substantially increasing worldwide, as wellas the number of studies on landslides (e.g. Gokceoglu and Sezer,2009; Gutiérrez et al., 2010). However, research on landslides inEast Africa is still rather restricted, although it is a region wherelandsliding is a widespread phenomenon. Steep slopes, high annualrainfall, increasing population pressure and deforestation, earth-quakes and extreme rainfall make most areas in East Africa very sen-sitive to landslides (Glade and Crozier, 2005). Moreover, most of thelandslides have a significant economic, social and geomorphologicimpact (e.g. Ngecu and Mathu, 1999). There are still regions in EastAfrica where quantitative data of landslides are missing, althoughthese data are necessary to undertake remedial measures and torestrict the negative consequences for the inhabitants.

In East Africa, landslides have been reported in Uganda (e.g. Kitutuet al., 2009; Knapen, et al., 2006; Ngecu et al., 2004), Kenya (e.g.

32 16 322980.. Broothaerts).

rights reserved.

Davies, 1996; Ngecu and Ichang'i, 1999; Ngecu and Mathu, 1999;Ngecu et al., 2004; Westerberg and Christiansson, 1999), Tanzania(e.g. Lundgren, 1978; Marwa and Kimaro, 2005; Temple andRapp, 1972; Westerberg and Christiansson, 1999), Rwanda (e.g.Moeyersons, 1989, 2003), D.R.Congo (e.g. Moeyersons et al., 2004,2010), and Ethiopia (e.g. Abebe et al., 2010; Ayalew, 1999; Ayalewand Yamagishi, 2004; Ayenew and Barbieri, 2005; Coltorti et al.,2009; Fubelli, et al., 2008; Moeyersons, et al., 2008; Nyssen et al.,2002, 2004, 2006; Temesgen, et al., 2001; Van Den Eeckhaut et al.,2009; Vilimek et al., 2010; Woldearegay et al., 2005). All these land-slides are caused by various causal factors, however some factorsare common for most of the landslide sensitive areas. These factorscan be divided in preconditions, inherent static factors that act as cat-alysts to allow other factors to act more effectively; preparatory fac-tors, which make the slope susceptible to movement withoutactually initializing it; and triggering factors, which finally initiatethe movement (Glade and Crozier, 2005). For East Africa, the mostimportant preconditions for landslides are steep slopes, deep weath-ered soils with high clay content and high annual rainfall. The mostimportant preparatory factors reported for landslides in East Africaare all kinds of human activities such as deforestation and otherland use changes but also constructions and excavations. Finally,

Fig. 2. Example of a recent, shallow landslide in the Gilgel Gibe catchment. Landslidearea is ca. 700 m², landslide volume is ca. 1400 m³. (August 24, 2009).

128 N. Broothaerts et al. / Catena 97 (2012) 127–136

most important triggering factors are earthquakes and extremerainfall.

The Gigel Gibe catchment (Fig. 1), located in the South-West ofEthiopia, is one of the regions in Ethiopia that are heavily affected bylandslides (Fig. 2). Landslides in this region have been reported byAyalew (1999) and Abebe et al. (2010). These landslides cause manysocial, economic and geomorphologic problems. One of the problemsis the filling up with sediments of the Gilgel Gibe reservoir, con-structed for the Gilgel Gibe I hydro-electric power plant (Devi et al.,2008), for which the landslides may be an important sediment source.However, quantitative data on the landslides in the Gilgel Gibe catch-ment are absent as well as clear insight into the local causes for thelandslides. Hence, more research about these landslides is needed toundertake specific measures against the landslides and to restricttheir consequences for the Gilgel Gibe reservoir and on the socio-economic level in general.

Therefore this study investigate the spatial and temporal patterns,causes and consequences of landslides in the Gilgel Gibe catchment,and assesses the contribution of the landslides to the high sedimentload in the rivers of the Gilgel Gibe catchment.

2. Study area

Landslides are reported in all hilly areas of the Gilgel Gibe catch-ment (Abebe et al., 2010; Ayalew, 1999). This research concentrateson a study area of 14 km², located in the upper parts of the GilgelGibe catchment (Fig. 1). The study area has altitudes ranging between2000 and 2800 m.a.s.l., and is bounded by latitude 7°29′30′′–7°26′00′′N and longitude 36°51′30′′–36°53′00′′ E. According to the geologicalmap of Ethiopia (Tadesse et al., 2003), the study area consists of volca-nic rocks of the Eocene and Paleocene, rhyolites, trachytes, rhyoliticand trachytic tuffs, ignimbrites agglomerates and basalts. However,the spatial occurrence of the different geological materials isvery complex and heterogeneous and not known in detail. Themajor soil types in the study area are Nitisols, Acrisols and Vertisols

Fig. 1. Location of the study area in Ethio

(FAO-Unesco, 1974). Mean annual rainfall in the study area is ca.2000 mm (National Meteorological Agency of Ethiopia, 2009). Around60% of the rainfall occurs in the rainy season, lasting from Mayuntil September. Mean annual air temperature is ca. 19 °C (NationalMeteorological Agency of Ethiopia, 2009).

In the study area, almost all the land has been taken into cultiva-tion. The main land use type in the study area is agricultural cropping,mainly wheat, teff, barley, faba bean, sorghum and maize. Next tothese cropping activities, the farmers keep certain plots as grazingland for their livestock. The plots are mostly small and enclosed byhedges or tree rows. Total forest cover in the study area is ca. 11%,however this is mainly bush vegetation which is extensively used

pia and in the Gilgel Gibe catchment.

129N. Broothaerts et al. / Catena 97 (2012) 127–136

by the farmers. Only the highest parts of the study area are stillcovered by forest.

