Runoff on slopes with restoring vegetation: A case study from the Tigray highlands, Ethiopia

This article was originally published in a journal published byElsevier, and the attached copy is provided by Elsevier for the

author’s benefit and for the benefit of the author’s institution, fornon-commercial research and educational use including without

limitation use in instruction at your institution, sending it to specificcolleagues that you know, and providing a copy to your institution’s

administrator.

All other uses, reproduction and distribution, including withoutlimitation commercial reprints, selling or licensing copies or access,

or posting on open internet sites, your personal or institution’swebsite or repository, are prohibited. For exceptions, permission

may be sought for such use through Elsevier’s permissions site at:

http://www.elsevier.com/locate/permissionusematerial

Autho

r's

pers

onal

co

pyRunoff on slopes with restoring vegetation: Acase study from the Tigray highlands, Ethiopia

Katrien Descheemaeker a,*, Jan Nyssen a,b, Jean Poesen c, Dirk Raes a,Mitiku Haile b, Bart Muys d, Seppe Deckers a

a Division Soil and Water Management, K.U. Leuven, Geo-Instituut, Celestijnenlaan 200E, B-3001 Leuven, Belgiumb Department of Land Resource Management and Environmental Protection, Mekelle University, P.O. Box 231,Mekelle, Ethiopiac Physical and Regional Geography Research Group, K.U. Leuven, Geo-Instituut, Celestijnenlaan 200E, B-3001Leuven, Belgiumd Division Forest, Nature and Landscape, K.U. Leuven, Geo-Instituut, Celestijnenlaan 200E, B-3001 Leuven, Belgium

Received 24 June 2005; received in revised form 11 May 2006; accepted 12 May 2006

Summary Daily runoff depths from 28 plots (5 m · 2 m) recorded during a 2-year period in thesemi-arid to subhumid highlands of Tigray were analyzed to study the effect of vegetation res-toration in exclosures and to identify other factors influencing runoff production. Plots are dis-tributed over three study sites and located in different land use types and on differentcombinations of soil type, vegetation cover and slope gradient. Runoff was found to be signif-icantly reduced when a degraded area is allowed to rehabilitate after closure. Runoff depth issignificantly correlated with event variables such as rain depth, rainfall intensity, storm dura-tion and soil moisture content. Total vegetation cover is the most important plot variableexplaining about 80% of the variation in runoff coefficients through an exponential decay func-tion. Also the runoff generating rainfall threshold has a positive correlation with total vegeta-tion cover. Runoff was found to be negligible when the vegetation cover exceeds 65%. Otherimportant variables affecting runoff production in the study sites are soil organic matter, soilbulk density, litter cover and slope gradient.

�c 2006 Elsevier B.V. All rights reserved.

KEYWORDSExclosure;Rangeland;Grazing;Eucalyptus;East Africa;Surface hydrology

Introduction

The natural environment of the northern Ethiopian high-lands is seriously threatened by land degradation (e.g. Hur-ni, 1990; Herweg and Stillhardt, 1999; Nyssen et al., 2004),mainly caused by the combined effects of deforestation,

0022-1694/$ - see front matter �c 2006 Elsevier B.V. All rights reserved.doi:10.1016/j.jhydrol.2006.05.015

* Corresponding author. Tel.: +32 16329765; fax: +32 16329760.E-mail address: [email protected] (K.

Descheemaeker).

Journal of Hydrology (2006) 331, 219–241

ava i lab le a t www.sc iencedi rec t . com

journal homepage: www.elsevier .com/ locate / jhydrol

Autho

r's

pers

onal

co

py

overgrazing, expansion of cropland and unsustainable use ofnatural resources. The alarming rate of desertification(UNEP, 1994) in Tigray region justifies the attention and ef-forts that have been paid to various techniques of soil andwater conservation (Herweg and Stillhardt, 1999; SCRP,2000; Nyssen et al., 2004). Apart from the construction ofphysical barriers, also vegetation restoration is an option(Chadhokar and Solomon, 1988; Fiedler and Gebeyehu,1988), as it is believed to be a powerful tool for site rehabil-itation (Hongo et al., 1995; Gao et al., 2002; Aerts et al.,2004; Zhang et al., 2004). Since the 1980s, this is achievedby the local government and communities through the clos-ing of strongly degraded areas, commonly located on steepslopes. These exclosures (Kebrom et al., 1997; Le Houerou,2000; Wisborg et al., 2000; Tenna et al., 2001) are areas setaside, where agriculture and grazing are forbidden. Themain objective is environmental rehabilitation through res-toration of natural vegetation. Also, income generation andcommunity support through food and cash for work areaimed at, besides production of grass for fodder and thatch-ing, as well as wood and non-wood forest products.

Runoff production is an important process in land degra-dation, causing soil erosion and influencing the soil waterbalance and hydrology of the catchment. Recently, exclo-sures have become a controversial land use, suspect to causemore pressure on already scarce grazing land in the study re-gion. Therefore, it is important to obtain more insight intothe effects of the set-aside policy on runoff production.

Many authors have discussed the runoff behavior of dif-ferent land use types and the effects of land use changeon runoff production (e.g. Calder et al., 1995; Kosmaset al., 1997; Narain et al., 1998; Cammeraat and Imeson,1999; Castro et al., 1999; Vacca et al., 2000; Bellot et al.,2001; Kang et al., 2001; McDonald et al., 2002; Dagnachewet al., 2003; Pardini et al., 2003; Dunjo et al., 2003, 2004).When afforestation/reforestation or vegetation restorationis concerned, it is commonly concluded that runoff ratesand peak flows are reduced (e.g. Mapa, 1995; Zhou et al.,2002; Huang and Zhang, 2004; Zhang et al., 2004), but alsobase flows may decrease as a result of increased evapo-transpiration (Bruijnzeel, 2004). In a review on rainfall–run-off modeling in arid and semi-arid regions, Pilgrim et al.(1988) claim that there is a serious lack of data on runoffproduction in these regions. They stress the importance ofan increased knowledge on the impact of vegetation, landmanagement and grazing practices on runoff production tosupport decision making in land use planning. Since this re-view, a significant number of studies have been conductedon runoff processes in relation to vegetation and other vari-ables in semi-arid regions, but the majority of them focus onthe Mediterranean environment (e.g. Yair and Lavee, 1985;Sala and Calvo, 1990; Sorriso-Valvo et al., 1995; Nicolauet al., 1996; Castillo et al., 1997; Kosmas et al., 1997;Sole-Benet et al., 1997; Cerda, 1997a,b, 1998; Bochetet al., 1998; Lopez-Bermudez et al., 1998; Martınez-Menaet al., 1998; Puigdefabregas et al., 1999; Lasanta et al.,2000; Vacca et al., 2000; Archer et al., 2002; Desir, 2002;Yair and Kossovsky, 2002; Calvo-Cases et al., 2003; Pardiniet al., 2003; Dunjo et al., 2003, 2004). As Pilgrim et al.(1988) point out, there is a great diversity in hydrologicalcharacteristics within (semi)arid regions, so that it wouldbe unrealistic to assume the findings of these studies to

be unconditionally applicable to Tigray. Not only the differ-ence in rainy season (winter versus summer) but also thedifference in elevation of both regions call for specific re-search in the Tigray highlands.

Mapa (1995), Descroix et al. (2001) and Archer et al.(2002) argue that relatively more studies refer to runoffcharacteristics in arable land than to natural vegetationand rangeland areas. Studies on runoff processes in range-lands have been conducted mainly in North America (e.g.Wilcox and Wood, 1988, 1989). In East Africa however,where the high population density of people and livestockin the highlands indicate the importance of rangelands, suchstudies are rather scarce. Gutierrez and Hernandez (1996)further indicate that there is great uncertainty on theamount of vegetation cover needed to reduce runoff fromsemi-arid rangelands.

Nyssen et al. (2004) provide an overview of runoff studiesconducted in Ethiopia and Eritrea over various scales ofspace and time. For a watershed in the northwestern high-lands, Woldeamlak and Sterk (2005) found a decreasing evo-lution in stream flow, which was attributed to changes inland cover, land use, degradation of the watershed and ahigher dry-season water consumption by an increasing pop-ulation. Also Dagnachew et al. (2003) investigated the ef-fect of land use changes on the hydrological response of acatchment in South Central Ethiopia. Mwendera and Mo-hamed Saleem (1997), Mwendera et al. (1997) and Taddeseet al. (2002a,b) investigated in Ethiopia the impact of graz-ing on vegetation attributes, soil physical and hydrologicalprocesses. Their findings indicate that with increasing graz-ing pressure, vegetation cover decreases and soil compac-tion increases, leading to lower infiltration rates andhigher runoff. Within the framework of the Soil Conserva-tion Research Programme (SCRP, 2000) in Ethiopia, up to12 years long data series were collected on runoff measuredat different plot scales. These studies mainly focused oncultivated areas to ascertain the runoff behavior of differ-ent cultivation practices, soil types, climates and to evalu-ate the effects of various soil and water conservationpractices in farmers’ fields.

For the case of the northern Ethiopian highlands in par-ticular, no runoff studies focusing on natural vegetation inrangelands have been conducted. Despite the importanceof exclosures as a soil and water conservation strategy,the effects on runoff of naturally regenerating vegetationcolonizing degraded hillslopes remain uninvestigated.

The overall objective of this study is therefore to betterunderstand runoff production in exclosures and hence tocontribute to improved land management practices for soiland water conservation in the Tigray highlands. More spe-cific objectives are (1) to quantify the impact of exclosureestablishment on runoff production from steep hillslopesand (2) to identify influencing variables and their impor-tance in controlling runoff processes.

Materials and methods

Study area

The study area is located in the Dogu’a Tembien (Tembienhighlands) district in central Tigray, near the district capital

220 K. Descheemaeker et al.

Autho

r's

pers

onal

co

py

Hagere Selam (13�39 0N, 39�10 0E, altitude 2650 m a.s.l.). Thelithology of the study area is composed of Mesozoic sedi-mentary rocks (Adigrat sandstone, Antalo limestone andAmba Aradam sandstone), covered by Tertiary flood basalts(Bosellini et al., 1997). The subhorizontal layering of rockformations of different resistance to weathering results ina stepped landscape, where flats alternate with steepescarpments (Nyssen et al., 2002). Mean annual rainfall isaround 700 mm, mostly concentrated in the rainy seasonfrom June to September (Nyssen et al., 2005).

Within this area, three representative study sites wereselected, each of them located on different lithologies:i.e. May Ba’ati (limestone, partly covered by non-calcareoussediments), Adewro (basalt), and Kunale (mixed sandstoneand limestone lithology, partly covered by transported ver-tic clay material) (Fig. 1).

Within each study site, several hillslope sections coveredby different land use types have been selected for compar-ative analysis: grazing land, exclosure of different ages andchurch forest (Figs. 2 and 3). The grazing lands are degradedareas where overgrazing has led to the disappearance ofmost vegetation so that severe erosion is taking place,removing nearly all fertile soil. In the protected exclosureareas, the recovery process invariably starts with a rapid in-crease in diversity and cover of herbaceous species, whileafter 3–5 years shrub and tree species gain importanceand start to suppress the herb layer (Eweg et al., 1998).Age of closure goes hand in hand with an increase in vegeta-tion cover and density and with an improvement in soil fer-tility (Descheemaeker et al., in press). The church forest isa sacred place where the vegetation has been protectedsince long time. Taller trees with strongly overlapping can-opies are found here.

Homogeneity in lithology and slope aspect within eachsite provides the chance to isolate explanatory factors suchas vegetation cover. As the grazing lands represent the sit-

uation of worst degradation in each study site, they are usedas a control to compare with the restoring forest areas.

Instrumentation and data collection

Each study site is equipped with one tipping bucket raingauge with a precision of 0.2 mm. A datalogger recordsthe moment of each tip and allows for rainfall intensity cal-culations. For each study site, two additional rain gaugeswere constructed, from which the daily readings comple-ment the records of the tipping bucket rain gauge and allowfor continuous records of rainfall in case of tipping bucketor datalogger failure. In each site the rain gauges are spacedout so that it is possible to adjust daily rainfall depths incase of considerable spatial variation.

Twenty-eight runoff plots of 5 m · 2 m were constructed(of which 15 in May Ba’ati, 8 in Kunale and 5 in Adewro).Their locations (Fig. 2) were not randomly chosen but se-lected to include all combinations of soil type, slope gradi-ent and vegetation type and cover that can be found in thehillslope sections of the study sites. Their coding (Tables 1and 2) reflects site (the first letter of the code correspondswith the first letter of the study site) and land use type (XY,XM, XO, RA, EU and FO for young exclosure, middle-agedexclosure, old exclosure, grazing land, Eucalyptus forestand church forest, respectively). Plots from different studysites and land use types are presented in Fig. 4. Stones wereused to construct 15 cm high cemented walls around theplot (Fig. 5). No considerable effect of these walls on runoffproduction was observed. The area enclosed by the cemen-ted walls was then carefully measured and the plan pro-jected area calculated. A metal container was dug intothe soil to collect all the runoff water from the plot(Fig. 5). For each container a volume-depth relation wasestablished and local secondary school students weretrained to empty daily the container after measuring the

Figure 1 Location of the three study sites and the study region (indicated as a triangle) in the Northeast of Ethiopia (inset top leftcorner).

