Analysis of factors determining sediment yield variability in the highlands of northern Ethiopia L. Tamene a, * , S.J. Park b , R. Dikau c , P.L.G. Vlek a a Center for Development Research (ZEF), University of Bonn, Walter-Flex-Str.3, 53113, Bonn, Germany b Department of Geography, College of Social Sciences, Seoul National University, Seoul 151-742, South Korea c Department of Geography, University of Bonn, Meckenheimer Allee 166, 53115, Bonn, Germany Received 11 April 2005; received in revised form 16 October 2005; accepted 20 October 2005 Available online 15 December 2005 Abstract In many developing countries, sustainable land management and water resources development are threatened by soil erosion and sediment-related problems. In response to such threats, there is an urgent need for improved catchment-based erosion control and sediment management strategies. The design and implementation of such strategies require data on erosion rates and understanding of the factors that control the delivery of sediment through the catchment system. In this study, reservoir sedimentation and corresponding catchment attribute data were used to investigate the major factors controlling sediment yield variability in a mountainous dryland region of northern Ethiopia. Sediment yield data were acquired for representative 11 catchments above reservoirs. Geomorphological and anthropogenic catchment attributes were extracted for each reservoir from different sources including digital elevation models, satellite images and field surveys. Different statistical analyses such as Pearson’s correlation, principal components and multiple regression were implemented to analyze the relationship between sediment yield and catchment characteristics and to determine the major factors controlling the variability of sediment yield. The results show that terrain form, gully erosion, surface lithology, and land cover explain most of the variability in sediment yield among the catchments. The implications of the results, for relevant management intervention targeted at ameliorating the major causative factors of erosion, are also outlined. D 2005 Elsevier B.V. All rights reserved. Keywords: Reservoir survey; Catchment attributes; Statistical analysis; Sediment yield variability; Northern Ethiopia 1. Introduction Agricultural productivity in Ethiopia is highly influ- enced by erratic and unpredictable rainfall, which led to recurrent droughts claiming thousands of human and livestock lives (e.g., Degefu, 1987; Hurni, 1993). To mitigate such crisis, it is necessary to minimize the deleterious impact of rainfall variability through the provision of adequate water supply and its proper uti- lization (Lawrence et al., 2004). Therefore, the govern- ment of Ethiopia, in collaboration with international organizations, launched a massive surface-water har- vesting scheme through the construction of micro- dams. For instance, from 1996 to 2001, over 50 micro-dams were built in the Tigray administrative region of northern Ethiopia. The construction of the micro-dams resulted in var- ious economical, hydrological, and ecological benefits including increase in food production, easy access to drinking water for people and livestock, rise in the ground water level and issuance of new springs (e.g., 0169-555X/$ - see front matter D 2005 Elsevier B.V. All rights reserved. doi:10.1016/j.geomorph.2005.10.007 * Corresponding author. Fax: +49 228 731889. E-mail address: lulseged _ [email protected] (L. Tamene). Geomorphology 76 (2006) 76 – 91 www.elsevier.com/locate/geomorph
16
Embed
Analysis of factors determining sediment yield variability in the highlands of northern Ethiopia
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
www.elsevier.com/locate/geomorph
Geomorphology 76
Analysis of factors determining sediment yield variability in the
highlands of northern Ethiopia
L. Tamene a,*, S.J. Park b, R. Dikau c, P.L.G. Vlek a
a Center for Development Research (ZEF), University of Bonn, Walter-Flex-Str.3, 53113, Bonn, Germanyb Department of Geography, College of Social Sciences, Seoul National University, Seoul 151-742, South Korea
c Department of Geography, University of Bonn, Meckenheimer Allee 166, 53115, Bonn, Germany
Received 11 April 2005; received in revised form 16 October 2005; accepted 20 October 2005
Available online 15 December 2005
Abstract
In many developing countries, sustainable land management and water resources development are threatened by soil erosion
and sediment-related problems. In response to such threats, there is an urgent need for improved catchment-based erosion control
and sediment management strategies. The design and implementation of such strategies require data on erosion rates and
understanding of the factors that control the delivery of sediment through the catchment system. In this study, reservoir
sedimentation and corresponding catchment attribute data were used to investigate the major factors controlling sediment yield
variability in a mountainous dryland region of northern Ethiopia. Sediment yield data were acquired for representative 11
catchments above reservoirs. Geomorphological and anthropogenic catchment attributes were extracted for each reservoir from
different sources including digital elevation models, satellite images and field surveys. Different statistical analyses such as
Pearson’s correlation, principal components and multiple regression were implemented to analyze the relationship between
sediment yield and catchment characteristics and to determine the major factors controlling the variability of sediment yield.