3. Materials and methods

3.1. Data collection

In order to get more insight in to the spatial pattern of the land-slides in the study area, an intensive field survey was made duringAugust and September 2009, at the end of the rainy season. Thefocus was on the recent landslides. They included active landslides,which are currently moving (which include first-time movementsand reactivations), and landslides which were not moving at the mo-ment of investigating (inactive landslides) but showed clear signs ofmovements within the last 20 years. In addition to these recent land-slides, also ancient landslides could be observed in the field. Theyinclude all landslides that have not been active for a very long time(more than 20 years), but are still visible in the landscape (Crudenand Varnes, 1996). Although these ancient landslides were still visiblein the landscape, it was very difficult to collect reliable quantitativedata about these landslides because erosion processes and ploughingmade them less clear and the local farmers could not give informationabout these landslides during the interviews. Therefore, the ancientlandslides were not included in the analysis.

All present landslides were mapped with GPS (Garmin eTrex Leg-end HCx). Landslides were classified during the fieldwork according

Fig. 3. Spatial pattern of landslides (n=60) in the study area. Each dot refers to the centrectangle indicates location of Fig. 11 B.

to Varnes (1978), based on their visible morphological characteristics.The area of the landslides was obtained by GPS, the volume of dis-placed hillslope material was estimated as the product of the areaand the estimated mean depth of the surface of rupture of the land-slide. A relative volume–area scaling for the landslides in the studyarea was assessed, based on:

V ¼ αAγ ð1Þ

where V is the volume of a landslide, A is the area of a landslide, γ isthe scaling exponent and α the intercept (Larsen et al., 2010). Themean soil loss (ton ha−1 y−1) due to landslides in the study areawas based on estimations of the volume of landslide material thatwas eroded by the rivers. To illustrate the impact of the landslides inthe study area on the river pattern, the gravel bars in one river section(Fig. 3) were mapped during the field survey, using GPS. A frequency–area distribution, which indicates the number of landslides thatoccur at different size-intervals, was constructed for the landslides inthe study area. The frequency density, f(AL), was derived as:

f ALð Þ ¼ δNL=δAL ð2Þ

where δNL/δAL is the number of landslides with areas between AL andAL+δAL (Van Den Eeckhaut et al., 2007).

Based on field investigations and interviews with local farmers,the year of first activation of the landslides - which allows us to

re of a landslide recorded during the field survey in the rainy season of 2009. Dotted

130 N. Broothaerts et al. / Catena 97 (2012) 127–136

discuss the temporal pattern of the landslides - and the socio-economic consequences were determined.

3.2. Statistical analysis

A digital elevation model (DEM) of the study area was created bydigitising the contour lines of the 1:50 000-scale topographic map(Ethiopian Mapping Agency, 1980). The contour lines have a verticalinterval of 20 m. The created DEM has a spatial resolution of 33 m. Inorder to assess the importance of the topographic factors (distance torivers, altitude a.s.l., planimetric curvature, aspect and slope gradient)as precondition for landslides in the study area, the values of thesefactors for slopes affected by landslides were compared with thevalues for the whole study area. Both values were deduced from theDEM of the study area. Based on the differences between these values,the importance of the topographic factors as causal factors for land-slides was determined. Chi-square tests were used to test whetherthe differences are significant or not. Cramer's V tests were calculatedto test the strength of the association.

4. Results and discussion

4.1. Spatial distribution

Sixty recent landslides were mapped and characterized in thestudy area of 14 km² (Fig. 3) Most of the recent landslides are locatednear rivers and at the lowest points in the landscape. Next to therecent landslides, also 23 ancient landslides were mapped in thestudy area, although the database of the latter is not complete as itwas difficult to distinguish them in the field and to collect reliabledata. Therefore, the ancient landslides will not be included in thefurther analyses.

4.2. Landslide characteristics

All recent landslides were classified according to the classificationof Varnes (1978) (Fig. 4). 75% of the landslides occurred in debrismaterial - a mixture of fine earth and rock fragments, resulting fromweathering processes and ancient landslides. Most of the landslidesin the study area are debris slides, debris falls and debris slumps.Together with the complex landslides they represent more than 75%of all mapped landslides. These landslide types are also reported inother areas in the Ethiopian Highlands, although rock fall, which hasalso been reported to occur in the Ethiopian Highlands, does not

Fig. 4. Classification of the mapped landslides (n=60) occurring in the study area,according to the classification of Varnes (1978).

occur in this study area (e.g. Abebe et al., 2010; Ayalew, 1999;Ayalew and Yamagishi, 2004; Ayenew and Barbieri, 2005; Fubelli etal., 2008; Nyssen et al., 2002, 2006; Van Den Eeckhaut et al., 2009;Woldearegay et al., 2005).

The mean area of all landslides was around 5 000 m², with a stan-dard deviation of 11 000 m² (Fig. 5). In total, the mapped landslidesoccupy ca. 0.275 km², which is almost 2% of the study area(14 km²). Because it is expected that not all recent landslides that oc-curred in the study area are still visible up to now, this figure is an un-derestimation of the total area occupied by landslides occurring in thelast 20 years. The volume of hillslope material displaced by the indi-vidual landslides varies widely in the study area and is summarizedin Fig. 5. The smallest landslide displaced only 20 m³ of material, thelargest around 250 000 m³. The average is 19 000 m³, with a standarddeviation of 42 000 m³. Together, the 60 landslides in the study areaof 14 km² displaced about 1 million m³ of slope material, which isfor the same reason as above an underestimation of the real cumula-tive displaced volume of slope material of the last 20 years. Based onEq. (1), the relative volume–area scaling for the landslides in thestudy area was assessed as V=1.9 *A1.1 (R²=0.9; N=51) (Fig. 5).This compares well with the exponent γ between 1.1 and 1.3 foundby Larsen et al. (2010) for the shallow soil-based landslides. Indeed,all mapped landslides were soil based landslides according to the sub-division of Larsen et al. (2010) (Fig. 4).