Runoff on slopes with restoring vegetation: A case study from the Tigray highlands, Ethiopia 221

Autho

r's

pers

onal

co

py

Figure 2 Topography of the three study sites with the selected hillslope sections and their land use.

Figure 3 Typical land use types under study in the May Ba’ati study site: (1) degraded grazing land, (2) young exclosure, (3) oldexclosure, (4) church forest. Except for the latter, all pictures are taken in the dry season.

222 K. Descheemaeker et al.

Autho

r's

pers

onal

co

py

Table 1 Characteristics of runoff plots

Site Plot code Land use type Soil type (WRB) Slope(%)

Clay(%)

Silt(%)

Sand(%)

Rc(%)

Rm(%)

Ksat(m d�1)

SOM(%)

BDtot(Mg m�3)

BDfe(Mg m�3)

WHC(%)

May Ba’ati MXY1 Young exclosure Thaptocalci-Humic Calcisol 65 37 25 39 62 11 0.51 2.3 1.28 1.24 12MXY2 Young exclosure Hypercalcic Calcisol 85 25 27 48 44 15 0.42 1.1 1.43 1.39 13MXY3 Young exclosure Thaptocalci-Humic Calcisol 35 48 40 12 38 30 0.51 1.3 1.36 1.24 12MXO1 Old exclosure Thaptocalci-Humic Calcisol 55 41 33 26 30 22 0.51 3.5 1.33 1.24 12MXO2 Old exclosure Petri-Skeletic Calcisol 40 44 45 11 31 38 0.52 6.4 1.34 1.17 11MXO3 Old exclosure Humi-Endocalcaric Phaeozem 30 52 32 16 11 1 0.29 6.1 1.21 1.21 13MXO4 Old exclosure Humi-Calcaric Cambisol 50 43 31 26 33 24 0.61 3.8 1.05 0.93 12MXO5 Old exclosure Humi-Endocalcaric Phaeozem 40 62 21 17 22 18 0.77 3.3 1.23 1.16 16MXO6 Old exclosure Humi-Endocalcaric Phaeozem 15 52 32 16 14 5 0.29 5.0 1.22 1.20 13MXO7 Old exclosure Humi-Calcaric Cambisol 50 43 31 26 21 13 0.61 5.1 0.99 0.93 12MXO8 Old exclosure Thaptocalci-Humic Calcisol 110 41 33 26 46 29 0.51 3.9 1.36 1.24 12MXO9 Old exclosure Humi-Calcaric Cambisol 70 43 31 26 27 36 0.61 4.4 1.12 0.93 12MFO Church forest Humi-Endocalcaric Phaeozem 35 62 21 17 3 15 0.77 8.3 1.22 1.16 16MRA1 Grazing land Hypercalcic Calcisol 20 30 32 38 44 17 0.42 1.3 1.44 1.39 13MRA2 Grazing land Hypercalcic Calcisol 40 30 32 38 42 15 0.42 1.2 1.43 1.39 13

Kunale KXY1 Young exclosure Humi-Episkeletic Phaeozem 55 41 21 38 20 19 0.33 2.4 1.10 1.01 14KXY2 Young exclosure Thaptocalci-Humic Cambisol 70 35 22 43 24 10 0.39 3.7 1.11 1.06 14KXY3 Young exclosure Humi-Episkeletic Phaeozem 70 43 23 34 28 16 0.33 2.0 1.13 1.06 9KXM1 Middle-aged exclosure Humi-Episkeletic Phaeozem 55 41 21 38 8 19 0.33 2.4 1.10 1.01 14KXM2 Middle-aged exclosure Thaptocalci-Humic Cambisol 75 35 22 43 31 31 0.39 3.7 1.21 1.06 14KXO1 Old exclosure Thaptocalci-Humic Phaeozem 45 37 21 42 7 18 0.36 2.5 1.23 1.15 13KXO2 Old exclosure Thaptocalci-Humic Phaeozem 50 37 21 42 17 26 0.36 2.5 1.26 1.15 13KRA Grazing land Verti-Humic Cambisol 50 43 23 34 33 16 0.33 1.7 1.13 1.06 9

Adewro AXY1 Young exclosure Humi-Skeletic Phaeozem 70 13 29 57 23 26 1.14 5.4 1.02 0.89 8AXY2 Young exclosure Humi-Skeletic Phaeozem 80 13 29 57 46 27 1.14 5.4 1.02 0.89 8AEU1 Eucalyptus forest Humi-Skeletic Phaeozem 40 16 28 56 33 45 1.14 5.3 1.14 0.89 8AEU2 Eucalyptus forest Humi-Skeletic Phaeozem 70 15 28 56 12 32 1.14 5.3 1.05 0.89 8ARA Grazing land Pachi-Skeletic Phaeozem 45 21 57 22 36 48 0.33 4.8 1.10 0.82 12

Mean value (standard deviation) over all plots n.a. 54(21) 37(13) 29(8) 34(14) 28(14) 28(14) 0.56(0.27) 3.7(1.9) 1.20(0.13) 1.10(0.16) 12(2)Minimum value n.a. 15 13 21 11 3 1 0.29 1.1 0.99 0.82 8Maximum value n.a. 110 62 45 57 62 48 1.14 8.3 1.44 1.39 14

All soil variables refer to the topsoil (0–15 cm).WRB: World reference base for soil resources (FAO et al., 1998); Rc: cover of soil surface by rock fragments (%); Rm: rock fragment content of topsoil by mass (%); Ksat: saturated hydraulic conductivity (m d�1); SOM:soil organic matter content (%); BDtot: soil bulk density (Mg m�3); BDfe: bulk density of the fine earth fraction (Mg m�3); WHC: water holding capacity (%).

Runoff

onslo

peswith

resto

ringve

getatio

n:Aca

sestu

dyfro

mtheTigray

high

lands,

Ethiopia

223

Autho

r's

pers

onal

co

py

Table 2 Vegetation cover attributes of the runoff plots

Site Plot code Shrub and tree canopy cover, VEGst (%) Grass and herb cover, VEGgh (%) Total vegetation cover, VEGtot (%) VEGtot,w(%)

Littercover(%)

Dry June July August September Dry June July August September Dry June July August September

May Ba’ati MXY1 25 30 30 29 46 5 34 34 33 49 27 50 51 50 57 49 11MXY2 18 22 22 29 29 3 27 26 29 36 21 42 44 46 55 44 30MXY3 43 49 62 62 66 3 12 15 37 32 43 52 62 74 70 65 35MXO1 42 45 53 64 68 2 18 15 12 14 42 48 56 64 68 58 41MXO2 38 39 60 59 66 3 10 17 15 15 38 41 61 59 66 56 58MXO3 60 65 65 73 84 12 40 40 48 65 64 77 77 86 100 82 68MXO4 45 49 58 60 69 3 15 15 14 22 45 54 60 60 70 59 42MXO5 62 67 62 69 86 3 11 12 12 21 62 67 62 69 86 68 72MXO6 39 39 45 45 55 7 35 35 60 63 44 59 63 76 84 69 51MXO7 58 64 63 78 82 2 15 20 12 12 58 64 65 80 82 73 64MXO8 38 39 45 59 65 5 12 14 15 22 38 46 50 59 67 54 17MXO9 68 72 88 94 95 1 2 5 4 6 68 72 88 95 95 88 55MFO 90 100 100 100 100 0 0 0 1 2 90 100 100 100 100 99 87MRA1 0 0 0 1 0 2 22 29 23 26 2 22 29 23 26 23 5MRA2 12 20 21 19 20 3 12 15 22 25 14 28 31 37 38 33 3

Kunale KXY1 27 23 39 42 43 13 70 92 90 95 37 75 92 90 95 86 26KXY2 10 11 23 21 26 9 60 65 68 76 15 63 70 68 76 67 20KXY3 8 9 12 12 25 15 55 59 62 79 19 55 59 62 79 61 15KXM1 20 25 38 46 55 14 70 74 77 87 31 71 81 87 87 80 34KXM2 20 20 39 45 48 5 55 62 60 73 25 60 74 77 86 73 33KXO1 80 80 95 95 97 5 25 25 25 26 80 80 95 95 97 92 79KXO2 65 67 86 90 95 8 12 19 20 26 65 67 86 90 95 85 62KRA 7 14 15 15 18 5 52 65 70 77 12 52 65 70 77 64 2

Adewro AXY1 25 26 41 45 47 5 30 48 45 35 24 45 50 55 52 50 20AXY2 14 25 35 39 40 5 15 36 33 35 18 40 58 58 60 53 19AEU1 10 10 12 15 16 9 32 63 64 69 19 41 63 63 69 58 15AEU2 50 55 65 60 65 9 10 28 47 53 53 55 70 78 84 72 36ARA 8 9 10 8 5 10 55 60 62 60 17 55 60 62 60 58 5

Mean valueover all plots(standard deviation)

35 (24) 38 (25) 46 (27) 49 (28) 54 (29) 6 (4) 29 (21) 35 (24) 38 (25) 43 (27) 38 (22) 56 (17) 65 (17) 69 (18) 74 (19) 65 (17) 36 (25)

Minimum value 0 0 0 1 0 0 0 0 1 2 2 22 29 23 26 23 2Maximum value 90 100 100 100 100 15 70 92 90 95 90 100 100 100 100 99 87

Data collected from May 2003 to April 2004.Dry, June, July, August, September: indicate the period of the year for which the vegetation cover is applicable. Dry comprises the months February–May. The months October–January are not represented as in thisperiod, there is almost no rainfall (see Fig. 7) nor runoff. VEGtot,w is the weighted average total vegetation cover, calculated basing the weights of each month on the fraction of the annual rainfall recorded in thatparticular period.

224

K.Desch

eemae

keretal.

Autho

r's

pers

onal

co

py

depth of the collected runoff water. Daily runoff depths andrunoff coefficients (ratio of runoff depth to rainfall depth)were then calculated. Quality control and precision checksof the daily rainfall and runoff data recorded by the localstudents guaranteed that errors were very small.

Daily rain and runoff depths were recorded for each plotover two rainy seasons from May 2003 until September 2004.In Adewro, the tipping bucket rain gauge failed to recordproperly from August 2004 onwards, so that intensities

could not be calculated for this period in Adewro. Runoffdepths are not available for one plot in the young exclosureof Kunale during August 2004 due to breakdown of therunoff box lid. In Adewro runoff data for June 2004 were lostfor three plots.

For each plot, possible explanatory variables for runoffgeneration were recorded:

• Slope gradient.

Figure 4 Runoff plots (5 m · 2 m) in different situations: young (MXY1) and old (MXO3) exclosure in May Ba’ati and grazing land(ARA) and Eucalyptus forest (AEU1) in Adewro.

Runoff on slopes with restoring vegetation: A case study from the Tigray highlands, Ethiopia 225

Autho

r's

pers

onal

co

py

• Percentage vegetative soil cover, measured monthlyfrom May 2003 to April 2004 along a tape meter andsplit into tree and shrub canopy cover (VEGst) and

grass and herb cover (VEGgh). To take into accountseasonal changes in canopy density a visual estimateof cover percentage was made monthly for eachcrown. This means that leafless shrubs with densebranch system still have some soil cover, as theircrown cover can be estimated at for example 20%.The crown cover of deciduous species will stronglyfluctuate over the year, while for evergreen species

Figure 5 Illustration of a runoff plot (5 m · 2 m) in thedegraded grazing land of the May Ba’ati study site (plot codeMRA1).

M J J A S O N D J F M A

% v

eget

ativ

e so

il co

ver

0

20

40

60

80

100

MXO1 - VEGstMXO1 - VEGghMXO9 - VEGstMXO9 - VEGghMXY2 - VEGstMXY2 - VEGghMRA2 - VEGstMRA2 - VEGghMFO - VEGstMFO - VEGgh

2003 2004

Figure 6 Evolution of shrub and tree canopy cover (VEGst)and grass and herb cover (VEGgh) of five plots, representativeof different land use types present in the May Ba’ati study site(grazing land (MRA2), young exclosure (MXY2), old exclosure(MXO1, MXO9) and church forest (MFO)).

May Ba'ati

M J J A S O N D J F M A M J J A S

prec

ipita

tion

(mm

)

0

50

100

150

200

250

2003 2004

Kunale

M J J A S O N D J F M A M J J A S

prec

ipita

tion

(mm

)

0

20

40

60

80

100

120

140

160

180

2003 2004

Adewro

M J J A S O N D J F M A M J J A S

prec

ipita

tion

(mm

)

0

50

100

150

200

250

300

2003 2004

Figure 7 Rainfall in May Ba’ati, Kunale and Adewro during theobservation period (May 2003–September 2004).