The results show that terrain form, gully erosion, surface lithology, and land cover explain most of the variability in sediment yield
among the catchments. The implications of the results, for relevant management intervention targeted at ameliorating the major
Proportion of total bare land (TBARE) �0.171 �0.957 �0.034 �0.045Proportion of cultivable land (CULT) �0.186 �0.944 �0.057 �0.183Proportion of dense cover (DENSE) 0.401 0.891 �0.039 �0.050Proportion of erodible lithology (EL) 0.197 �0.889 0.264 0.069
Proportion of open grazing land (NGL) 0.233 0.239 0.065 0.767
Drainage density (DD) 0.141 0.469 0.163 �0.623The first four PCs explain about 88% of the variability in the dataset. Bold values indicate highest eigenvector score for a given variable of the first
two significant components.
L. Tamene et al. / Geomorphology 76 (2006) 76–91 85
4.4. Multiple regression
Before applying multiple regression, it was neces-
sary to keep the number of explanatory variables (20)
lower than the number of cases (11) to avoid the
problem of inflated R2 (Phippen and Wohl, 2003). It
was also necessary to minimize the effect of autocorre-
lation of environmental factors on multiple regression
analysis (e.g., Phippen and Wohl, 2003). The PCA
made it possible to reduce the dimensionality of the
data and to address the problem of multicollinearity.
From the two significant PCs (Table 6), catchment
attributes with a high eigenvector value were selected,
and the actual values of the factors were used for
regression. The applied Model 1 in Table 7 shows
that over 80% of the variability in sediment yield can
be explained by the height difference (HD) and the
cultivated land percentage (CULT). A similar level of
significance was achieved (over 80%) when only HD
and dense cover percentage (DENSE) were used
(Model 2). Inclusion of factors from PC3 and PC4
into the regression analysis resulted in very minor con-
tributions of about 1%, confirming the insignificance of
these PCs. The Model 3 uses the gully/bank-erosion-
related factor (SBCR) for PC1 instead of HD, and the
variability explained has increased to over 90% (Table
7). A similar level of significance is achieved when
SBCR and DENSE are used in the regression (not
shown). Again, the contribution of the parameters
representing PC3 and PC4 to the overall regression
was very minor (less than 3%).
4.5. Step-wise regression
In order to further explore the role of factors affect-
ing the variation in observed sediment yield, a step-
wise regression was run using 18 out of the 20 variables
shown in Table 3, while the multicollinearity between
independent variables was kept to the minimum. Two
factors, mean elevation and the circularity ratio, were
not used because they are negatively correlated with
most of the terrain attributes which are positively and
significantly correlated with logSSY. The best predic-
tive equation to estimate SSY to reservoirs is:
logSSY ¼ 0:007SBCRþ 0:003ELþ 0:002RG
� 0:007BUSHþ 2:33
R2 ¼ 0:96; a b 0:05 ð7Þ
Eq. (7) shows that terrain-, lithology-, gully- and LUC-
related factors explain most of the variation in SSY,
Table 7
Multiple regression coefficients estimated by each unstandardized statistical model for the 11 reservoirs
Model Independent variable b estimate Standard error P-value Pearson’s r
Model 1 (R2=0.83, P=0.002) Height difference 0.0016 0.000 0.000 0.64*
Cultivated land 0.015 0.003 0.002 0.31
Constant 1.59
Model 2 (R2=0.83, P=0.002) Height difference 0.0019 0.000 0.000 0.64*
Dense cover �0.015 0.003 0.002 0.31
Constant 2.731
Model 3 (R2=0.92, P b0.001) Gully/bank erosion 0.011 0.001 0.000 0.92**
Cultivated land 0.0056 0.002 0.03 0.31
Constant 0.26
**Significant at 0.01. *Significant at 0.05. Inclusion of the elements of PC3 and PC4 does not change the significant level as such (R2 =0.84 for
models 1 and 2, and R2 =0.95 for model 3).
L. Tamene et al. / Geomorphology 76 (2006) 76–9186
which agrees with the result of the PCA. When regres-
sion was run after excluding gully erosion, which is the
most significantly correlated variable with SSY, the
following relationship can be obtained:
logSSY ¼ 0:0011HDþ 0:009ELþ 0:019
R2 ¼ 0:87; a b 0:001 ð8Þ
5. Discussion
5.1. Correlation analysis
The height difference (HD) and ruggedness number
(RG) show positive correlation with logSSY (Table 5)
because they reflect potential energy available to detach
and transport soil particles (Sarangi et al., 2003). The
proportion of erodible lithology (EL) plays an impor-
tant role in the siltation of reservoirs. Erodible lithology
types showing high SSY are shale and marl, and less
erodible lithology types are sandstone and metavolca-
nics. Similar observations are reported elsewhere (e.g.,
Lahlou, 1988; Fargas et al., 1997; Phippen and Wohl,
2003). Gully erosion/bank collapse (SBCR) plays a
very significant role in the siltation of reservoirs.