The frequency–area distribution for the landslides in the studyarea is shown in Fig. 6 together with the theoretical curves proposedby Malamud et al. (2004). These theoretical curves are based onthree complete landslide inventories which were all approximatedby the same probability density function (Malamud et al., 2004).The frequency–area distribution of the recent landslides in thestudy area is best described by a negative power–law relation withan exponent of −1.02. This distribution deviates from the theoreticalcurves proposed by Malamud et al. (2004) and from the distributionfound by different authors for landslides in mountainous regionswhich found a positive power–law relation for landslides of thesame magnitude (for an overview, see Van Den Eeckhaut et al.,2007). Following Van Den Eeckhaut et al. (2007), this negativepower–law relation suggests that the recent landslides in the studyarea are mainly caused by anthropogenic factors which have animpact at a local scale. Such factors, acting at a local scale, decreasethe stability of only a limited part of hillslopes. Therefore, smallerlandslides are more likely to occur and the number of landslidesdecreases with increasing landslide area, following a negativepower–law relation. This is in contrast with landslides caused bynatural factors which have a regional impact such as heavy rainfall,rapid snowmelt or earthquakes, causing a positive power–law rela-tion for small landslides.

Fig. 5. Volume hillslope material displaced by the mapped landslides, versus landslidearea (n=51).

Fig. 6. Landslide frequency density as a function of landslide area for 55 mapped recentlandslides; and the general landslide probability distributions proposed by Malamud etal. (2004) for landslide magnitude (ml) 1 to 5.

131N. Broothaerts et al. / Catena 97 (2012) 127–136

4.3. Temporal distribution

The year of first activation of the recent landslides was determinedbased on field observations and interviews with local farmers. Land-slides are grouped per decade due to uncertainties in the data provid-ed by the local farmers. For the landslides which were first activatedin the last 10 years, more reliable data could be collected (Fig. 7).Most of the recent landslides remain active up to now, with differentphases of activity and phases of inactivity within one rainy season. Nosignificant relation could be found between the number of activatedlandslides in a particular year and the annual rainfall, the occurrenceof rainfall peaks in a year or El Niño years.

Some authors (e.g. Ayalew, 1999; Davies, 1996) indicate an in-crease in landslide frequency in East Africa during the last decades.Based on our data (Fig. 7) this could not been confirmed for thisstudy area. The larger number of landslides after the year 2000,compared to the previous periods, only indicates a larger numberof reported landslides as the older landslides are not always visibleanymore in the landscape.

4.4. Causal factors

4.4.1. Preconditions

4.4.1.1. Topographic factors. The frequency distributions (Fig. 8) showthat the distribution of most of the topographic values for slopesexhibiting landslides deviated from those for the whole study area.Chi-square tests confirm that the differences are significant (signifi-cance level 0.95) for the topographic factors distance to river, altitude,aspect, and plan curvature. Cramer's V tests indicate the distance tothe rivers as the most important topographic factor. The distance to

Year of activation

2000

2001

2009

2008

2007

2006

2005

2004

2003

2002

2

10

4

6

0

8

Nu

mb

er o

f la

nd

slid

es

12

14

16

Unkno

wn

Befor

e 19

90

1990

- 19

99

Fig. 7. Number of landslides activated per year (n=60).

rivers is important in the study area because of the high river densitywith rivers occurring in narrow V-shaped valleys which may lead toslope instability near the rivers and finally to a landslide. There isalso a significant overrepresentation of recent landslides at lower al-titudes. This is caused by the negative correlation between altitudeand distance to rivers; therefore altitude is not a direct precondition.Next, significantly more landslides occurred on slopes orientedtowards the north-west and north-east. However, a good reason forthe overrepresentation of landslides oriented towards the north-west and north-east could not be found. Also the plan curvature of aslope is indicated as a precondition for landslides. At planconcaveslope segments, water can stagnate, infiltrate and accumulate,which can facilitate landsliding. Further, there is a significant overrep-resentation of landslides on slopes with low slope gradients. This isremarkable, as most studies in East Africa found an overrepresenta-tion of landslides on steep slopes (e.g. Ayalew, 1999; Knapen et al.,2006). The overrepresentation of recent landslides on low slope gra-dients found here is caused by the fact that those slopes are locatednear the rivers. River incisions and bank erosion make slopes unsta-ble, even slopes with low slope gradient. Therefore, it is clear that inour study area slope gradient is less important as a preconditionthan the distance to rivers. However further away from the rivers,the slope gradient becomes important. When excluding the areaslying closer than 100 m to the rivers, a significant (significance level0.90) overrepresentation of recent landslides on steep slopes appears,which indicates steep slopes as a precondition further away from therivers. Compared to landslides in other East African regions (Table 1),the critical slope for landslides in the study area is rather low. Land-slides in the study area occur on slopes with a gradient between8.5° and 45°, as determined during fieldwork. The recent landslidesthat occurred on a very low slope gradient (b20°) are almost all land-slides that occurred near to a river channel. Because of a river incision,the lateral support of the river banks disappears and the slope next tothe river becomes unstable, even with a low slope gradient. This indi-cates again that river incision is an important factor in causing land-slides in the study area. Further away from the rivers, landslidesoccur on steeper slopes, with slope gradients between 16° and 45°.These are gradients which are comparable with other studies in EastAfrica (Table 1).