226 K. Descheemaeker et al.

Autho

r's

pers

onal

co

py

these fluctuations will be more subtle. With respect tothe grass and herb layer, standing dead grasses arealso included in the vegetative soil cover, as theycan still intercept rainfall and offer resistance to run-off. To calculate total vegetative cover (VEGtot), thecover percentages of canopies overlapping each otherand/or the understorey vegetation were added withoutexceeding the maximum of 100%. To come up with onevalue of vegetation cover per plot, a weighted averagevegetation cover was calculated for each category(VEGst,w, VEGgh,w, VEGtot,w). The weighing factorsgiven to each monthly vegetation cover estimate arebased on the fraction of the annual rainfall recordedin that particular period. This procedure is based onthe reasoning that the vegetation cover prevailing ina month with high rainfall is more important in reduc-ing runoff than the vegetation cover in a month withnegligible rainfall.

• Percentage litter cover, visually determined once in therainy season.

• Topsoil (upper 15 cm) variables:– Texture, by the sieve-pipette method (Sheldrik and

Wang, 1993).– Saturated hydraulic conductivity (Ksat), using the

inversed auger hole method (Kessler and Oosterbaan,1974).

– Soil organic matter content (SOM), calculated basedon soil organic carbon content. Total C was determinedby the dry combustion method (Tiessen and Moir, 1993)using a VARIO MAX total element analyzer (Elementar,Hanau, Germany). Organic C content was then calcu-lated by subtracting the C content present in CaCO3

from the total C content.– Dry soil bulk density (BDtot), determined using a par-

ticular method for stony soils described in Blake andHartge (1990). Bulk density of the fine earth fraction(BDfe) is subsequently calculated following Poesen andLavee (1994).

– Soil water retention following the procedures descri-bed by Van Reeuwijk (2002). Water holding capacity(WHC) is subsequently calculated as the soil moisturecontent at wilting point subtracted from the moisturecontent at field capacity.

– Rock fragment content (Rm), determined as mass per-centage after sieving and weighing (Childs and Flint,1990).

• Surface stoniness (Rc), determined through measuringthe length of all stones along five line transects of 1 m.

• Soil moisture content on a weekly basis in the year 2004using Time Domain Reflectometry (Topp et al., 1980). Forthis, a TRIME-FM3 device (IMKO, Ettlingen, Germany)with tube access probe was used.

Data analysis

To obtain a detailed understanding of influencing factorsand their role in runoff processes, explanatory variableswere split into those related to the plot on the one handand event variables, related to rainfall and soil water con-tent on the other hand. Relations between runoff depth,runoff coefficient and explanatory variables were investi-gated through correlation and regression analysis. Thenon-parametric Kruskal–Wallis test was used to testwhether runoff coefficients for the land use types are signif-icantly different. Stepwise regression was used to identifythe set of variables that most effectively predicts the totalrunoff coefficient. All statistical analyses were performedwith SPSS 11.0 (SPSS, 2001).

Results and discussion

Plot characteristics

Environmental, soil and vegetation characteristics for all 28experimental plots are summarized in Tables 1 and 2. In theMay Ba’ati study site, Calcisols (FAO et al., 1998) are typicalsoil types for the more degraded areas, whereas Cambisols(FAO et al., 1998) and Phaeozems (FAO et al., 1998) areencountered in the old exclosure. The latter were formedunder influence of vegetation in deposited sediment mate-rial (Descheemaeker et al., 2006). In Kunale, most commonsoil types are Phaeozems and Cambisols, while in Adewro,only Phaeozems were found. Most soils are relatively richin organic matter, which is indicated by the qualifier Humic.This applies to soils containing more than 1% of organic car-bon to a depth of 50 cm (FAO et al., 1998). Skeletic on theother hand refers to soils containing between 40 and90 mass% of rock fragments to a depth of 100 cm (FAOet al., 1998).

Monthly records of soil cover of the different vegetationstrata for the period May 2003–April 2004 provide a detailed

Table 3 Frequency distribution (%) of rainfall intensity calculated in 1 min time intervals in the three study sites for theobservation period May 2003–September 2004

Intensity (mm h�1) May Ba’ati Kunale Adewro

% Cumulative % % Cumulative % % Cumulative %

<1.2 25.3 25.3 28.5 28.5 23.9 23.91.2–6 60.4 85.8 58.9 87.4 62.4 86.36–12 5.7 91.4 5.5 92.9 6.5 92.812–18 2.9 94.4 2.3 95.2 2.7 95.518–30 2.5 96.9 2.5 97.8 2.2 97.830–60 2.2 99.1 1.8 99.6 1.6 99.4>60 0.9 100.0 0.4 100.0 0.6 100.0

Runoff on slopes with restoring vegetation: A case study from the Tigray highlands, Ethiopia 227

Autho

r's

pers

onal

co

py

image of seasonal vegetation cover change for each plot(Table 2). Under influence of climatic conditions, the veg-etation goes through an annual cycle, during which thecover increases substantially in the rainy season (Fig. 6).In the dry season, grass and herb cover is on average 15%of the cover in the rainy season, while for shrubs and treesless seasonal fluctuation is noticed as dry season cover isstill 45% of the wet season cover. Analyzing Table 2 andFig. 6 further shows that plots with high shrub and treecover have low grass and herb cover, while plots in grazingland and young exclosures are characterized by highergrass and herb cover, as shrubs and trees are not (yet)capable of suppressing understorey vegetation.

Rainfall

Rainfall in the study region follows a unimodal rainfall pat-tern with a dry period from October to May and a concen-tration of rainfall in the months June–September (Nyssenet al., 2005). Rainfall depth in the observation period fromMay 2003 to September 2004 was highest in Adewro(965 mm). May Ba’ati and Kunale received 892 and774 mm of rain respectively during these 17 months(Fig. 7). Except for August 2004, monthly rainfall depthsin the recorded period are well below the average monthlyrainfall depth presented by Nyssen et al. (2005), so that itis concluded that the observation period was relativelydry.

Based on the tipping bucket data it is concluded thatrainfall intensity during a storm is quite variable. Rainintensity is only high during a small part of a storm and,as also noticed by Nyssen et al. (2005), this high intensityperiod can occur either at the beginning, in the middle orat the end of a storm. Overall, rainfall intensity calcu-lated in 1 min time intervals is low in the three studysites, as about 60% of all rain fell with an intensity be-tween 1.2 and 6 mm h�1. More than 90% of the rain hasan intensity less than 12 mm h�1 and high intensity rainfallis rare as less than 1% of the recorded rainfall depth hasan intensity over 60 mm h�1 (Table 3). This adds evidenceto the statement of Pilgrim et al. (1988) that convectivehigh intensity rainstorms are not typical in semi-aridareas.

Nyssen et al. (2005) classified the months October toMay as dry, June and September as rainy with moderateconcentration of big rains and July and August as rainy withvery high concentration of big rains according to the clas-sification scheme proposed by Daniel Gamachu (1977).Characteristics of the storm events were determined forthese different groups of months (Table 4). Average stormduration varies from roughly half an hour in the dry seasonto about 50 min in July and August. The biggest rainfallevent was recorded in Adewro (44 mm), while biggeststorms in May Ba’ati and Kunale produced 35 mm and26 mm, respectively. The highest rainfall intensity over aperiod of 30 min (I30) was recorded in May Ba’ati(49 mm h�1), but also in Kunale and Adewro maximum val-ues for this variable were over 40 mm h�1. Average valuesfor the mean intensity of a storm are below 4 mm h�1 in allsites and average values for I30 are below 10 mm h�1, indi-cating again the relatively low rainfall intensity of storms

Table

4Storm

charac

teristicsfordry

(October–

May)

andrainymonths(June+Se

ptemberan

dJu

ly+Augu

st)forthethreestudysites

Site

Period

Numberof

storm

sDuration(m

in)

Depth

(mm)

Mean

intensity

(mm

h�1)

I30(m

mh�1)

Ave

rage

(stdev)

Ran

geAve

rage

(stdev)

Ran

geAve

rage

(stdev)

Ran

geAve

rage

(stdev)

Ran

ge

MayBa’ati

October–

May

61

28.9(43.2)

10–252

1.9(4.6)

0.2–

26.9

2.5(3.4)

0.6–

19.3

9.1(10.5)

0.7–

33.2

June+Se

ptember

8531.3(29.2)

10–139

2.5(4.4)

0.2–

22.9

3.5(4.2)

0.7–

17.8

9.9(11.3)

0.7–

43.7

July

+Augu

st17

551.3(65.1)

10–405

3.3(6.1)

0.2–

34.6

3.3(4.5)

0.4–

32.1

8.9(10.8)

0.6–

49.1

Kunale

October–

May

51

36.9(53.5)

10–261

1.4(2.4)

0.2–

13.2

2.4(2.9)

0.4–

18.3

4.0(3.7)

0.6–

13.4

June+Se

ptember

8338.8(33.5)

10–140

3.0(4.5)

0.2–

23.9

3.5(4.3)

0.4–

29.9

9.8(9.6)

0.5–

42.6

July

+Augu

st18

050.0(73.0)

10–509

2.6(4.3)

0.2–

26.4

2.9(3.8)

0.4–

26.3

6.8(7.0)

0.6–

29.9

Adewro

October–

May

67

30.7(43.5)

10–260

1.6(2.9)

0.2–

152.8(3.0)

0.6–

16.6

6.3(6.0)

1.0–

17.7

June+Se

ptember

7744.0(41.6)

10–177

3.0(4.4)

0.2–

21.8

3.4(3.3)

0.7–

14.1

8.3(9.1)

0.7–

40.7

July

+Augu

st15

649.5(65.5)

10–428

2.6(5.3)

0.2–

442.9(4.7)

0.4–

32.2

6.2(8.2)

0.6–

43.5

I30:

max

imal

30min

rainfallintensity.

228 K. Descheemaeker et al.

Autho

r's

pers

onal

co

py

Table 5 Monthly runoff depth (R, mm) and runoff coefficient (RCm, %) for all plots during different periods of the year

Site Plot Jun-03 Jun-04 Jul-03 Jul-04 Aug-03 Aug-04 Sep-03 Sep-04 Oct03–May04

Rain (mm) 50 68.7 93.6 117.8 150.2 215.3 63.5 26.8 64.5

R RCm R RCm R RCm R RCm R RCm R RCm R RCm R RCm R RCm

May Ba’ati MXY1 6.5 13.0 6.0 8.7 34.7 37.1 15.5 13.1 24.9 16.6 56.0 26.0 8.3 13.0 1.5 5.5 0.5 0.7MXY2 13.2 26.4 11.7 17.1 32.2 34.4 38.1 32.3 32.1 21.4 50.5 23.4 12.7 20.0 1.5 5.7 1.7 2.6MXY3 6.8 13.6 14.6 21.3 27.6 29.5 34.4 29.2 26.5 17.7 56.1 26.1 17.6 27.7 1.8 6.7 3.6 5.6MXO1 4.8 9.5 3.5 5.1 8.4 9.0 10.2 8.6 6.6 4.4 12.7 5.9 2.9 4.6 0.5 1.7 2.1 3.3MXO2 4.8 9.7 2.2 3.2 5.5 5.9 8.4 7.1 11.3 7.5 9.5 4.4 2.9 4.6 0.4 1.4 1.8 2.7MXO3 0.0 0.0 0.0 0.1 0.4 0.4 0.9 0.7 1.8 1.2 0.6 0.3 4.2 6.6 0.0 0.0 0.0 0.0MXO4 0.8 1.6 4.2 6.1 1.2 1.3 5.7 4.9 2.3 1.5 8.4 3.9 1.7 2.6 1.1 4.0 1.0 1.5MXO5 0.0 0.0 0.5 0.7 0.0 0.0 1.6 1.3 0.8 0.6 1.7 0.8 1.0 1.6 0.5 2.1 0.0 0.0MXO6 0.8 1.6 0.1 0.2 5.2 0.6 0.6 0.5 0.9 0.6 1.1 0.5 0.5 0.8 0.0 0.0 0.4 0.6MXO7 0.1 0.3 0.1 0.2 2.0 2.1 2.1 1.8 0.1 0.1 0.9 0.4 0.5 0.8 0.0 0.0 0.0 0.0MXO8 1.5 2.9 2.0 2.9 0.5 0.5 4.4 3.7 3.8 2.5 12.9 6.0 2.1 3.3 1.5 5.7 0.6 0.9MXO9 0.2 0.3 0.2 0.3 1.2 1.2 2.3 2.0 0.9 0.6 0.9 0.4 0.3 0.5 0.2 0.6 0.2 0.3MFO 0.0 0.1 0.0 0.0 0.1 0.1 0.4 0.4 0.1 0.1 0.6 0.3 0.1 0.2 0.0 0.0 0.1 0.1MRA1 26.3 52.6 40.1 58.3 47.0 50.2 80.1 68.0 81.0 53.9 122.2 56.8 40.7 64.1 11.9 44.3 23.3 36.2MRA2 21.5 42.9 27.9 40.6 32.1 34.3 66.3 56.3 61.9 41.2 110.5 51.3 27.2 42.9 5.9 21.8 12.6 19.5

