here - Mekelle University

About the Journal
Momona Ethiopian Journal of Science (MEJS) is established in December 2008 by Mekelle
University to promote the cause of advanced study and research in all branches of physical,
biological, chemical, geological and computational sciences. It provides a platform for the
scientists to exchange ideas among themselves and interact with industry. It is a peer reviewed
international interdisciplinary electronic science ejournal and will be available on the WWW
with free access. It is published in English, half yearly and constitutes a volume a year. It
operates in a manner similar to a conventional paper journal but with no cost to the user to read
or publish manuscripts.
MEJS provides a means for communication and exchange of research results and other more
general information among scientists and others interested in various aspects concerning Earth
Science, Physics, Chemistry, Biology, Mathematics and related disciplines. It tries to provide an
international forum for the scientific discussion of techniques and approaches leading to a higher
degree of efficiency in scientific knowledge and its application on the ground. At the same time,
MEJS being ejournal it facilitates rapid review, speedy processing of manuscripts and timely
dissemination of information.
Background
Publication of research results and exchange of information is essential to the growth of any
discipline. There are several International and National journals available for scientists to publish
their research results. Among them some are highly specialized and dedicated to a particular
profession or discipline. MEJS tries to provide platform for scientists related to Natural and
Computational Sciences in general and Earth Scientists, Physicists, Biologists, Chemists and
Mathematicians in particular and covers all aspects from basic to applied nature. Contrary to the
paper journal, MEJS is launched as an EJOURNAL primarily with the desire to expand its base
to more and more countries around the world and serve the international scientific community by
providing a low cost journal. The idea of ejournal is in tune with the recent advances made in the
electronic publishing sector which has resulted in the publication thousands of journals
electronically globally and available on the World Wide Web (WWW). This makes the present
initiative possible and provides a suitable mechanism for widespread dissemination of
information.
1
Focus and Scope of the Journal
Momona Ethiopian Journal of Science (MEJS) focuses on the research related to the Earth,
Physical, Chemical, Biological and Computational Sciences. The scope of the journal includes
all aspects of geology, chemistry, physics, biology and mathematics. It publishes original peerreviewed scientific papers covering both basic and applied aspects of science including
interdisciplinary. The ejournal is devoted primarily to original research papers that present work
of national or international significance. It will also include short communications, invited
review papers, general review articles, book reviews, letter to the Editor, comments and critique
of published material, information related to conferences and any other relevant topics from time
to time. Manuscripts from scientists working in academic institutions, industry, companies and
as consultants in Ethiopia or elsewhere, having national and international appeal are accepted for
publication. The number of papers published is going to vary depending on the number
submitted and approved by the peer review process.
Peer Review
The manuscripts will be reviewed initially by the Editorial Board. Later by two experts,
specialists in the respective fields working in academic institutions, industries, companies and as
consultants either in Ethiopia or elsewhere. The Board will identify suitable experts preferably
international on the basis of availability. If the comments of the reviewers related to quality, size
and type of the manuscript differ drastically then a third reviewer will be consulted. The majority
opinion will prevail. In extreme or special cases the Board will take the final decision.
Quality Assurance
Quality being critical and essential for survival and sustenance of the journal, the Board will
make all possible efforts to establish a reputation as a quality journal and takes responsibility for
quality control. The Editor-in-Chief is responsible for overall operation of the ejournal and the
Board for reviewing manuscripts, selecting qualified and expert reviewers and to maintain
quality standards. The Board in view of keeping the focus of the journal and gaps in the scientific
knowledge will make sure broad geographic and subject matter representation.
2
Submission of Manuscripts
All manuscripts in three copies should be submitted to the Editor-in-Chief or any member of the
editorial board (MEJS, College of Natural and Computational Sciences, P.O. Box. 231, Mekelle
University, Mekelle, Ethiopia). In the case of electronic form which is preferred, the manuscript
should be in single file (inclusive of text, tables and figures) either in PDF or Word format
without any password control and sent to [email protected]. The manuscript presented as a paper
in any proceeding, or submitted to other journal for publication needs clearance before it is
considered for publication in MEJS. A copyright form duly signed by the author or authors
should be sent to the Editor-in-Chief once the manuscript is accepted for publication.
Dr. Tadesse Dejenie
Editor-in-Chief
© College of Natural and Computational Sciences, Mekelle University
3
Compositional Differences between Felsic Volcanic Rocks from the Margin
and Center of the Northern Main Ethiopian Rift
*Kurkura Kabeto1, 2, Yoshihiro Sawada2, Barry Roser2
Department of Earth Science, College Natural and Computitional Sciences, Mekelle University,
P.O.Box 231, Mekelle, Ethiopia. (*[email protected]/ [email protected])
1
2
Department of Geosciences, Shimane University, Matsue 690-8504, Japan
ABSTRACT
Pliocene felsic rift margin and Quaternary rift center volcanic rocks from the northern Main
Ethiopian Rift (MER) exhibit contrasts in major and trace element contents and Sr-Nd isotopic
ratios. Quaternary rift center felsic volcanic rocks are mainly peralkaline trachytes and rhyolites,
whereas Pliocene felsic rift margin volcanic rocks are represented by benmoreites, weakly
peralkaline trachytes and rare rhyolites.