4.4.1.2. Lithology and soils. The geological and pedological propertiesof the study area are important preconditions for the recent land-slides in the study area for different reasons. (i) According to theFAO-Unesco soil map (1974) the major soil types in the study areaare Acrisols, Vertisols and especially Nitisols. The latter can be aprecondition for landslides because of their swell shrink properties(e.g. Inganga et al., 2001; Knapen et al., 2006). Because the distribu-tion of the soil types in the study area is not known, it was impossibleto detect a possible overrepresentation of landslides on the Nitisols.(ii) For 16 landslides, the clay content of the surface of rupture wasdetermined (Fig. 9). For 10 out of 16 landslides, the clay contentis above 30%. This high clay content in the soil can be an importantprecondition for landslides, because of the chemical and physicalproperties of clay minerals (e.g. Yalcin, 2007). (iii) On most placesin the study area, there exists a distinct boundary between differentgeological layers and between weathered and less weathered materi-al. This change in texture and material may cause a perched watertable around their contact during a high rainfall event, which can re-sult in favourable conditions for landslides (e.g. Temesgen et al.,2001). (iv) According to Schelstreate (2010), differences in clay-mineralogy in the subsoil can be preconditions for recent landslides.At some soil profiles in the study area, there is a change from a dom-inance of 1:1 phyllosilicate like kaolinite towards a dominance of 1:2phyllosilicate like smectite underneath. It is assumed that this changecan be a precondition for the landslides in the study area. Namely, inthe material above the surface of rupture, infiltrating rainwater is less

Chi-square Cramer's V

Chi² P-value

Distance to rivers 38 <0.0001 0,43

Altitude 29 0,0001 0,38

Planimetric curvature 62 <0.0001 0,55

Aspect 17 0,016 0,29

Slope gradient 7,9 0,7 0,2

Slope gradient (>100mfrom rivers)

15 0,1 0,3

Landslide slopes (n=60 landslides)Whole study area

0

10

20

30

40

50

60

70

Fre

qu

ency

(%

)F

req

uen

cy (

%)

Fre

qu

ency

(%

)

Fre

qu

ency

(%

)F

req

uen

cy (

%)

Fre

qu

ency

(%

)

Distance to rivers (m)

0

10

20

30

40

50

convex straight concave

planimetric curvature

0

5

10

15

20

0 -

5

5_10

10_1

5

15-2

0

20-2

5

25-3

0

30-3

5

35-4

0

40-4

5

45-5

0

> 5

0

Slope gradient (%)

0

10

20

30

40

50

Altitude (m.a.s.l.)

0

10

20

30

40

N NE E SE S SW W NW

Aspect

0

5

10

15

20

25

Slope gradient (%) for slopes more than100m away from a river

Fig. 8. Graphical and statistical comparison of the topographic factors distance to river, altitude, planimetric curvature, aspect, slope gradient and slope gradient for slopes morethan 100 m away from a river; both for slopes with landslides and all slopes in the study area. All values of the topographic factors where deduced from a DEM (spatial resolution33 m).

132 N. Broothaerts et al. / Catena 97 (2012) 127–136

retained due to the dominance of the 1:1 phyllosilicate. The materialbeneath is dominated by 1:2 phyllosilicate, clay that starts to swellwhen wetted. This swelling layer can form the surface of rupture ofthe landslide. The water is retained above the swelling clays, and alandslide can occur (Schelstreate, 2010). (v) In order to get a roughidea of the underlying geology of the landslides, during the fieldworkthe underlying bedrock for 19 out of 60 recent landslides was deter-mined by a preliminary visual analysis. The lithology of 11 of these19 landslides was determined as trachyte, 2 as trachyte tuff and 6 asbasalt. This suggests that landslides occur more on trachyte than onbasalt. Indeed, as trachyte contains more silica than basalt, trachytecan weather more easily to 2:1 phyllosilicate with typical swell-shrink capacities which may cause landslides. To confirm that the oc-currence of trachyte is a precondition for landslides, a more detailedgeological map is needed to estimate the distribution of trachyte inthe study area and to compare it with the location of landslides.

4.4.1.3. Annual rainfall. In the study area, annual rainfall is rather high:around 2000 mm y−1 (National Meteorological Agency of Ethiopia,2009). The high annual rainfall and the concentration of the rain inone rainy season (from May till September) causes saturation of thesoil, and positive pore pressure in the soil during the rainy seasonwhich can then serve as a catalyst to allow other causal factors toact more effectively in causing a landslide. The high annual rainfallshould therefore be seen as a precondition for landslides in thestudy area.

4.4.1.4. Land use. In the study area 65% of the landslides occurred oncropland, 21% on so-called bush land – land with shrub and fewtrees extensively used by the farmers – and 10% on grazing land.The exact distribution of land use in the whole study area is unknown,only an estimation of the total cropland area as 40% of the total area isavailable (Tadesse et al., 2006). This indicates an overrepresentation

Table 1Slope gradients for landslide occurrence in East Africa (based on Knapen et al., 2006).

Region (Country) Slope gradient Source

Tropics 28–45° Thomas, 1994Ethiopian Highlands 20–45° Ayalew, 1999Northern EthiopianHighlands

19–56° Nyssen et al., 2002

10–45° Woldearegay et al., 2005>2° Van Den Eeckhaut et al., 2009

Wendo Genet (Ethiopia) 10–20° Temesgen et al., 2001Central Kenyan Highlands 2–53° Westerberg and Christiansson, 1999

>20° Ngecu and Ichang'i, 1999>35° Davies, 1996

Western Uluguru(Tanzania)

30–40° Lundgren, 1978

28–44° Temple and Rapp, 1972Mount Elgon (Uganda) 14–41° Knapen et al., 2006Gilgel Gibe catchment(Ethiopia)

8–45° This study

Fores

t in ‘8

4 - f

ores

t in ‘0

5

Fores

t in ‘8

4 - n

o fo

rest

in ‘05

No fo

rest

in ‘84

- fo

rest

in ‘05

No fo

rest

in ‘84

- no

fore

st in

‘05

40

30

20

10

0

50

60

Landslides Study area

Fre

qu

ency

(%

)

Fig. 10. Graphical and statistical comparison of the forest and other land cover types(no forest) in 1984 and 2005, for the landslides formed since 1984 (n=49) and thewhole study area.