Rain (mm) 41.0 71.7 110.4 77.0 108.6 170.9 86.3 44.3 37.0

R RCm R RCm R RCm R RCm R RCm R RCm R RCm R RCm R RCm

Kunale KXY1 0.7 1.7 0.8 1.0 2.7 2.5 0.7 0.9 2.6 2.4 1.3 0.8 10.9 12.7 0.4 0.8 0.0 0.0KXY2 0.8 1.8 0.2 0.3 8.3 7.5 0.3 0.4 7.5 6.9 n.a. n.a. 16.5 19.1 0.0 0.0 0.0 0.0KXY3 0.3 0.7 0.2 0.3 4.4 4.0 1.4 1.8 35.0 32.3 12.3 7.2 40.8 47.2 0.4 0.9 0.0 0.0KXM1 0.2 0.6 0.0 0.0 0.4 0.4 0.1 0.1 1.2 1.1 0.6 0.3 3.8 4.4 0.1 0.1 0.0 0.0KXM2 0.0 0.1 0.2 0.2 0.6 0.5 0.2 0.3 1.0 1.0 0.5 0.3 2.0 2.3 0.3 0.6 0.0 0.0KXO1 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0KXO2 0.0 0.0 0.0 0.0 1.4 1.3 0.2 0.2 0.0 0.0 0.0 0.0 2.2 2.5 0.0 0.0 0.0 0.0KRA 1.3 3.2 2.0 2.8 7.3 6.6 2.1 2.7 34.4 31.7 27.6 16.2 26.1 30.2 2.2 4.9 0.0 0.0

Rain (mm) 83.1 77.3 133.5 102.4 152.4 288.4 72.8 13.1 50.6

R RCm R RCm R RCm R RCm R RCm R RCm R RCm R RCm R RCm

Adewro AXY1 2.2 2.6 n.a. n.a. 4.2 3.2 3.7 3.7 6.4 4.2 16.1 5.6 3.0 4.1 0 0.0 1.6 3.2AXY2 0.2 0.2 n.a. n.a. 9.5 7.2 5.4 5.3 9.2 6.0 48.5 16.8 3.8 5.3 0 0.0 1.7 3.3AEU1 25.5 30.7 10.0 13.0 25.0 18.7 5.7 5.5 6.5 4.3 31.8 11.0 2.7 3.7 0 0.0 2.9 5.8AEU2 7.9 9.5 2.5 3.3 6.6 4.9 2.5 2.5 4.7 3.1 17.5 6.1 0.0 0.0 0 0.0 0.3 0.5ARA 24.9 29.9 n.a. n.a. 17.8 13.4 30.8 30.1 32.2 21.1 135.9 47.1 29.6 40.7 0 0.0 4.3 8.5

For each site and period rainfall depth (mm) is also indicated. (n.a.: no data available).

Runoff

onslo

peswith

resto

ringve

getatio

n:Aca

sestu

dyfro

mtheTigray

high

lands,

Ethiopia

229

Autho

r's

pers

onal

co

py

in the study region. During the peak of the rainy season, tak-ing place in the months July and August, higher rainfallamounts and longer lasting storms are recorded as com-pared to June and September (Table 4).

Surface runoff

Field observations reveal that runoff in our study sites ismainly Hortonian. Runoff response to rainfall is quick andno indications of saturated runoff were noted.

Table 5 gives an overview of monthly rainfall depth,runoff depths and runoff coefficients for all plots. Runoffamounts and coefficients vary considerably between plots,but also among sites and recorded periods. However, fromthe data presented in Table 5, it can already be deducedthat degraded areas and young exclosures generally pro-duce more runoff than older exclosures. It further seemsnot unlikely that an effect of rainfall and moisture condi-tion is present, as the monthly runoff coefficients differconsiderably between periods. The large difference be-tween plots is also evident from Fig. 8, illustrating thatcumulative runoff depths over two rainy seasons vary fromless than 2 mm in the church forest of May Ba’ati to almost400 mm in a plot in the degraded grazing land in the samestudy site. The cumulative runoff curves mainly rise in theperiod between July and September, as rainfall eventsoccurring in the dry period generally do not produce sub-stantial runoff. The latter is also deduced from the resultsin Table 5, indicating that the monthly runoff coefficientsare smaller in the dry period as compared to the othermonths. On the one hand this can be explained by the factthat storms in the dry period are on average shorter, lessintense and producing less rainfall as compared to stormsin the rainy season (Table 4). On the other hand, dry soilshave higher infiltration capacity than wet soils in the rainyseason.

Rainfall thresholds for runoff generation were deter-mined for each plot making use of graphs plotting daily run-off depth against daily rainfall depth (Fig. 9). The rainfallthreshold is the rainfall depth for which the probability tohave no runoff becomes very small. The slope of the rain-fall–runoff curve beyond the threshold value (Fig. 9) wasdetermined through least squares curve fitting. This slopegives an indication to what degree runoff depth increaseswith increasing rainfall depth once the rainfall threshold isexceeded. The higher the rainfall threshold and the lowerthe slope of the curve, the higher the absorption capacityof the ecosystem. Table 6 gives an overview of the rainfallthreshold and the slope of the rainfall–runoff curve for allplots. The smallest rainfall threshold values are found inthe grazing lands of the May Ba’ati study site, where a rain-fall event of about 3 mm causes runoff in most cases. Thelargest threshold values on the other hand indicate that insome places in the old exclosure of May Ba’ati a storm eventproducing more than 20 mm of rain is needed for runoff tooccur. The slope of the rainfall–runoff curve also widelyvaries among plots: young exclosures and degraded grazinglands yield slope values, which generally lie between 0.1and 0.8, while values for older exclosures are smaller than0.1. It was also investigated whether soil moisture conditionhad any influence on the rainfall threshold, but this did notseem the case.

Rainfall threshold values for (semi)arid regions found inthe literature typically lie in the same range, varying from3 to 16 mm of rain (Cordery et al., 1983, cited in Pilgrimet al., 1988; Romero Dıaz et al., 1988; Karnieli and Ben-Asher, 1993; Martınez-Mena et al., 1998; Desir, 2002). Ascompared to humid areas rainfall threshold values for(semi-)arid areas are low, caused by rapid time to ponding(Pilgrim et al., 1988). This illustrates the lower interceptioncapacity of semi-arid canopies and the lower infiltrationcapacity of soils in drier environments. The latter is relatedto the common presence of a soil crust and a usually sparservegetation cover, leading to lower organic matter contentand less structure in soils.

Effect of exclosure on runoff

For each site separately, differences in runoff generationbetween land use types were tested for their significance.For this, all plots located in the same land use were groupedand only rainfall events that caused runoff in at least oneplot were considered. The non-parametric Kruskal–Wallis

A M J J A S O N D J F M A M J J A S O N D

cum

ulat

ive

runo

ff (m

m)

0

100

200

300

400

cum

ulat

ive

runo

ff (m

m)

0

2

4

6

8

MRA2

MXY2

cum

ulat

ive

runo

ff (m

m)

0

10

20

30

40

50

60 MXO1

MXO9

MFO

cum

ulat

ive

rain

fall

(mm

)

0

200

400

600

800

2003 2004

Figure 8 Cumulative rainfall depth and runoff depth for fiveplots, representative of different land use types present in theMay Ba’ati study site (grazing land (MRA2), young exclosure(MXY2), old exclosure (MXO1, MXO9) and church forest (MFO)).Cumulative runoff is presented in separate graphs because ofscale differences.

230 K. Descheemaeker et al.

Autho

r's

pers

onal

co

py

test revealed that for the sites Kunale and May Ba’ati, allland use types have significantly different runoff coeffi-cients (Table 7). In Adewro however, the grazing land had

significantly higher runoff coefficients than the young exclo-sure and the Eucalyptus forest, but the last ones did not dif-fer among each other.

MXO1

Rainfall (mm)

0 5 10 15 20 25 30 35 40

Run

off (

mm

)

0

1

2

3

4

5

6MRA2

Rainfall (mm)

0 5 10 15 20 25 30 35 40 45

Run

off (

mm

)

0

5

10

15

20

25

30

Figure 9 Illustration of the determination of the rainfall threshold (arrow) and the slope of the rainfall–runoff curve for two plots(MXO1, old exclosure and MRA2, grazing land) in the May Ba’ati study site.

Table 6 Rainfall threshold (T, mm) and slope of the rainfall–runoff curve (slopePR, mm runoff/mm rainfall) for each plot in thethree study sites (n.a.: not applicable as for this plot, no runoff event was recorded) with indication of R2 of the regressionfunction described by T and slopePR, significance level (p) and number of observations (n)

Site Plot code T slopePR R2 p n

May Ba’ati MXY1 5 0.388 0.71 0.000 182MXY2 6 0.571 0.82 0.000 182MXY3 4 0.439 0.83 0.000 182MXO1 7 0.116 0.61 0.000 182MXO2 9 0.132 0.65 0.000 182MXO3 27 0.077 0.12 0.000 182MXO4 7 0.062 0.55 0.000 182MXO5 12 0.019 0.39 0.000 182MXO6 9 0.011 0.26 0.000 182MXO7 16 0.024 0.17 0.000 182MXO8 9 0.096 0.72 0.000 182MXO9 15 0.022 0.20 0.000 182MRA1 3 0.808 0.93 0.000 182MRA2 3 0.691 0.90 0.000 182MFO 16 0.006 0.43 0.000 182

Kunale KXY1 8 0.13 0.36 0.000 176KXY2 8 0.28 0.61 0.000 147KXY3 6 0.43 0.38 0.000 176KXM1 9 0.044 0.40 0.000 176KXM2 10 0.033 0.52 0.000 176KXO1 n.a. n.a. 0.000 176KXO2 17 0.17 0.55 0.000 176KRA 4 0.29 0.43 0.000 176

Adewro AXY1 7 0.063 0.67 0.000 162AXY2 7 0.17 0.37 0.000 162AEU1 7 0.24 0.58 0.000 181AEU2 8 0.096 0.62 0.000 181ARA 7 0.52 0.60 0.000 162

Runoff on slopes with restoring vegetation: A case study from the Tigray highlands, Ethiopia 231

Autho

r's

pers

onal

co

py

The results in Table 7 show that in all sites, closing a de-graded area for grazing leads to a significant decrease inrunoff generation and this decrease continues as closure ex-tends over time. The Eucalyptus plantation however is anexception to this rule, as despite its age of about 20 years,its runoff coefficient is even slightly higher than the one inthe young exclosure, closed for only 5 years. In the litera-ture different explanations are given for the higher runoffcoefficients of Eucalyptus forests compared to certain otherland use types. Zhou et al. (2002) attribute the higher runoffcoefficient under Eucalyptus trees to a lower litter coverand less understorey vegetation compared to mixed forests.On the other hand, Sorriso-Valvo et al. (1995) blame high lit-ter cover of the broad eucalyptus leaves for causing imme-diate runoff by effectively shedding water as runoff and,more indirectly, for suppressing grass vegetation, therebyinhibiting an increased infiltration rate. Kosmas et al.(1997) and Vacca et al. (2000) also describe the favorableconditions for runoff and erosion under Eucalyptus forestas related to prevailing bare soil conditions due to a de-creased understorey vegetation. For the same Eucalyptusforest in Adewro Descheemaeker et al. (2006) found a lim-ited understorey vegetation, which was inversely correlatedwith Eucalyptus tree canopy cover. This corroborates thefindings of Fiedler and Gebeyehu (1988) that Eucalyptus for-ests in Ethiopia in areas with rainfall less than 750 mm havea weakly developed understorey and forest floor which mayenhance runoff and erosion. The inverse relation betweenundergrowth and Eucalyptus canopy cover is widely dis-cussed in literature (Poore and Fries, 1985) and often attrib-uted to competition for water (mainly in drier areas) andallelopathic effects of Eucalyptus litter and roots. Also inthis study, the four plots in Adewro representing the Euca-lyptus forest on the one hand and the young exclosure onthe other hand, do not present such differences in vegeta-tion and litter cover (Table 2) as would be expected fromtheir difference in establishment age (almost 20 versus 5).Besides vegetative soil cover, also the interception capacityof the vegetation can explain differences in runoff and an-other explanation might be that the Eucalyptus forest inAdewro is more accessible for occasional intrusions of live-stock, causing soil trampling and soil structure loss with in-creased runoff as a consequence.

Comparing our results with findings elsewhere is notstraightforward as the conditions influencing runoff genera-tion (vegetation, soil, precipitation regime, etc.) often varybetween the different studies. Experimental plot dimen-sions are also important determinants, as runoff amount isstrongly influenced by scale effects (Bergkamp, 1998). How-ever, when considering studies in semi-arid regions whererunoff collection has been conducted in medium-sizedplots, some comparison is possible.