Dense networks of gullies (e.g., Fig. 2) increase slope
collapses and catchment connectivity, and facilitate
sediment delivery (e.g., Ownes and Slaymaker, 1993;
Wasson, 1994; Trimble, 1995; Walling et al., 1998;
Poesen et al., 2003). Livestock disturbances of gully
floors and banks as well as trampling of areas nearby
reservoirs (Fig. 2d) worsen gully erosion in the study
area. Similar observations are reported in other regions
(e.g., Trimble and Mendel, 1995; Lloyd et al., 1998).
Among the terrain-related attributes, the mean slope
(MES) shows a very poor correlation with sediment
yield in the study area, although their strong correla-
tions have often been reported (e.g., Hicks et al., 1996;
Schiefer et al., 2001). This observation suggests that the
role of slope is masked due to its association with dense
surface cover and less erodible lithology. In the study
areas, MES is positively correlated with dense surface
cover (DENSE), because the conservation and affores-
tation efforts have been concentrated on steep slopes.
Most steep slopes are also less accessible and therefore
less exposed to human and livestock disturbances. MES
is also negatively correlated with easily erodible lithol-
ogy (Table 5), which is also observed by Kirkby et al.
(2003) and Mills (2003). In addition, most steep slopes
are located away from the reservoirs so their influence
on reservoir siltation could be limited (Verstraeten and
Poesen, 2001a).
If it is difficult to assess the separate effects of
different factors, it may be necessary to stratify the
sites (Lu and Higgitt, 1999; Rustomji and Prosser,
2001). The correlation between some catchment attri-
butes and logSSY improved when three catchments
(Laelaywukro, Korir and Teghane) with relatively bet-
ter management, surface cover, and more resistant li-
thology (Fig. 5) were excluded from analysis. For
instance, catchment area (R2=0.80), mean slope
(R2=0.73), drainage length (R2=0.71), and elongation
ratio (R2=0.84) became significantly correlated with
logSSY (a b0.05). Catchment area (A) is positively
correlated with sediment yield as opposed to the
buniversalQ area�SSY relationship (e.g., Walling,
1983; Verstraeten et al., 2003). This may be because
of its positive and significant correlation with most of
the terrain attributes that are positively correlated to
logSSY (Table 5). The relationship between catchment
area and sediment yield could also be positive when the
main sediment source is from channels and floodplain
depositions (e.g., Walling and Webb, 1996), which is an
observed phenomenon in the study sites (Tamene et al.,
submitted for publication).
Fig. 5. Scatter plot of specific sediment yield against mean catchment slope showing a difference between less-managed and well-managed
catchments.
L. Tamene et al. / Geomorphology 76 (2006) 76–91 87
5.2. Principal component analysis
According to the constants in Eq. (6), sediment yield
is more sensitive to changes in terrain- and gully-related
factors (PC1) than to changes in LUC and lithology
(PC2) (Fig. 4a,b), which is also reflected in the corre-
lation result (Table 5). Fig. 4a,b shows that the first two
PCs account for about 92% of the total variability in
sediment yield (a b0.05). The PCA revealed the effect
of LUC on sediment yield (e.g., Fig. 4b), which was not
the case in the correlation analysis (Table 5).
Fig. 4a indicates that sites with low HD and SBCR
correspond to low sediment yield while those with high
HD and SBCR correspond to high sediment yield. On
the other hand, Fig. 4b indicates that those sites with
low proportion of dense cover or high proportion of
bare land and less resistant lithology correspond to high
sediment yield.
Comparison of Fig. 4a,b with Fig. 4c,d shows the
significant role of the factors of the two PCs in the
arrangement of sites within the PC spaces. Catch-
ments located at the lower end of PC1 have generally
similar characteristics, mainly low HD and low SBCR
(Fig. 4a,c). Adiakor (1) is located at the lowest end of
PC1 due to its large hypsometric integral. Gindae (5),
Grashito (6) and Maidelle (9) occupy a proximity
space because of their similar SBCR values. Adike-
nafiz (2) and Gerebmihiz (3) are located at the upper
end of PC1 mainly because they have high HD and
SBCR.