133N. Broothaerts et al. / Catena 97 (2012) 127–136

of landslides on cropland. Moreover, no landslides have beenobserved or reported to occur on slopes with forest vegetation, most-ly located at the highest points in the study area. Therefore, land usecan indeed be a precondition for recent, shallow landslides in thestudy area. Also deep-seated landslides (>3 m deep) in the studyarea may be influenced by the land use and cultivation, as intensiveagriculture causes higher runoff coefficients, higher peak flow dis-charges and hence more river incisions which may destabilize slopesclose to the rivers. This interaction will be further addressed in thenext two paragraphs.

4.4.2. Preparatory factors

4.4.2.1. Deforestation. The impact of a forest cover on slope stability iscomplex. However in general one could say that a forest cover has apositive effect on the slope stability, especially for shallow landslides.The most important effect is the extra cohesion of the soil, given bythe roots of the trees. Deforestation can therefore lower the stabilityof a slope, and can shift the slope to a marginally stable state(Selby, 1993). Deforestation has been reported as one of the mainpreparatory factors for landslides in East Africa (e.g. Davies, 1996;

Fig. 9. Fine-earth particle-size distribution of the soil material at or near the surface ofrupture, for 16 recent landslides.

Inganga et al., 2001; Knapen et al., 2006; Nyssen et al., 2002). A com-parison of Landsat images of 1984 and 2005 by Getahun et al. (2010)shows a significant deforestation in the study area in this period. Theforest cover drops from 7.8 km² (56% of the study area) in 1984 toonly 1.6 km² (11% of the study area) in 2005; a decline in forestcover of almost 80% in a period of 21 years. No information is avail-able about the rates of deforestation before 1984 and after 2005.When comparing the places which are deforestated between 1984and 2005 and the position of the recent landslides which occurredsince 1984 (Fig. 10), it is clear that there is a significant (significancelevel 0.90) overrepresentation of landslides on areas without forest,and especially on recently deforestated areas. Deforestation cantherefore be indicated as a preparatory factor for recent, shallowlandslides (b3 m deep) in the study area by removal of the stabilizingforest vegetation. Deforestation in the study area not only have an im-pact on shallow landslides but also on deep-seated landslides ( >3 mdeep), as deforestation increases runoff and hence increases the oc-currence of peak flow discharges in the rivers, which causes bank ero-sion leading to deep-seated landslides close to the rivers. Thisinteraction will further be addressed in the next paragraph.

4.4.2.2. Bank erosion and river incision. As shown above, there is a sig-nificant overrepresentation of landslides very close to the rivers(Fig. 8). 42 out of 60 recent landslides (70%) are located next to ariver, from which 32 landslides (which is 53% of all landslides in thestudy area) are directly influenced by river incision or bank erosion.River incision or bank erosion shifts the slopes next to a river to amarginally stable state by removing the lateral support and can there-fore be seen as an important preparatory factor for 53% of the recentlandslides. River incision and bank erosion frequently occur in thestudy area especially in the last 20 years, according to local farmers.This is probably related to deforestation and intensification of agricul-ture in the last 20 years, as these land use changes increased the run-off and finally initiated bank erosion. Therefore deforestation not onlycaused more shallow landslides, but also a lot of deep-seated land-slides close to the rivers. A similar relation between deforestation,river incisions and landslides was found in other areas in East Africa(e.g. Moeyersons et al., 2004, 2010; Woldearegay et al., 2005).

4.4.2.3. Terraces. Three landslides (5% of all landslides in the studyarea) are reported to have been preceded by the construction of ter-races. The construction of terraces is not widespread in the studyarea. However at sites where they have been constructed, it maycause water stagnation and increased infiltration, which can lead toan increased pore water pressure and a higher landslide risk. During

Table 2Suspended sediment load upstream and downstream landslide (ls) zone for 10 samplingplaces.

Disharge (m³/s) atmoment of sampling

Suspended sediment concentration (g/l)

Before ls zone After ls zone

3.5 0.2 0.34.5 3.1 3.14 2.4 2.63 0.2 0.14.5 2.7 6.80.06 0.2 7.60.03 0.1 0.20.06 0.1 1.00.07 1.3 4.60.08 1.1 0.7

134 N. Broothaerts et al. / Catena 97 (2012) 127–136

a long period of heavy rainfall, these places become extremely unsta-ble and landslides can occur.

4.4.3. Triggering factors

4.4.3.1. Heavy rainfall. Ninety percent of the landslides occurredduring or after a heavy rainfall event, therefore heavy rainfall is con-sidered to be the main triggering factor for recent landslides inthe study area. Heavy rainfall increases the total weight of the slopematerial, causes the water table to rise and increases the pore pres-sure in fine-grained sediments, and can therefore trigger a landslide(e.g. Abebe et al., 2010). It was not possible to assess a rainfall thresh-old for triggering a landslide in the study area, because of lack ofdetailed landslide dates.

4.4.3.2. Seismicity. Although the dates of occurrence of recent land-slides and earthquakes in the study area cannot be compared due tolack of temporal detail in the data, earthquakes are not thought tobe an important triggering factor for recent landslides in the studyarea. First, the frequency of the earthquakes in the study area(EELPA, 1997) is much smaller than the frequency of the recentlandslides. Moreover, the study area is at least at a distance of90 km from the most active seismic centres, namely the East AfricanRift structures. Finally, during the interviews none of the respondentsindicated that a landslide occurred during or immediately after anearthquake.