In an experimental site near Addis Ababa, Mwenderaet al. (1997) found runoff coefficients for 13 mm rainfallevents, which ranged from 36% to 60% and 39% to 72% formedium and heavy grazing intensities, respectively. Themeasurements were conducted in 10 m · 10 m plots onslopes ranging from 2% to 7%. When only considering stormevents yielding between 10 and 20 mm rainfall, we foundaverage runoff coefficients of 54% in May Ba’ati, 17% in Ku-nale and 32% in Adewro for their respective grazing lands.Our plots have a much higher slope gradient (20 up to50%, Table 1) and soil types (see Table 1) are clearly differ-ent from the Vertisols at Debre Zeit (Mwendera et al.,1997). In their overview of hydrological data obtained onlimestone in Mediterranean environments, Calvo-Caseset al. (2003) cite the study of Diamantopoulos et al.(1996) who measured a runoff coefficient of 80% in a2 m · 10 m runoff plot in grazed shrubland for a rainfallevent of 30 mm. Considering the two plots located in graz-ing land in limestone area in May Ba’ati, six storm events ofbetween 25 and 35 mm rain caused high runoff amountsresulting in an average runoff coefficient of 77%. In contrastwith these strongly similar results, Lavee et al. (1998) founda very low runoff coefficient of 0.7% for grazing land in a re-gion in Israel with an average annual rainfall depth of620 mm. These measurements were done on 3 m · 21 mplots.

Considering runoff measurements in medium-sized,bounded runoff plots in non-grazed shrublands, runoff coef-ficients are generally found to be lower than 5% (Bergkamp,1998; Lopez-Bermudez et al., 1998; Romero Dıaz et al.,1999; Puigdefabregas et al., 1999). Except for the youngexclosure in May Ba’ati these findings from the Mediterra-nean region are in accordance with our results from variousexclosure sites in Tigray (Table 5).

Table 7 Average daily runoff coefficients (%) for different land use types in the three study sites

Land use type May Ba’ati Kunale Adewro

Average n* Average n* Average n*

Degraded rangeland 34.8 a 195 11.8 a 91 11.4 a 162Young exclosure 13.4 b 294 4.7 b 254 2.5 b 324Middle-aged exclosure 0.6 c 182Old exclosure 1.7 c 882 0.1 d 182Eucalyptus forest 3.2 b 367Church forest 0.1 d 98

Within one site, land use types with no letters in common have significantly different runoff coefficients at the 0.05 level (based on theKruskal–Wallis test).

* Different values for n result from different numbers of plots in the land use types and from gaps in data series. Only rainfall eventscausing runoff in at least one plot of the study site were considered.

232 K. Descheemaeker et al.

Autho

r's

pers

onal

co

py

For Eucalyptus forests in the Mediterranean region Kos-mas et al. (1997) found runoff coefficients ranging between0.6% and 8.2%, based on medium-sized runoff plot measure-ments. The runoff coefficient of 3.2% we found for the Euca-lyptus plantation in Adewro lies within this range and isfurthermore highly similar to the coefficient of 3% thatwas found by Sorriso-Valvo et al. (1995) for Eucalyptus for-est with high understorey grass cover in Italy.

Variables influencing runoff

Event variables

Correlations between event variables and runoff amountwere determined for all rainfall events and for each plotseparately (Table 8). Rainfall amount and intensity are inall plots more or less equally and significantly correlatedto runoff amount and Spearman’s rank correlation coeffi-

cients generally range between values of 0.6 and 0.8 (Table8). Both storm duration and soil moisture content are aswell significantly correlated to runoff amount in most plots,but in all cases less pronounced than rainfall amount andintensity. This finding indicates that runoff is producedthrough infiltration-excess or Hortonian runoff generationmechanisms. Several authors reported that Hortonian runoffgeneration is generally prevailing in semi-arid areas (e.g.Dunne, 1978; Pilgrim et al., 1988; Martınez-Mena et al.,1998; Kang et al., 2001). The strong correlations betweenrunoff and rainfall characteristics are not surprising as manyauthors demonstrate similar relationships (e.g. Sala, 1988;Northcliff et al., 1990; Lopez-Bermudez et al., 1998; Martı-nez-Mena et al., 1998; Kang et al., 2001; Descroix et al.,2001; Desir, 2002; Dunjo et al., 2004). Others (e.g. Le Biss-onnais et al., 1995; Gutierrez and Hernandez, 1996; Peugeotet al., 1997; Desir, 2002) also stress the importance of soil

Table 8 Correlation coefficients (Spearman’s q) per plot between daily runoff depth (mm) and daily precipitation (prec) (mm),daily maximal 30 min intensity (I30) (mm h�1), daily storm duration (min) and weekly soil moisture content (%)

Prec I30 Duration Soil moisture

May Ba’ati MXY1 0.79** 0.78** 0.58** 0.36**

MXY2 0.81** 0.80** 0.57** 0.40**

MXY3 0.83** 0.82** 0.60** 0.36**

MXO1 0.79** 0.79** 0.55** 0.33**

MXO2 0.81** 0.81** 0.60** 0.28**

MXO3 0.46** 0.74** 0.50** 0.20*

MXO4 0.75** 0.74** 0.50** 0.20*

MXO5 0.58** 0.55** 0.46** 0.20*

MXO6 0.65** 0.65** 0.47** 0.06MXO7 0.55** 0.55** 0.40** 0.29**

MXO8 0.73** 0.73** 0.50** 0.34**

MXO9 0.74** 0.70** 0.62** 0.27**

MFO 0.61** 0.63** 0.41** 0.43**

MRA1 0.87** 0.85** 0.63** 0.25**

MRA2 0.84** 0.83** 0.59** 0.30**

n 182 182 182 105

Kunale KXY1 0.73** 0.72** 0.59** 0.24*

KXY2 0.62** 0.61** 0.46** 0.11KXY3 0.72** 0.66** 0.63** 0.45**

KXM1 0.61** 0.60** 0.45** 0.37**

KXM2 0.62** 0.62** 0.43* 0.28**

KXO1 n.a. n.a. n.a. n.a.KXO2 0.38* 0.34** 0.26** 0.00KRA 0.79** 0.76** 0.64** 0.41**

n 176 (147) 176 (147) 176 (147) 95 (66)

Adewro AXY1 0.74** 0.70** 0.48** 0.44**

AXY2 0.75** 0.68** 0.53** 0.45**

AEU1 0.72** 0.60** 0.51** 0.19AEU2 0.23** 0.16 0.13 0.33**

ARA 0.76** 0.70** 0.59** 0.31**

n 181 (162) 152 (133) 152 (133) 97 (79)

For each site, the number of observations (n) is given after the list of plots (numbers between brackets refer to plot KXY2 for Kunale andto AXY1, AXY2 and ARA for Adewro).n.a.: not applicable; as for this plot no runoff event was recorded.

* Significant at the 0.05 level (2-tailed).** Significant at the 0.01 level (2-tailed).

Runoff on slopes with restoring vegetation: A case study from the Tigray highlands, Ethiopia 233

Autho

r's

pers

onal

co

py

moisture content, showing a positive correlation with runoffamount.

Plot variables

To offset effects of rainfall in analyzing the relation be-tween runoff production and plot variables, the runoff coef-ficient was considered. Two types of runoff coefficient areused in this analysis: firstly the monthly runoff coefficient(RCm), calculated as the ratio of monthly runoff depth tomonthly rainfall depth and secondly the total runoff coeffi-cient (RCtot) as the ratio of total runoff depth to total rain-fall depth for the observation period.

Special attention is paid to the effects of vegetationcover on runoff generation as it is believed to be an impor-tant factor influencing runoff production.

The RCtot was related to the weighted vegetation covervariables. This revealed significant correlations withVEGst,w (r = �0.643) and VEGtot,w (r = �0.766) and no cor-relation with VEGgh,w. Refining the analysis by consideringmonthly data, correlation analysis revealed that when usingthe total set of monthly data, the RCm is significantly corre-lated with shrub and tree cover (VEGst,m) (r = �0.626) andtotal vegetation cover (VEGtot,m) (r = �0.640) and not withgrass and herb cover (VEGgh,m). However, dealing witheach plot separately to investigate whether vegetationgrowth and increasing soil cover during the rainy seasoncan significantly attenuate runoff production, no or evenpositive correlations were found between the runoff coeffi-cients and the vegetation cover variables. This can be ex-plained by the fact that denser vegetation covers arereached near the peak of the rainy season, which coincideswith wetter soil moisture conditions and bigger storms.Apparently these event variables exert a more importantinfluence on runoff production than vegetation cover. Thefact that refining the analysis by considering monthly datadid not yield higher correlations is also due to opposite influ-ences exerted on the one hand by the vegetation cover vari-ables and on the other hand by the event variables.

Another striking and unexpected result is the fact thatgrass and herb cover was not found to attenuate runoff pro-duction, which is in contradiction with a large body of evi-dence (e.g. Abrahams et al., 1995; Fiener and Auerswald,2003) stating that grasses are very effective in controllingrunoff. Grasses increase surface roughness, which promotesinfiltration. Hence the widespread use of grass strips in con-trolling runoff (Van Dijk et al., 1996). However, the expla-nation for our case is simply found by comparing plots:the more degraded plots (such as the ones in young exclo-sures and grazing lands) have higher grass cover (Table 2),but they are also characterized by higher runoff as com-pared to plots located under a restored vegetation cover(Tables 5 and 7). Besides other factors, this is related to dif-ferences in the vegetation’s capacity to intercept, slowdown and attenuate the impact of raindrops. It is logicalthat a relatively low grass and herb layer does not havethe same capacity as a well developed shrub and tree layer.Therefore, in order to analyze the effect of grass and herbcover the effect of tree and shrub cover has to be elimi-nated first. To achieve this, plots with VEGst,w < 40% werefocused at in a correlation analysis, which revealed a signif-icant negative relation between VEGgh,w and RCtot

(r = �0.652). This means that in situations were there isno dominant shrub and tree canopy cover, grasses and herbsare found to significantly temper runoff production.

Table 9 presents the results of a correlation analysis be-tween the RCtot and various plot variables. The weightedvegetation cover variables VEGst,w and VEGtot,w are moststrongly correlated with RCtot, closely followed by a thirdvariable related to vegetation, namely litter cover. Soil re-lated variables such as bulk density and soil organic mattercontent are also significantly correlated to RCtot, as well assurface stoniness. From Table 9 it is also clear that some ofthese plot variables are intercorrelated, so that unequivocaljudgements about their influence on runoff production cannot be made at this stage.

The relation between total vegetation cover and theRCtot is best described by an exponential decay function,which explains almost 80% of the variation (Fig. 10). Thisbears resemblance to the result obtained by Francis andThornes (1990) stating that runoff exponentially increaseswith decreasing vegetation cover. Based on the scatterplotin Fig. 10, four vegetation cover classes can be distin-guished: class 1 (VEGtot,w 6 40%); class 2 (40% < VEG-tot,w 6 65%); class 3 (65% < VEGtot,w 6 80%) and class 4(VEGtot,w > 80%). A Kruskal Wallis multiple comparison re-vealed that class 3 and 4 are significantly different concern-ing their average RCtot from class 1 and class 2. Class 3 and4 do not significantly differ from one another and the sameapplies to classes 1 and 2, although the latter is probablydue to the limited amount of observations for class 1. Somestudies mention a vegetation threshold above which runoffis totally controlled by vegetation. Based on the scatter inFig. 10 and the results of the Kruskal Wallis multiple com-parison test, it is confirmed that plots with a VEGtot,w be-low 65% have a significantly higher RCtot (17.2% on average)than plots with a denser vegetation (1.3% on average).Therefore, it can be concluded that for steep hillslopes inTigray, the vegetation threshold value lies around 65%.Once total vegetative soil cover surpasses this threshold,runoff becomes almost negligible. Three references, whosethresholds vary from 50% (Gifford, 1985), over 60% (Orr,1970) to 70% vegetative cover (Lang, 1979) are cited by Gut-ierrez and Hernandez (1996). The latter differentiate be-tween the growing and the dormant season to put forwardthreshold values of 50% and 75%, respectively. While North-cliff et al. (1990) recognized a lower vegetation cover of30% as a major threshold in runoff and erosion, still othersstate values of 55% (Snelder and Bryan, 1995) and 60% (De-sir, 2002), so that it can be concluded that our vegetationthreshold value of 65% is within the commonly adoptedrange.

From Table 9 it is clear that various plot variables arecorrelated with RCtot, so that an attempt was made to im-prove the predictive power of the regression using only VEG-tot,w as explanatory variable by including other variablesinto a multiple linear regression. Care was taken not tointroduce problems of multicollinearity in doing so. As wedemonstrated before that runoff production depends onthe vegetation cover class, it was decided to conduct thisexercise for groups of classes in which the most dense veg-etation cover class was omitted consecutively.

The resulting regression functions and a list of variablessignificantly correlated with RCtot are given in Table 10.