Most of the felsic rift margin volcanic rocks have greater Al2O3, K2O, Nb, Zr, Rb, and Sr, and
lesser CaO, Zr/Nb, and CaO/Al2O3 than rift center volcanic rocks. These contrasts may have
been inherited from differences in the compositions of their parental basic magmas, which were
produced by variable degrees of partial melting.
In both series, the felsic volcanic rocks generally have higher initial Sr- isotopic (0.7038-0.7073)
ratios than their basic equivalents (0.7035-0.7046). Nd- isotopic ratios of most felsic rift center
samples (0.5129-0.5126) are similar to their associated basic volcanic rocks. In contrast, the Ndisotopic ratios (0.5128-0.5124) of felsic rift margin volcanic rocks are commonly lower than
their companion basic volcanic rocks (0.512806-0.512893), and are relatively lower than rift
center equivalents. The elemental and Sr-Nd isotopic compositions of the volcanic rocks suggest
that fractional crystallization from differing basic parents accompanied by a limited assimilation
(AFC) was the dominant process controlling the genesis of the MER felsic volcanic rocks.
Keywords: Ethiopia; Northern Main Ethiopian Rift; Bimodal Volcanism; parental difference;
Sr-Nd Isotopes, Fractional Crystallisation
1. INTRODUCTION
The Main Ethiopian Rift (MER), the Southwestern Ethiopian Rift Zone (SWERZ), the Tana Rift
and the Afar region represent the northernmost part of the East African Rift System (Fig. 1). The
Ethiopian volcanic province is dominated by up to 300,000 km3 of generally fissure-fed MidTertiary basic (SiO2 < 53 wt.%) volcanic rocks, and minor associated felsic (SiO2 > 53 wt.%)
products. However, the proportion of felsic products in the Ethiopian rift valley itself is high,
reaching about 90% of the total volume (Mohr, 1992).
4
Between 45 and 22 Ma, volcanic activity in the Ethiopian plateau (Figs. 1a and b) was
characterized by fissural flows. Central volcanoes covered the fissural flows beginning at about
30 Ma and 15-13 Ma, and erupted intermittently into the Pleistocene (Morton et al., 1979;
WoldeGabreil et al., 1990; Wolde, 1996; Stewart and Rogers, 1996; Chernet et al., 1998; Pik et
al., 1998; Ayalew et al., 2002). Recent geochemical and isotopic studies have focused on the
Oligocene-Miocene to Quaternary basic-felsic volcanism that accompanied the formation of the
MER. These studies have proposed the involvement of distinct mantle components in various
proportions, and the importance of the Afar mantle plume and lithospheric mantle in the sources
of the basic lavas (Fig. 1).
Figure 1. Maps showing the
location of the study area. a) Index
map showing surface expression of
the Ethiopian rift system and
volcanic cover (Wolde, 1996) and
the Ethiopian Plateau (EP, Stewart
and Rogers, 1996). b) Volcanic and
tectonic structures of the MER. The
white ellipses are the study areas. N
(northern), C (central), and S
(southern) sectors of the MER;
SWERZ is the southwestern
Ethiopian Rift Zone. c) Sample
sites, felsic volcanic centers and
volcanic and tectonic structures of
the Addis-Nazreth region. YTVL
Yerer Tulu Welel Volcano Tectonic
Lineament from Abebe et al. 1998.
Although the previous studies have
provided
a
well-documented
framework for the genesis of the
basic
magmas
in
MER,
the
associated felsic magmas are not
well
studied.
Recently,
several
geochemical and isotopic studies of
felsic products have been carried out in the rift center at Debre Zeit, Gedemsa, and Asela-Ziway
5
(Fig. 1) (Gasparon et al., 1993; Peccerillo et al., 1995; Trua et al., 1999; Abebe et al., 1998;
Boccaletti et al., 1999), on volcanic rocks which are mostly younger than 2 Ma. The felsic
volcanic products at the rift margin (Addis Ababa) and rift center (Nazreth) regions and their
compositional variations are not yet well studied.
This paper presents new elemental and Sr-Nd isotope data for selected volcanic sequences from
the northern sector of the MER (Fig. 1). Based on the petrographic, geochemical, and isotopic
characteristics of the volcanic units, we describe the compositional differences between the
volcanic rocks at the rift margin and rift center, and discuss the petrogenetic relationships
between basic and felsic magmas in order to assess the influence of basic parents and continental
crust in the genesis of the felsic melts.