4.4.3.3. Bank erosion and river incision. Four recent landslides (7% of allrecent landslides) were mainly triggered by bank erosion and riverincision, and these factors can therefore be seen as the triggering fac-tor for these landslides. However bank erosion and river incision ismainly a preparatory factor for the recent landslides in the study area.

4.5. Consequences

Although no fatalities due to landslides are reported in the studyarea, the impact on the inhabitants is enormous and calls for mea-sures. First, most landslides occur on cropland and make the landpartly useless for several years. Consequently, land becomes scarcerand the pressure on the remaining land becomes higher. Land short-age is indeed one of the main problems farmers in the Gilgel Gibecatchment are faced with. In total, during the last 20 years at least0.2 km² of cropland has been damaged, which is 3% of the total crop-land area in the study area. Because the reported landslides are an un-derestimation of the real amount of landslides, this damage is anunderestimation of the real percentage of damage to cropland. Sec-ondly, some landslides are reported to cause damage to houses androads. Finally, the landslides caused the death of some cattle, whichcan be a disaster for the families involved.

Next to these socio-economic consequences, landslides in thestudy area have important geomorphic consequences. In total the re-cent landslides occupied an area of around 275 000 m², which is al-most 2% of the total study area. Together, the 60 recent landslides inthe study area of 14 km² replaced about 1 million m³ of slope materialduring the last 20 years, which equals a displaced volume of 50 tonha−1 y−1. This is an underestimation because not all landslides thathappened during the last 20 years are still visible in the landscape.Further, the recent landslides have an important impact on the riverchannels. The accumulation zone of 45 recent landslides (which is75% of the total number of recent landslides) enters directly into a(small) river, where the river can pick up material from these land-slides. In order to quantify the effects of these landslides on thesuspended sediment load in the rivers, the suspended sediment con-centration upstream and downstream the landslide was determinedand compared (Table 2). For 3 out of the 10 samples the differenceis clearly visible and significant. Especially during rainfall events the

increase in sediment load was high during rainfall events. Thesethree samples indicate that especially during rainfall events, whenthe discharge of the rivers is higher, the increase in sediment loadcan be very high. This increase can be attributed to landslides, be-cause no other important sediment sources could be detected in thesurroundings. It was estimated that all rivers together in the studyarea eroded ca. 225 000 m³ of soil in total, which was delivered tothe rivers by those 45 recent landslides during the last 20 years.This volume corresponds to a total soil loss by river erosion of 11ton ha−1 y−1. This is even an underestimation because the reportednumber of landslides in the study area is an underestimation of thetrue number of landslides in the last 20 years.

Further, the recent landslides have an import impact on the riverpatterns. The landslides can also deliver large rock fragments (up to80 cm in diameter), originating from the underlying regolith, intothe rivers. The rivers can pick up these rock fragments during peakflows and transport them over a short distance. Most of these largerock fragments are deposited just downstream the landslide zone,where gravel bars are formed. These gravel bars can totally changethe river pattern. Examples of large gravel bars can be found justdownstream the study area (Fig. 11), where many large rock frag-ments are deposited.

5. Conclusions

The study area within the Gilgel Gibe catchment is an unstable re-gion, intensively affected by landslides. These landslides are the resultof a complex combination of preconditions, preparatory factors andtriggering factors. The typical lithological and pedological properties,the high annual rainfall, the cultivation of most parts of the land andto a lesser extent steep slopes, make the study area very sensitive tolandslides and act as catalysts to allow other factors to act more effec-tively. However the main precondition for the recent landslides is thedistance to rivers. Hillslopes close to a river are less stable than theslopes further away from the rivers. This is mainly caused by theriver incisions and bank erosion, which in turn is probably relatedto land use changes in the catchment like deforestation and intensifi-cation of agriculture. Therefore, deforestation not only caused shallowlandslides, but also deep-seated landslides close to rivers. Finally, al-most all landslides are triggered by heavy rainfall.

It is clear that the landslides in the study area are a significantproblem for the local farmers. Therefore, it is important to look for re-liable and sustainable measures. Total reforestation with deep-rootedtrees can possibly reduce the landslide risk, although this is not a re-alistic measure due to the shortage of arable land in this highly pop-ulated area. Partly reforestation of the riparian zone can locallyenhance stability, and can protect slopes from destabilization byriver incisions and bank erosion. Further deforestation of the riparianzone in this type of regions should be avoided in any case.

Fig. 11. Illustration of the impact of landslides on sediment supply to a river and its resulting change of a river channel. A. Location of the river section (B) just downstream the studyarea with indication of the mapped landslides. B. Detailed map of the gravel bars as mapped on August 31, 2009. Stream direction is from left to right. C. Downstream view of theriver section shown in (B).

135N. Broothaerts et al. / Catena 97 (2012) 127–136

Moreover the landslides caused an overall displaced volume of 50ton ha−1 y−1 during the last 20 years. Many landslides transport soilmaterial directly into the river and cause a higher sediment load inthe rivers. The total soil loss to the rivers was estimated at 11 tonha−1 y−1 during the last 20 years. Both measurements are underes-timations. Knowing that the study area is only 14 km² and that alsoin other places of the Gilgel Gibe catchment landslides do occur(Ayalew, 1999; Abebe et al., 2010; own observations), the landslidesneed to be considered as an important sediment source for the GilgelGibe reservoir. However, more research is needed on other landslideareas in the catchment and especially on the contribution of the land-slides to the increased sediment load in the rivers, so that the impactof landslides on the sediment delivery into the Gigel Gibe reservoircan be calculated. If one wants to extend the lifespan of the reservoir,

effective measures to control landsliding in the catchment need tobe undertaken, especially because other hydro-electric dams areplanned to be built in this catchment in the near future.