234 K. Descheemaeker et al.

Autho

r's

pers

onal

co

py

Table 9 Pearson correlation coefficients for total runoff coefficient, calculated over the 2-year observation period and plot variables for all plots in the three study sites(n = 28)

VEGgh,w VEGst,w VEGtot,w Slope Rc Rm BDfe BDtot Clay Silt Ksat WHC Litter SOM

RCtot �0.040 �0.643** �0.766** �0.234 0.585** 0.036 0.444* 0.534** �0.323 0.363 �0.145 �0.058 �0.619** �0.506**VEGgh,w �0.575** 0.079 0.050 �0.105 �0.050 �0.331 �0.373 �0.236 �0.182 �0.254 �0.155 �0.476** �0.206VEGst,w 0.729** �0.076 �0.565** �0.074 �0.009 �0.076 0.500** �0.192 0.088 0.333 0.905** 0.399*

VEGtot,w �0.078 �0.764** �0.068 �0.325 �0.407* 0.465* �0.366 �0.118 0.324 0.713** 0.363Slope 0.298 0.205 �0.222 �0.161 �0.373 �0.187 0.223 �0.273 �0.296 �0.085Rc 0.220 0.195 0.337 �0.331 0.337 0.047 �0.279 �0.695** �0.374*Rm �0.494** �0.129 �0.459* 0.492** 0.390* �0.391* �0.198 0.198BDfe 0.917** 0.360 �0.062 �0.448* 0.481** 0.051 �0.449*BDtot 0.214 0.146 �0.367 0.373 �0.044 �0.418*Clay �0.186 �0.488** 0.660** 0.598** 0.050Silt �0.037 �0.204 �0.203 0.209Ksat �0.537** �0.002 0.476*

WHC 0.462* �0.113Litter 0.016

RCtot: total runoff coefficient (%); VEGgh,w: weighted grass and herb cover (%); VEGst,w: weighted shrub and tree cover (%); VEGtot,w: weighted total vegetation cover (%); Slope: slopegradient (%); Rc: surface stoniness (%); Rm: stoniness of the topsoil (mass%); BDfe: bulk density of the fine earth fraction (Mg m�3); BDtot: soil bulk density (Mg m�3); Clay: topsoil claycontent (%); Silt: topsoil silt content (%); Ksat: saturated hydraulic conductivity (m d�1); WHC: water holding capacity (%); Litter: litter cover (%); SOM: topsoil organic matter content (%).

* Correlations are significant at the 0.05 level (2-tailed).** Correlation are significant at the 0.01 level (2-tailed).

Runoff

onslo

peswith

resto

ringve

getatio

n:Aca

sestu

dyfro

mtheTigray

high

lands,

Ethiopia

235

Autho

r's

pers

onal

co

py

Several conclusions can be drawn from Table 10. Total veg-etation cover and shrub and tree cover are the most impor-tant explaining variables in all groups of plots.

The positive impact of vegetation cover on runoff reduc-tion is extensively discussed in the literature. This vegeta-tion impact is explained on the one hand by a directeffect as a canopy cover intercepts raindrops, thus dissipat-ing their energy and creating infiltration pathways (Morganet al., 1986; Castillo et al., 1997; Descroix et al., 2001).Moreover, vegetation tempers the water flow velocity, thusincreasing the opportunities for the water to infiltrate (Mor-gan et al., 1986; Bochet et al., 1998). On the other hand andmore indirectly, vegetation positively influences soil physi-cal properties through the incorporation of organic matter,so that infiltration rate is increased (see also below) (Dunneand Dietrich, 1980; Morgan et al., 1986; Bochet et al., 1998;Puigdefabregas et al., 1999).

From the correlation and regression analysis, summa-rized in Tables 9 and 10, it was proven that runoff produc-tion is mainly controlled by vegetation cover. Also severalother authors (Thurow et al., 1986; Northcliff et al., 1990;Bohm and Gerold, 1995; Gutierrez and Hernandez, 1996;Cerda, 1998) emphasize that vegetation cover is the primaryexplanatory variable for runoff production.

The suppressing effect of vegetation cover on runoff pro-duction is also illustrated through its positive linear relationwith the rainfall threshold (Fig. 11). The relation betweenVEGtot,w and the slope of the rainfall–runoff curve is bestdescribed by an exponential decay function and closelyresembles the one in Fig. 10, which is not surprising as theindependent variable is the same and in both cases thedependent value is a ratio of runoff depth over rainfalldepth.

Left aside the dominance of vegetation in controllingrunoff, other variables also play a role, which can be unrav-elled based on Table 10. Although they are significantly cor-related with RCtot some variables are not included in thebest fitting regression function because they do not improvethe fit of the curve or because of multicollinearity prob-

VEGtot,w

0 20 40 60 80 100

RC

tot

0

10

20

30

40

50

60

70

000.0,28

78.0

23.17469.22

)045.0(

===

+−= −

pn

R

ey x

Figure 10 Scatter plot and best fitting regression line relatingthe runoff coefficient determined for the total observationperiod (RCtot) and the weighted average total vegetation cover(VEGtot,w).

Table

10

Ove

rview

ofbest

fittingregressionfunctionsandva

riab

lessign

ifica

ntlyco

rrelatedwithRCtot,

fordifferentgroupsofplots

Plots

included

nBest

fittingregressionfunctionsa

R2b

Correlatedva

riab

lesc

Allplots

28RCtot=165

.4�

37.6

ln(VEGtot,w)

0.72

VEGtot,w;VEGst,w

;litter;

Rc;

BDtot;

SOM;BDfe

28RCtot=14

8�

31.4

ln(VEGtot,w)�

7.4

ln(SOM)

0.77

VEGtot,w<80

%21

RCtot=19

6�

45.6

ln(VEGtot,w)

0.73

VEGtot,w;VEGst,w

;SO

M;litter;

BDtot;

Rc;

BDfe

21RCtot=16

6.1�

35.6

ln(VEGtot,w)�

8.9ln(SOM)

0.79

21RCtot=15

2.7�

30.1

ln(VEGtot,w)�

7.3ln(SOM)�

3.6ln(litter)

0.83

VEGtot,w6

65%

15RCtot=19

4.4�

45.1ln(VEGtot,w)

0.65

VEGtot,w;VEGst,w

;SO

M;BDtot;

slope;WHC;litter;

BDfe

15RCtot=16

6.2�

35.5ln(VEGtot,w)�

9.4ln(SOM)

0.73

15RCtot=18

4�

27.9

ln(VEGtot,w)�

8.9

ln(SOM)�

12.2ln(slope)

0.82

15RCtot=17

4.1�

23.2

ln(VEGtot,w)�

7.6ln(SOM)�

13.4

ln(slope)�

0.2litter

0.87

RCtot:

totalrunoff

coefficient(%);

VEGst,w

:weightedshruban

dtreeco

ver(%);

VEGtot,w:weightedtotalve

getationco

ver(%);

slope:slopegrad

ient(%);

Rc:

surface

stoniness

(%);

BDfe:

bulk

density

ofthefineearth

frac

tion(M

gm�3);

BDtot:

soilbulk

density

(Mgm�3);

WHC:waterholdingca

pac

ity(%);

litter:

litterco

ver(%);

SOM:topsoilorganic

matterco

ntent(%).

aOnly

variab

lesareincludedthat

donotca

use

problemsofmulticollinearity;

allregressionfunctionsaresign

ifica

ntat

p=0.00

0.bFo

rmultiple

regressions,

thead

justedR2is

give

n.

cVariablesareenumeratedin

orderofdecreasingco

rrelationwithRCtot.

236 K. Descheemaeker et al.

Autho

r's

pers

onal

co

py

lems. The latter is the case for Rc, which is also stronglycorrelated to the vegetation cover variables and to littercover in the first two groups of plots. For plots with VEG-tot,w < 65% the correlation between Rc and other plot vari-ables disappears, just as the correlation between Rc andRCtot. Although many authors discuss the influence of sur-face stoniness on runoff generation (e.g. Casenave and Val-entin, 1989; Wilcox and Wood, 1989; Poesen et al., 1990;Poesen and Ingelmo-Sanchez, 1992; Valentin and Casenave,1992; Poesen and Lavee, 1994; Poesen et al., 1994; Valen-tin, 1994; Gutierrez and Hernandez, 1996; Cerda, 2001)we must conclude from our results that for the case of steephillslopes in Tigray surface stoniness is not determining run-off production.

As densely vegetated plots are excluded from the analy-sis, other variables gain importance. One example is waterholding capacity, which is also correlated to bulk densityand Ksat, indicating that soil variables are more stronglyinfluencing runoff under a less dense vegetation cover. An-other interesting evolution when plots with a denser vegeta-tion cover are omitted is the inclusion of slope gradient asan explaining variable in the regression function. Slope gra-dient is not correlated with any other plot variable and dis-plays a negative relation with the runoff coefficient.Although most studies report the opposite, similar findingsconcerning the role of slope gradient in runoff processescan be found in the literature. Viramontes (1993, cited inDescroix et al., 2001) found smaller runoff coefficients onsteeper slopes, just as Descroix et al. (2001) for slope gradi-ents above 27%. Also Gresillon (1994, cited in Descroixet al., 2001) and Bohm and Gerold (1995) did not observehigher runoff on steeper slopes. These contra-intuitive re-sults can be related to the existence of a positive relationbetween slope gradient and infiltration rate, as demon-strated through laboratory research conducted by Poesen(1984). Different topsoil properties, such as a decreasedsurface sealing, arise through increased soil erosion on stee-per slopes and as such explain the negative correlation be-tween runoff coefficient and slope gradient.

Other correlations between plot variables and RCtot aremore easily explained. Litter cover acts similarly to a can-opy cover in capturing raindrops and dissipating their en-

ergy. Besides that, it is very effective in retarding thewater flow, so that its negative correlation with RCtot isnot surprising and also documented by other authors (Salaand Calvo, 1990; Vertessy et al., 1993; Gutierrez and Her-nandez, 1996; Zhou et al., 2002). Just like Gutierrez andHernandez (1996) and Calvo-Cases et al. (2003) we alsofound a positive correlation between bulk density andRCtot, which is explained by the fact that high bulk densityvalues usually correspond to lack of soil structure and assuch also to lower infiltration rates and higher runoffdepths. The negative correlation between soil organic mat-ter content and RCtot can also be reduced to an effect ofsoil structure on infiltration rate. Soils with high organicmatter content tend to have a better structure which favorsinfiltration and reduces runoff. Similar results reported bymany others (Young and Onstad, 1978; Wood et al., 1987;Le Bissonnais et al., 1995; Gutierrez and Hernandez, 1996;Descroix et al., 2001; Calvo-Cases et al., 2003) justify theincorporation of this variable in all of the multiple regres-sion functions in Table 10.

Conclusions

When degraded steep hillslopes in the Tigray highlands areallowed to rehabilitate, runoff production significantly de-creases. As soon as grazing is banned, the recovery processis noticeable as already in young exclosures runoff produc-tion is significantly lower than in grazing lands. Even whenthe rehabilitation process is prolonged to 15 or 20 years,still an appreciable improvement in runoff reduction ismade. The main reason for this runoff attenuation is therestoration of the natural vegetation, taking place in theexclosures. This was proven through correlation and regres-sion analysis both indicating total vegetation cover as themain factor influencing runoff production. Moreover, itwas shown that as soon as the total vegetation cover in anexclosure surpasses 65%, runoff becomes negligible. This isrelated to the exponential decay function between totalvegetation cover and total runoff coefficient, which entailsthat as vegetation cover increases, runoff coefficients rap-idly tend to zero. Another indication for the primary role

VEGtot,w (%)

0 20 40 60 80 100

T (

mm

)

0

5

10

15

20

25

30

001.0,27

42.0

18.013.02

===

+=

pn

R

xy

VEGtot,w

0 20 40 60 80 100

slop

ePR

0.0

0.2

0.4

0.6

0.8

1.0

000.0,27

82.0

514.2046.02

043.0

===

+−= −

pn

R

ey x

(a) (b)

Figure 11 Scatter plots and best fitting regression lines relating the rainfall threshold (T) (a) and the slope of the rainfall–runoffcurve (slopePR) (b) with weighted average total vegetation cover (VEGtot,w).

Runoff on slopes with restoring vegetation: A case study from the Tigray highlands, Ethiopia 237

Autho

r's

pers

onal

co

py

of vegetation cover, is its positive relation with the rainfallthreshold. Also other plot variables, such as litter cover, soilorganic matter content and bulk density are correlated withthe runoff coefficient. When focusing on less densely vege-tated locations other variables, such as slope gradient, startplaying a role in runoff processes.

Event variables such as rainfall depth, rainfall intensity,storm duration and soil moisture content determine the run-off depth that can be expected at a certain location in anexclosure. From the relative importance of their correlationcoefficients with runoff depth, it was concluded that theprevailing runoff generation mechanism must be Hortonian.

From these findings it is concluded that exclosure is asuccessful soil and water conservation strategy, significantlyreducing runoff through vegetation restoration. The result-ing higher infiltration benefits plant growth and biomassproduction and can also lead to groundwater recharge, thusreplenishing deeper lying water resources. Another impor-tant advantage of the decrease in runoff is that lower lyingcropland becomes less subject to damaging floods from theformerly degraded steep hillslopes. Whether these benefitswill compensate for the loss of already scarce grazing landin the local population’s perspective, depends on how theexclosure areas will be managed and to what degree localfarmers will be entitled to sustainably use these resourcesin the future.