Similar interpretation can be given to the distribution
of sites in relation to PC2 (Fig. 4b,d). Catchments are
generally clustered into two categories mainly based on
surface cover and lithology. Laelaywukro (8), Korir (7)
and Teghane (11) are located at the higher position of
PC2 because these sites have relatively good surface
cover and more resistant surface lithology compared to
the others. Laelaywukro (8) is located at the higher end
of PC2 mainly due to its relatively high BUSH cover,
high MES and low CULT relative to the others. Grashito
(6), Maidelle (9) and Gerebsegen (4) are located at the
lower end of PC2 mainly because of their poor surface
cover and shale- and marl-dominated susceptible sur-
face lithology. The position of Adikenafiz (2) and Ger-
ebmihiz (3) as well as Majae (10) and Adiakor (1) in the
PC2 space is a result of their CULT or TBARE.
5.3. Multiple regression analysis
The results of the regression analysis (models 1 to 3
and Eqs. (7) and (8)) show that terrain form, surface
lithology, surface cover and gullies play significant
roles in the siltation of reservoirs. The catchments
characterized by pronounced terrain, less resistant li-
thology and wide spread gullies show high sediment
yield to reservoirs in the study area. Most of the studied
L. Tamene et al. / Geomorphology 76 (2006) 76–9188
catchments are characterized by rugged terrain in their
upslopes and predominantly cultivated land or bare
grazing areas in their lower slopes. The high HD accel-
erates flow energy, and the poor surface cover offers the
minimum protection. As a result, most of the flood-
plains are being eroded contributing sediment genera-
tion and enhancing efficient transport of sediment from
upslopes.
The main purpose of multiple regression analysis is
to assess the relationship between several independent
or predictor variables and a dependent variable. Once
the relationships are established, the equations can be
used to predict the status of a dependent variable in
relation to the independent variables. The equations
presented in Table 7 and Eq. (7) or Eq. (8), for example,
can be used to predict SSY of other similar catchments.
However, detailed analysis for a larger number of sites
will be required to establish a robust prediction model
of annual sediment yield.
5.4. Sediment yield controlling factors and
management implications
Based on the rate of sediment deposition (Table 4)
and the reservoir live storage capacity shown in Table 1,
most of the reservoirs will be filled with sediment within
less than 50% of their intended service time. For in-
stance, the Adikenafiz, Gerebmihiz and Grashito reser-
voirs have lost over 40% of their live storage capacity
within about 25% of their expected service time. The
above two and the Gindae and Maidelle reservoirs have
also lost more than 100% of their dead storage capacity
in less than a quarter of their expected life time. There-
fore the planned food security improvement scheme will
be under threat unless relevant preventive measures are
put in place.
The benefits of proper catchment management and
conservation can be demonstrated by Fig. 5. Three
reservoirs with comparatively proper catchment man-
agement and conservation (pink symbols in the upper
left graph) show relatively low sediment deposition
despite high terrain potential for erosion.
This study shows that terrain-from, surface lithology,
gullies and LUC play significant roles in determining
SSY variability, and gully erosion may be the most
important. The four catchments with high SSY have a
very high problem of gully erosion, whereas three
catchments with low SSY have very limited evidences
of gully erosion. Shibru et al. (2003) also shows that
gullies alone produced an annual soil loss rate of 25 t
ha�1 year�1 in eastern Ethiopia. Therefore, attention
needs to be paid to the rehabilitation/stabilization of
gullies and their banks, and prevent their destabilization
due to livestock trampling, although soil loss assess-
ment programs only infrequently take account of the
contribution of gully erosion (e.g., Liggitt and Fincham,
1989; Poesen et al., 2003). The location of reservoirs
should also be carefully determined to avoid severe
siltation. Reservoirs are often located at the confluences
of a few streams to efficiently collect water. However, if
one of the streams is characterized by heavy gullying, it
may lead to a higher risk of accelerated siltation.
Because only lumped attributes of catchments were
used in this study, the spatial distribution of the factors
responsible for siltation within each catchment could
not be examined. However, we could specify the terrain
attributes that land managers should focus on to tackle
the rapid sedimentation of the reservoirs. For instance,
covering the upland non-cultivable areas with vegeta-
tion looks effective (Tamene, 2005).
6. Conclusions
Identification of the major causative factors of ero-
sion that accelerate siltation in reservoirs is necessary to
guide targeted management. Against this background,
different statistical analyses were performed to assess the
role of different catchment attributes in the siltation of
reservoirs in the drylands of northern Ethiopia. The
results show that pronounced terrain steepness, easily
detachable slope material, poor surface cover, and
gullies accelerate siltation in the reservoirs. These factors
often co-exist, but the distribution of gullies is the most
influential. Some of the catchments show lower sedi-
ment yields despite their steep terrain because catchment