Acknowledgements

This study was conducted in the framework of Soil Fertility projectof the IUC-JU (VLIR), a cooperation between Jimma University(Ethiopia) and different Flemish universities (together in the FlemishInteruniversity Council VLIR-UOS). Funding for the fieldwork wasprovided by a travel grant of the Flemish Interuniversity CouncilVLIR-UOS. Special appreciation is extended to all members of theVLIR-IUC project for support during the fieldwork, and to the farmersin the study area who provided useful information.

136 N. Broothaerts et al. / Catena 97 (2012) 127–136

References

Abebe, B., Dramis, F., Fubelli, G., Mohammed, U., Asrat, A., 2010. Landslides inthe Ethiopian highlands and the Rift margins. Journal of African Earth Sciences56 (4–5), 131–138.

Ayalew, L., 1999. The effect of seasonal rainfall on landslides in the highlands ofEthiopia. Bulletin of Engineering Geology and the Environment 58, 9–19.

Ayalew, L., Yamagishi, H., 2004. Slope failures in the Blue Nile basin, as seen fromlandscape evolution perspective. Geomorphology 57, 95–116.

Ayenew, T., Barbieri, G., 2005. Inventory of landslides and susceptibility mapping in theDessie area, northern Ethiopia. Engineering Geology 77, 1–15.

Coltorti, M., Pieruccini, P., Berakhi, O., Dramis, F., Asrat, A., 2009. The geomorphologicalmap of Mt. Amba Aradam Sothern Slope (Tigray, Ethiopia). Journal of Maps 7,56–65.

Cruden, M.D., Varnes, D.J., 1996. Landslide types and processes. In: Turner, K.A.,Schuster, R.L. (Eds.), Landslides: Investigation and Mitigation. National AcademyPress, Washington D. C.. Special report 247 / Transportation Research Board,National Research Council.

Davies, T.C., 1996. Landslide research in Kenya. Journal of African Earth Sciences 23 (4),541–545.

Devi, R., Tesfahune, E., Legesse, W., Deboch, B., Beyene, A., 2008. Assessment of siltationand nutrient enrichment of Gilgel Gibe dam, Southwest Ethiopia. BioresourceTechnology 99, 975–979.

Ethiopian Electric Light and Power Authority (EELPA), 1997. Gilgel Gibe hydroelectricproject: Environmental assessment main report. EELPA, Ministry of Mines andEnergy, Addis Ababa, Ethiopia.

Ethiopian Mapping Agency, 1980. Topographic map, sheet 0736 D2 –Metoso, 1:50000.Addis Ababa, Ethiopia.

FAO-Unesco, 1974. FAO/Unesco Soil Map of the World 1:5.000.000, Unesco, Paris.Fubelli, G., Abebe, B., Dramis, F., Vinci, S., 2008. Geomorphological evolution and

present-day processes in the Dessie Graben (Wollo, Ethiopia). Catena 75, 28–37.Getahun, K., Van Rompaey, A., Poesen, J., 2010. Detection and Spatial Analysis of

Recent Landuse Changes in Southwest Ethiopia. IUC Research Day. Jimma Univer-sity, Jimma, Ethiopia. September 18th 2010, (Available online: http://www.iucju.ugent.be/prog/poster/submitted/Kefelegn_Getahun.pdf).

Glade, T., Crozier, M.J., 2005. The nature of landslide hazard impact. In: Glade, T.,Anderson, M., Crozier, M.J. (Eds.), Landslide Hazard and Risk. John Wiley andSons, Chichester.

Gokceoglu, C., Sezer, E., 2009. A statistical assessment on international landslideliterature (1945–2008). Landslides 6, 345–351.

Gutiérrez, F., Soldati, M., Audemard, F., Balteanu, D., 2010. Recent advances in landslideinvestigation: issues and perspectives. Geomorphology 124, 95–101.

Inganga, S.F., Ucakuwun, E.K., Some, D.K., 2001. Rate of swelling of expansive soils: acritical factor in the triggering of landslides and damage to structures. DocumentaNaturae 136, 93–98.

Kitutu, M.G., Muwanga, A., Poesen, J., Deckers, J.A., 2009. Influence of soil propertieson landslide occurrences in Bududa district, Eastern Uganda. African Journal ofAgricultural Research 4 (7), 611–620.

Knapen, A., Kitutu, M.G., Poesen, J., Breugelmans, W., 2006. Landslides in a denselypopulated county at the foot slopes of Mount Elgon (Uganda): characteristicsand causal factors. Geomorphology 73, 149–165.

Larsen, I.J., Montgomery, D.R., Korup, O., 2010. Landslide erosion controlled by hillslopematerial. Natural Geoscience 3, 247–251.

Lundgren, L., 1978. Studies of Soil and Vegetation Development on Fresh LandslideScars in the Mgeta Valley, Western Uluguru Mountains, Tanzania, GeografiskaAnnaler, Series A. Physical Geography 60 (3/4), 91–127.

Malamud, B.D., Turcotte, D.L., Guzetti, F., Reichenbach, P., 2004. Landslide inventoriesand their statistical properties. Earth Surface Processes and Landforms 29,687–711.

Marwa, E.M.M., Kimaro, D.N., 2005. Geomorphological mapping for route selection androad construction at lower Kihansi, Tanzania. Bulletin of Engineering Geology andthe Environment 64, 333–339.

Moeyersons, J., 1989. A possible causal relationship between creep and sliding onRwaza Hill, Southern Rwanda. Earth Surface Processes and Landforms 14, 597–614.

Moeyersons, J., 2003. The topographic thresholds of hillslope incisions in southwesternRwanda. Catena 50, 381–400.