Acknowledgement

The research was conducted within the framework of theZala-Daget project, funded by the VLIR (Flemish Interuni-versity Council, project no. VLIR/EI 237). Katrien Deschee-maeker is funded through an FWO Vlaanderen Aspirantscholarship. The support by Mekelle University and the lab-oratory facilities and at the Faculty of Applied Bioscienceand Engineering of the K.U. Leuven are greatly appreciated.Furthermore, we want to thank the local Bureau of Agricul-ture branch, the authorities and exclosure guards of severalvillages of the Dogu’a Tembien district, Yikuno Amlak Tek-leberhan, Haileselassie Gebremedhin, Haftu Assefa and Ber-hanu Gebremedhin for their field assistance, and severalstudents of the Dogu’a Tembien villages for carefully read-ing out boxes and raingauges.

References

Abrahams, A.D., Parsons, A.J., Wainwright, J., 1995. Effects ofvegetation change on interrill runoff and erosion, Walnut Gulch,southern Arizona. Geomorphology 13, 37–48.

Aerts, R., Wagendorp, T., November, E., Mintesinot, Behailu,Deckers, J., Muys, B., 2004. Ecosystem thermal buffer capacityas an indicator of the restoration status of protected areas in theNorthern Ethiopian Highlands. Restoration Ecology 12, 586–596.

Archer, N., Hess, T., Quinton, J., 2002. The water balance of twosemi-arid shrubs on abandoned land in South-Eastern Spain aftercold season rainfall. Hydrology and Earth System Sciences 6,913–926.

Bellot, J., Bonet, A., Sanchez, J.R., Chirino, E., 2001. Likely effectsof land use changes on the runoff and aquifer recharge in asemiarid landscape using a hydrological model. Landscape andUrban Planning 55, 41–53.

Bergkamp, G., 1998. A hierarchical view of the interactions ofrunoff and infiltration with vegetation and microtopography insemiarid shrublands. Catena 33, 201–220.

Blake, G.R., Hartge, K.H., 1990. Porosity. In: Klute, A., Campbell,G.S. (Eds.), Methods of Soil Analysis. 1: Physical and Mineralog-ical Methods, Agronomy 9,1. American Society of Agronomy,Madison, WI, USA, p. 1188.

Bochet, E., Rubio, J.L., Poesen, J., 1998. Relative efficiency ofthree representative matorral species in reducing water erosionat the microscale in a semi-arid climate (Valencia, Spain).Geomorphology 23, 139–150.

Bohm, P., Gerold, G., 1995. Pedo-hydrological and sedimentresponses to simulated rainfall on soils of the Konya Uplands(Turkey). Catena 25, 63–75.

Bosellini, A., Russo, A., Fantozzi, P.L., Assefa, G., Solomon, T.,1997. The Mesozoic succession of the Mekele outlier (TigreProvince, Ethiopia). Memorie di Scienze Geologiche 49, 95–116.

Bruijnzeel, L.A., 2004. Hydrological functions of tropical forests:not seeing the soil for the trees? Agriculture, Ecosystems andEnvironment 104, 185–228.

Calder, I.R., Hall, R.L., Bastable, H.G., Gunston, H.M., Shela, O.,Chirwa, A., Kafunda, R., 1995. The impact of land use change onwater resources in sub-Saharan Africa: a modelling study of LakeMalawi. Journal of Hydrology 170, 123–135.

Calvo-Cases, A., Boix-Fayos, C., Imeson, A.C., 2003. Runoff gener-ation, sediment movements and soil water behaviour on calcar-eous (limestone) slopes of some Mediterranean environments insoutheast Spain. Geomorphology 50, 269–291.

Cammeraat, L.H., Imeson, A.C., 1999. The evolution and signifi-cance of soil-vegetation patterns following land abandonmentand fire in Spain. Catena 37, 107–127.

Casenave, A., Valentin, C., 1989. Les etats de surface de la zonesahelienne: influence sur l’infiltration, Orstom, Paris.

Castillo, V.M., Martınez-Mena, M., Albaladejo, J., 1997. Runoff andsoil loss response to vegetation removal in a semiarid environ-ment. Soil Science Society of America Journal 61, 1116–1121.

Castro, N.M.D.R., Auzet, A.-V., Chevallier, P., Leprun, J.-C., 1999.Land use change effects on runoff and erosion from plot tocatchment scale on the basaltic plateau of Southern Brazil.Hydrological Processes 13, 1621–1628.

Cerda, A., 1997a. The effects of patchy distribution of Stipatenacissima L. on runoff and erosion. Journal of Arid Environ-ments 36, 37–51.

Cerda, A., 1997b. Seasonal changes of the infiltration rates in aMediterranean scrubland on limestone. Journal of Hydrology198, 209–225.

Cerda, A., 1998. The influence of geomorphological position andvegetation cover on the erosional and hydrological processes ona Mediterranean hillslope. Hydrological Processes 12, 661–671.

Cerda, A., 2001. Effects of rock fragment cover on soil infiltration,interrill runoff and erosion. European Journal of Soil Science 52,59–68.

Chadhokar, P.A., Solomon Abate, 1988. Importance of revegetationin soil conservation in Ethiopia. In: Richwanich, S. (Ed.), LandConservation for Future Generations. Proceedings of the VthInternational Soil Conservation Conference, Bangkok, Thailand,pp. 1203–1213.

Childs, S.W., Flint, A.L., 1990. Physical properties of forest soilscontaining rock fragments. In: Gessel, S.P. et al. (Eds.),Sustained Productivity of Forest Soils. University of BritishColumbia, Faculty of Forestry Publ.,, Vancouver, BC, pp. 95–121.

Cordery, I., Pilgrim, D.H., Doran, D.G., 1983. Some hydrologicalcharacteristics of arid western New South Wales. In: Hydrologyand Water Resources Symposium 1983, Inst. Engrs. Austral.,National Conference Publ. no. 83/13, pp. 287–292.

Dagnachew, Legesse, Vallet-Coulomb, C., Gasse, F., 2003. Hydro-logical response of a catchment to climate and land use changes

238 K. Descheemaeker et al.

Autho

r's

pers

onal

co

py

in Tropical Africa: case study South Central Ethiopia. Journal ofHydrology 275, 67–85.

Descheemaeker, K., Nyssen, J., Rossi, J., Poesen, J., Mitiku, Haile,Raes, D., Muys, B., Moeyersons, J., Deckers, J., 2006. Sedimentdeposition and pedogenesis in exclosures in the Tigray highlands,Ethiopia. Geoderma 32, 291–314.

Descheemaeker, K., Nyssen, J., Poesen, J., Mitiku Haile, Muys, B.,Raes, D., Moeyersons, J., Deckers, J. Soil and water conserva-tion through forest restoration in exclosures of the Tigrayhighlands. Journal of the Drylands, in press.

Descroix, L., Viramontes, D., Vauclin, M., Gonzalez Barrios, J.L.,Esteves, M., 2001. Influence of soil surface features andvegetation on runoff and erosion in the Western Sierra Madre(Durango, Northwest Mexico). Catena 43, 115–135.

Desir, G., 2002. Hydrological response types for gypsiferous soils ina semi-arid region during nine years of continuous record.Hydrological Processes 16, 2685–2700.

Diamantopoulos, J., Pantis, J., Sgardelis, S., Iatrou, G., Pirintsos,S., Papatheodorou, E., Dalaka, A., Stamou, G.P., Cammeraat,L.H., Kosmas, C., 1996. The Petralona and Hortiatis field sites(Thessaloniki, Greece). In: Brandt, C.J., Thornes, J.B. (Eds.),Mediterranean Desertification and Land Use. Wiley, Chichester,pp. 229–246.

Dunjo, G., Pardini, G., Gispert, M., 2003. Land use change effectson abandoned terraced soils in a Mediterranean catchment, NESpain. Catena 52, 23–37.

Dunjo, G., Pardini, G., Gispert, M., 2004. The role of land use-landcover on runoff generation and sediment yield at a microplotscale, in a small Mediterranean catchment. Journal of AridEnvironments 57, 99–116.

Dunne, T., 1978. Field studies of hillslope flow processes. In: Kirkby,M.J. (Ed.), Hillslope Hydrology. Wiley, New York, NY, pp. 227–293.

Dunne, T., Dietrich, W.E., 1980. Experimental study of Hortonoverland flow on tropical hillslopes: 1. Soil conditions, infiltra-tion and frequency of runoff. Zeitschrift fur GeomorphologieN.F., Suppl. Bd., Surface Runoff and Erosion 35, 40–59.

Eweg, H.P.A., Van Lammeren, R., Deurloo, H., Woldu, Z., 1998.Analysing degradation and rehabilitation for sustainable landmanagement in the highlands of Ethiopia. Land Degradation andDevelopment 9, 529–542.

FAO, ISRIC, ISSS, 1998. World Reference Base of Soil Resources.World Soil Resources Reports 84, FAO, Rome, Italy.

Fiedler, H.J., Gebeyehu, Belay, 1988. Forests and their importancefor soil conservation in Ethiopia. Archiv fur Naturschutz undLandschaftsforschung 28, 161–175.

Fiener, P., Auerswald, K., 2003. Effectiveness of grassed waterwaysin reducing runoff and sediment delivery from agriculturalwatersheds. Journal of Environmental Quality 32, 927–936.

Francis, C.F., Thornes, J.B., 1990. Runoff hydrographs from threeMediterranean vegetation cover types. In: Thornes, J.B. (Ed.),Vegetation and Erosion. Wiley, Chichester, pp. 363–384.

Daniel Gamachu, 1977. Aspects of climate and water budget inEthiopia. Addis Ababa University Press, 71 pp.

Gao, Y., Qiu, G.Y., Shimizu, H., Tobe, T., Sun, B., Wang, J., 2002. A10-year study on techniques for vegetation restoration in adesertified salt lake area. Journal of Arid Environments 52, 483–497.

Gifford, G.F., 1985. Cover allocation in rangeland watershedmanagement (a review). In: Jones, E.B., Ward, T.J. (Eds.),Watershed Management in the Eighties, Proceedings of aSymposium. ASCE, New York, pp. 23–31.

Gresillon, J.M., 1994. Contribution a l’etude de la formation desecoulements de crue sur les petits bassins-versants. Diplomed’Habilitation a Diriger des Recherches, UJF-Grenoble 1.

Gutierrez, J., Hernandez, I.I., 1996. Runoff and interrill erosion asaffected by grass cover in a semi-arid rangeland of northernMexico. Journal of Arid Environments 34, 287–295.

Herweg, K., Stillhardt, B., 1999. Implications of land use and landcover dynamics for mountain resource degradation in theNorthwestern Ethiopian Highlands. The variability of soil erosionin the highlands of Ethiopia and Eritrea. Average and extremeerosion patterns. Soil Conservation Research Programme,Research Report 42, University of Berne, Switzerland.

Hongo, A., Matsumoto, S., Takahashi, H., Zou, H.Y., Cheng, J.M.,Jia, H.Y., Zhao, H.Y., 1995. Effect of exclosure and topographyon rehabilitation of overgrazed shrub-steppe in the Loessplateau of Northwest China. Restoration Ecology 3, 18–25.

Huang, M., Zhang, L., 2004. Hydrological responses to conservationpractices in a catchment of the Loess Plateau, China. Hydro-logical Processes 18, 1885–1898.

Hurni, H., 1990. Degradation and conservation of the resources inthe Ethiopian highlands. Mountain Research and Development 8,123–130.

Kang, S., Zhang, L., Song, X., Zhang, S., Liu, X., Liang, Y., Zheng,S., 2001. Runoff and sediment loss responses to rainfall and landuse in two agricultural catchments on the Loess Plateau ofChina. Hydrological Processes 15, 977–988.

Karnieli, A., Ben-Asher, J., 1993. A daily runoff simulation in semi-arid watersheds based on soil water deficit calculations. Journalof Hydrology 149, 9–25.

Kebrom, Tekle, Backeus, I., Skoglund, J., Zerihun, Woldu, 1997.Vegetation on hillslopes in southern Wello, Ethiopia: degrada-tion and regeneration. Nordic Journal of Botany 17, 483–493.

Kessler, J., Oosterbaan, R.J., 1974. Determining hydraulic condi-tions of soils. In: de Ridder, N.A. et al. (Eds.), DrainagePrinciples and Applications, Surveys and Investigations, vol. 3.International Institute for Land Reclamation and Improvement,Wageningen, The Netherlands, pp. 253–296.

Kosmas, D., Danalatos, N., Cammeraat, L.H., Chabart, M., 1997.The effect of land use on runoff and soil erosion rates underMediterranean conditions. Catena 29, 45–59.

Lang, R.D., 1979. The effects of ground cover on surface runoff fromexperimental plots. Journal of Soil Conservation 35, 108–114.

Lasanta, T., Garcıa-Ruiz, J.M., Perez-Rontome, C., Sancho-Marcen,C., 2000. Runoff and sediment yield in a semi-arid environment:the effect of land management after farmland abandonment.Catena 38, 265–278.

Lavee, H., Imeson, A.C., Sarah, P., 1998. The impact of climatechange on geomorphology and desertification along a Mediter-ranean arid transect. Land Degradation and Development 9,407–422.

Le Bissonnais, Y., Renaux, B., Delouche, H., 1995. Interactionsbetween soil properties and moisture content in crust forma-tion, runoff and interrill erosion from tilled Loess soils. Catena25, 33–46.

Le Houerou, H.N., 2000. Restoration and rehabilitation of arid andsemiarid Mediterranean ecosystems in North Africa and WestAsia: a review. Arid Soil Research and Rehabilitation 14, 3–14.

Lopez-Bermudez, F., Romero-Dıaz, A., Martınez-Fernandez, J.,Martınez-Fernandez, J., 1998. Vegetation and soil erosion undera semi-arid Mediterranean climate: a case study from Murcia(Spain). Geomorphology 24, 51–58.

Mapa, R.B., 1995. Effect of reforestation using Tectona grandis oninfiltration and soil water retention. Forest Ecology and Man-agement 77, 119–125.

Martınez-Mena, M., Albaladejo, J., Castillo, V.M., 1998. Factorsinfluencing surface runoff generation in a Mediterranean semi-arid environment: Chicamo watershed, SE Spain. HydrologicalProcesses 12, 741–754.

McDonald, M.A., Healey, J.R., Stevens, P.A., 2002. The effects ofsecondary forest clearance and subsequent land-use on erosionlosses and soil properties in the Blue Mountains of Jamaica.Agriculture, Ecosystems and Environment 92, 1–19.

Morgan, R.P., Finney, H.J., Lavee, H., Merrit, E., Noble, C.A., 1986.Plant cover effects on hillslope runoff and erosion: evidence

Runoff on slopes with restoring vegetation: A case study from the Tigray highlands, Ethiopia 239

Autho

r's

pers

onal

co

py

from two laboratory experiments. In: Abrahams, A.D. (Ed.),Hillslope Processes. Allen and Unwin, Boston, pp. 77–96.

Mwendera, E.J., Mohamed Saleem, M.A., 1997. Hydrologic responseto cattle grazing in the Ethiopian highlands. Agriculture,Ecosystems and Environment 64, 33–41.

Mwendera, E.J., Mohamed Saleem, M.A., Dibabe, A., 1997. Theeffect of livestock grazing on surface runoff and soil erosionfrom sloping pasture lands in the Ethiopian highlands. AustralianJournal of Experimental Agriculture 37, 421–430.

Narain, P., Singh, R.K., Sindhwal, N.S., Joshie, P., 1998. Waterbalance and water use efficiency of different land uses inwestern Himalayan valley region. Agricultural Water Manage-ment 37, 225–240.

Nicolau, J.M., Sole-Benet, A., Puigdefabregas, J., Gutierrez, L.,1996. Effects of soil and vegetation on runoff along a catena insemi-arid Spain. Geomorphology 14, 297–309.

Northcliff, S., Ross, S.M., Thornes, J.B., 1990. Soil moisture, runoffand sediment yield from differentially cleared tropical rainfor-est plots. In: Thornes, J.B. (Ed.), Vegetation and Erosion. Wileyand Sons, Ltd., New York, pp. 419–436.

Nyssen, J., Moeyersons, J., Poesen, J., Deckers, J., Mitiku, Haile,2002. The environmental significance of the remobilisation ofancient mass movements in the Atbara-Tekeze headwaters,Northern Ethiopia. Geomorphology 49, 303–322.

Nyssen, J., Poesen, J., Moeyersons, J., Deckers, J., Mitiku, Haile,Lang, A., 2004. Human impact on the environment in theEthiopian and Eritrean highlands – a state of the art. Earth-Science Reviews 64, 273–320.

Nyssen, J., Vandenreyken, H., Poesen, J., Moeyersons, J., Deckers,J., Mitiku, Haile, Salles, C., Govers, G., 2005. Rainfall erosivityand variability in the Northern Ethiopian Highlands. Journal ofHydrology 311, 172–187.

Orr, H.K., 1970. Runoff and erosion control by seeded and nativevegetation on a forest burn, Black Hills, South Dakota. US ForestService, Rocky Mountain Forest and Range Experiment Station.Fort Collins, CO, Research paper RM-60, 12 pp.

Pardini, G., Gispert, M., Dunjo, G., 2003. Runoff erosion andnutrient depletion in five Mediterranean soils of NE Spain underdifferent land use. The Science of the Total Environment 309,213–224.

Peugeot, C., Esteves, M., Galle, S., Rajot, J.L., Vandervaere, J.P.,1997. Runoff generation processes: results and analysis of fielddata collected at the East Central Supersite of the HAPEX-Sahelexperiment. Journal of Hydrology, 179–202.

Pilgrim, D.H., Chapman, T.G., Doran, D.G., 1988. Problems ofrainfall–runoff modelling in arid and semiarid regions. Hydro-logical Sciences Journal 33, 379–400.

Poesen, J., 1984. The influence of slope angle on infiltration rateand Hortonian overland flow volume. Zeitschrift fur Geomor-phologie, N.F. Supplementband 49, 117–131.

Poesen, J., Ingelmo-Sanchez, F., 1992. Runoff and sediment yieldfrom topsoils with different porosity as affected by rockfragment cover and position. Catena 19, 451–474.

Poesen, J., Lavee, H., 1994. Rock fragments in top soils: signif-icance and processes. Catena 23, 1–28.

Poesen, J., Ingelmo-Sanchez, F., Mucher, H., 1990. The hydrolog-ical response of soil surfaces to rainfall as affected by cover andposition of rock fragments in the top layer. Earth SurfacesProcesses and Landforms 14, 469–480.

Poesen, J., Torri, D., Bunte, K., 1994. Effects of rock fragments onsoil erosion by water at different spatial scales: a review. Catena23, 141–166.

Poore, M.E.D., Fries, C., 1985. The ecological effects of Eucalyptus.FAO, Rome, Forestry Paper 59, 87 pp.

Puigdefabregas, J., Sole, A., Gutierrez, L., del Barrio, G., Boer, M.,1999. Scales and processes of water and sediment redistributionin drylands: results from the Rambla Honda field site inSoutheast Spain. Earth-Science Reviews 48, 39–70.

Romero Dıaz, A., Lopez-Bermudez, F., Thornes, J.B., Francis, C.F.,Fisher, G.C., 1988. Variability of overland flow rates in a semi-arid mediterranean environment under matorral cover Murcia,Spain. Catena 13 (Suppl.), 1–11.

Romero Dıaz, A., Cammeraat, L.H., Vacca, A., Kosmas, C.,1999. Soil erosion at three experimental sites in the Medi-terranean. Earth Surface Processes and Landforms 24, 1243–1256.

Sala, M., 1988. Slope runoff and sediment production in twomediterranean mountain environments. Catena Supplement 12,13–29.

Sala, M., Calvo, A., 1990. Response of four different Mediterraneanvegetation types to runoff and erosion. In: Thornes, J.B. (Ed.),Vegetation and Erosion. Wiley and Sons, Ltd., New York, pp.347–362.

SCRP, D., 2000. Long-term monitoring of the agricultural environ-ment in six research stations in Ethiopia. Soil Erosion andConservation Database, vol. 7. Soil Conservation ResearchProject, Berne.

Sheldrik, B.H., Wang, C., 1993. Particle size distribution. In: Carter,M.R. (Ed.), Soil Sampling and Methods of Analysis, CanadianSociety of Soil Science. Lewis Publishers, Boca Raton, USA, pp.499–512.

Snelder, D.J., Bryan, R.B., 1995. The use of rainfall simulation teststo assess the influence of vegetation density on soil loss ondegraded rangelands in the Baringo District, Kenya. Catena 25,105–116.

Sole-Benet, A., Calvo, A., Cerda, A., Lazaro, R., Pini, R., Barbero,J., 1997. Influences of micro-relief patterns and plant cover onrunoff related processes in badlands from Tabernas (SE Spain).Catena 31, 23–38.

Sorriso-Valvo, M., Bryan, R.B., Yair, A., Iovino, F., Antronico, L.,1995. Impact of afforestation on hydrological response andsediment production in a small Calabrian catchment. Catena 25,89–104.

SPSS 11.0.1, 2001. SPSS for windows, release 11.0.1. SPSS Inc.Taddese, G., Saleem, M.A.M., Abyie, A., Wagnew, A., 2002a.

Impact of grazing on plant species richness, plant biomass, plantattribute, and soil physical and hydrological properties ofvertisol in East African highlands. Environmental Management29, 279–289.

Taddese, G., Saleem, M.A.M., Astatke, A., Ayaleneh, W., 2002b.Effect of grazing on plant attributes and hydrological propertiesin the sloping lands of the East African highlands. EnvironmentalManagement 30, 406–417.

Tenna Shitarek, Sintayehu Manaye, Berihun Abebe, 2001. Strength-ening user-rights over local resources in Wollo, Ethiopia. IIEDDrylands Programme, Issue paper no. 103, 23 pp.

Thurow, T.L., Blackburn, W.H., Taylor, C.A., 1986. Hydrologiccharacteristics of vegetation types as affected by livestockgrazing systems, Edwards Plateau, Texas. Journal of RangeManagement 39, 505–509.

Tiessen, H., Moir, J.O., 1993. Total and organic carbon. In: Carter,M.R. (Ed.), Soil Sampling and Methods of Analysis, CanadianSociety of Soil Science. Lewis Publishers, Boca Raton, USA, pp.187–200.

Topp, G.C., Davis, J.L., Annan, A.P., 1980. Electromagneticdetermination of soil water content: measurements incoaxial transmission lines. Water Resources Research 16,374–582.

UNEP, 1994. United Nations Convention to Combat Desertification,Berne, 71pp.

Vacca, A., Loddo, S., Ollesch, G., Puddu, R., Serra, G., Tomasi, D.,Aru, A., 2000. Measurement of runoff and soil erosion in threeareas under different land use in Sardinia (Italy). Catena 40, 69–92.

Valentin, C., 1994. Surface sealing as affected by various rockfragment covers in West Africa. Catena 23, 87–97.

240 K. Descheemaeker et al.

Autho

r's

pers

onal

co

py

Valentin, C., Casenave, A., 1992. Infiltration into sealed soils asinfluenced by gravel cover. Soil Science Society of AmericaJournal 56, 1167–1673.

Van Dijk, P.M., Kwaad, F.J.P.M., Klapwijk, M., 1996. Retention ofWater and sediment by grass strips. Hydrological Processes 10,1069–1080.

Van Reeuwijk, L.P., 2002. Procedures for Soil Analysis. ISRIC,Wageningen, The Netherlands.

Vertessy, R.A., Hatton, T.J., O’Shaughnessy, P.J., Jayasuriya,M.D.A., 1993. Predicting water yield from a mountain ash forestcatchment using a terrain analysis based catchment model.Journal of Hydrology 2, 665–700.

Viramontes, D., 1993. Redistribucion espacial del agua en elpaisaje: escurrimiento y erosion hydrica a traves de unatoposecuencia. Actas del Seminario Mapimi. Instituto de Eco-logiarOrstom, Gomez Palacio, Dgo, Mexico, pp. 143–159.

Wilcox, B., Wood, K., 1988. Hydrologic impacts of sheep grazing onsteep slopes in semiarid rangelands. Journal of Range Manage-ment 41, 303–306.

Wilcox, B., Wood, K., 1989. Factors influencing interrill erosionfrom semiarid slopes in New Mexico. Journal of Range Manage-ment 42, 66–70.

Wisborg, P., Shylendra, H.S., Gebrehiwot, K., Shanker, R., Tilahun,Y., Nagothu, U.S., Tewoldeberhan, S., Bose, P., 2000. Rehabil-itation of CPRS through re-crafting of village institutions: acomparative study from Ethiopia and India. Paper Presented at

the 8th Biennial Conference of the International Association forthe Study of Common Property (IASCP), Bloomington, Indiana,May 31–June 4, 2000, 26pp.

Woldeamlak, Bewket, Sterk, G., 2005. Dynamics in land cover andits effect on stream flow in the Chemoga watershed, Blue Nilebasin, Ethiopia. Hydrological Processes 19, 445–458.

Wood, J.C., Wood, M.K., Tromble, J.M., 1987. Important factorsinfluencing water infiltration and sediment production on aridlands in New Mexico. Journal of Arid Environments 12, 111–118.

Yair, A., Kossovsky, A., 2002. Climate and surface propertieshydrological response of small arid and semi-arid watersheds.Geomorphology 42, 43–57.

Yair, A., Lavee, H., 1985. Runoff generation in arid and semi-aridzones. In: Anderson, M.G., Burt, T.P. (Eds.), HydrologicalForecasting. John Wiley and Sons, New York, pp. 183–220.

Young, R.A., Onstad, C.A., 1978. Characterization of rill andinterrill eroded soils. Transactions of the American Society ofAgricultural Engineers 21, 1126–1130.

Zhang, B., Yang, Y., Zepp, H., 2004. Effect of vegetation restora-tion on soil and water erosion and nutrient losses of a severelyeroded clayey Plinthudult in southeastern China. Catena 57, 77–90.

Zhou, G.Y., Morris, J.D., Yan, J.H., Yu, Z.Y., Peng, S.L., 2002.Hydrological impacts of reafforestation with eucalypts andindigenous species: a case study in southern China. ForestEcology and Management 167, 209–222.

Runoff on slopes with restoring vegetation: A case study from the Tigray highlands, Ethiopia 241