Moeyersons, J., Tréfois, Ph., Lavreau, J., Alimasi, D., Badriyo, I., Mitima, B., Mudala, M.,Munganga, D.O., Nahimana, L., 2004. A geomorphological assessment of landslide

origin at Bukavu, Democratic Republic of the Congo. Engineering Geology 72,73–87.

Moeyersons, J., Van Den Eeckhaut, M., Nyssen, J., Gebreyohannes, T., Van de Wauw, J.,Hofmeister, J., Poesen, J., Deckers, J., Mitiku, H., 2008. Mass movement mappingfor geomorphological understanding and sustainable development: Tigray, Ethio-pia. Catena 75, 45–54.

Moeyersons, J., Tréfois, Ph., Nahimana, L., Ilunga, L., Vandecasteele, I., Byizigiro, V.,Sadiki, S., 2010. River and landslide dynamics on the western Tanganyika riftborder, Uvira, D.R. Congo: diachronic observations and a GIS inventory of tracesof extreme geomorphologic activity. Natural Hazard 53 (2), 291–311.

National Meteorological Agency of the Federal Democratic Republic of Ethiopia, 2009.Daily precipitation and temperature data. Ethiopian Government, Ethiopia. www.ethiomet.gov.et 2009.

Ngecu, W.M., Ichang'i, D.W., 1999. The environmental impact of landslides on the pop-ulation living on the eastern foot slopes of the Aberdare ranges in Kenya: a casestudy of Maringa Village landslide. Environmental Geology 38 (3), 259–264.

Ngecu, W.M., Mathu, E.M., 1999. The El-Nino-triggered landslides and their socioeco-nomic impact on Kenya. Environmental Geology 38 (4), 277–284.

Ngecu, W.M., Nyamai, C.M., Erima, G., 2004. The extent and significance of mass-movements in Eastern Africa: case studies of some major landslides in Ugandaand Kenya. Environmental Geology 46, 1123–1133.

Nyssen, J., Poesen, J., Moeyersons, J., Deckers, J., Haile, M., 2002. The environmental sig-nificance of the remobilisation of ancient mass movements in the Atbara-Tekezeheadwaters, Northern Ethiopia. Geomorphology 49, 303–322.

Nyssen, J., Poesen, J., Moeyersons, J., Deckers, J., Haile, M., Lang, A., 2004. Human impacton the environment in the Ethiopian and Eritrean highlands—a state of the art.Earth-Science Reviews 64, 273–320.

Nyssen, J., Poesen, J., Moeyersons, J., Deckers, J., Haile, M., 2006. Processes and ratesof rock fragment displacement on cliffs and scree slopes in an amba landscape,Ethiopia. Geomorphology 81, 265–275.

Schelstreate, 2010. Lithology andweathering products at the slip plane of landslides in SWEthiopia. MSc Thesis, Faculteit wetenschappen, Universiteit Gent. (Available online:http://lib.ugent.be/fulltxt/RUG01/001/458/729/RUG01-001458729_2011_0001_AC.pdf).

Selby, M.J., 1993. Hill Slope Materials and Processes, 2nd ed. Oxford University Press,Oxford.

Tadesse, S., Milesi, J.P., Deschamps, Y., 2003. Geology and mineral potential of Ethiopia:a note on geology and mineral map of Ethiopia. Journal of African Earth Sciences36.

Tadesse, M., Alemu, B., Bekele, G., Tebikew, T., Chamberlin, J., Benson, T., 2006. Atlas ofthe Ethiopian rural economy. Ethiopian Development Research Institute.

Temesgen, B., Mohammed, M.U., Korme, T., 2001. Natural hazard assessment using GISand remote sensing methods, with particular reference to the landslides in theWondogenet area, Ethiopia. Physics and Chemistry of the Earth 26 (9), 665–675.

Temple, P.H., Rapp, A., 1972. Landslides in the Mgeta area, Western Uluguru Moun-tains, Tanzania, Geografiska Annaler, Series A. Physical Geography 54 (3/4),157–193.

Thomas, F.M., 1994. Geomorphology in the Tropics: A Study of Weathering and Denu-dation in Low Latitudes. John Wiley and Sons, Chichester.

Van Den Eeckhaut, M., Poesen, J., Govers, G., Verstraeten, G., Demoulin, A., 2007. Char-acteristics of the size distribution of recent and historical landslides in a populatedhilly region. Earth and Planetary Science 256, 588–603.

Van Den Eeckhaut, M., Moeyersons, J., Nyssen, J., Abraha, A., Poesen, J., Haile, M.,Deckers, J., 2009. Spatial patterns of old, deep-seated landslides: a case-study inthe northern Ethiopian highlands. Geomorphology 105, 239–252.

Varnes, D.J., 1978. Slope Movement Types and Processes. In: Schuster, R.L., Krizek, R.J.(Eds.), Landslides: Analysis and Control. National Academy Press, WashingtonD.C. Special Report 176 / Transportation Research Board, National Research Council.

Vilimek, V., Zvelebil, J., Kalvoda, J., Sima, J., 2010. Landslide field research and capacitybuilding through international collaboration. Landslides 7 (3), 375–380.

Westerberg, L.A., Christiansson, C., 1999. Highlands in East Africa: Unstable Slopes,Unstable Environments? Ambio 28 (5), 419–429.

Woldearegay, K., Schubert, W., Klima, K., Mogessie, A., 2005. Landslide hazardsmitigation strategies in the northern highlands of Ethiopia. In: Hungr, O., Fell, R.,Couture, R., Eberhardt, E. (Eds.), Landslide Risk Management, SupplementaryVolume. Taylor & Francis Group, London, pp. 25–37.

Yalcin, A., 2007. The effect of clays on landslides: a case study. Applied Clay Science 38,77–85.