2. GEOLOGICAL SUMMARY OF THE ADDIS ABABA AND NAZRETH REGIONS
The study area is located in the center of the Ethiopian dome (Fig 1), and contains volcanic
sequences that are directly related to the northern Main Ethiopian Rift (MER) activity. Reevaluation of seismic refraction data for the region by Makris and Ginzburg (1987) revealed
thinning of the crust from 44 km thick at Addis Ababa to 30-26 km in the center of the MER to
the east. This led Wolde (1996) to regard the volcanic sequences at Debre Zeit and to its east as
rift center, and Miocene to Pliocene volcanic rocks in the Addis Ababa region, west of Debre
Zeit as rift margin.
On the basis of whole-rock and mineral K-Ar and
40
Ar-39Ar ages, Morton et al. (1979) and
Chernet et al. (1998) found that the volcanic cover extending from Addis Ababa to Nazreth
showed age progression from 22.8 Ma in plateau basalts to 0.21 Ma in the rift center volcanic
rocks. At the base of the rift margin volcanic rocks, alkaline-transitional basalt (22.8 Ma) of the
Plateau unit is in fault contact with the overlying Intoto unit (22.2-22 Ma) (Morton et al. 1979;
Chernet et al. 1998). The Intoto unit consists of trachyte-rhyolite flows and associated
ignimbrites at its base (22.2 Ma) and plagioclase phyric basalt (22 Ma) in its upper part. The
Early Miocene Plateau and Intoto units represent bimodal volcanic rocks, which were formed
during a localized terminal episode following the massive Oligocene fissural basalt activity of
the northwestern Ethiopian plateau and are present at the rift margin in the Addis Ababa region
(Morton et al., 1979; Begazi et al., 1993; Chernet et al., 1998). Because of their close
realationship, the Plateau and Intoto units are here collectively referred to as the Plateau unit
6
(22.8-22 Ma). The geochemical data for the Plateau unit is not discussed in this paper because
they are pre-rifting eruptions.
Volcanic activity resumed at the rift margin after a considerable hiatus between 22-10 Ma, by
eruption of transitional-alkaline basalts of the Addis Ababa unit (9-7 Ma). This was followed by
the welded trachytic tuffs (5.1 and 3.3 Ma) of the Nazreth unit, which are thought to have been
derived from the mostly trachytic Pliocene (4.6-3.09 Ma) Wechecha, Menagesha, Furi, and
Yerer volcanoes (Fig. 1). These centers are collectively designated as the Wechecha unit
(Chernet et al., 1998).
Predominantly felsic volcanic products were erupted in the Debre Zeit and Nazreth regions in the
rift center between 2.0-0.20 Ma. This sequence consists of the Nazreth, Keleta, Boku-Tede,
Bofa, Dera-Sodere, Gedemsa, Boseti, and Melkasa units in ascending order (Fig. 1; Boccaletti et
al., 1999). For simplicity, the felsic volcanic rocks of Nazreth, Keleta, Boku-Tede, and DeraSodere units are here grouped as the Nazreth unit, whereas the basaltic rocks of the Bofa and
Melkasa units are collectively termed the rift center basic volcanic rocks.
3. CLASSIFICATION AND PETROGRAPHIC SUMMARY
Most of the samples investigated here were fresh, and collected from lava flows, except for a few
welded ignimbrites from the rift center (Table 1). According to the TAS classification diagram
(Le Bas et al., 1986) the rock types in both areas range from basalt to rhyolite, but trachytes and
rhyolites predominate (Fig. 2). The rocks fall in numerous compositional fields, making
terminology cumbersome. Therefore, a restricted set of terms has been adopted here. Rock
samples with SiO2 < 53 wt.% are defined as basic, whereas samples with SiO2 > 53 wt.% are
classed as felsic. In common with many bimodal suites in continental rifts and oceanic islands,
the basic rocks at the rift margin are generally alkaline or transitional, whereas rift center
equivalents are typically transitional (Wolde, 1996) (Fig. 2). Strongly felsic alkaline rocks at the
rift margin and center are weakly peralkaline and strongly peralkaline, respectively (Table 1).
According to Macdonald (1974), most of the peralkaline felsic volcanic rocks are commendites
with rare pantellerites. However, most of the Gedemsa unit samples are pantelleritic (Peccerillo
et al., 1995) (Figure not shown).
7
Petrographically, the basic rocks in both rift margin and center are usually aphyric. Porphyritic
samples on the other hand are rare and contain about 15-25% phenocrysts. The phenocryst
minerals are olivine, clinopyroxene, and plagioclase with or without opaques. The groundmass
of the basic lavas consists of the above phenocryst phases and accessory glass, zircon, apatite,
and titanite, along with secondary alteration products including sericite/carbonate, iddingsite, and
hematite. Loss on ignition values in these samples differ little from the unaltered samples (Table
1).
Figure 2. Classification of volcanic
rocks of the northern MER according
to the scheme of Le Bas et al. (1986).
The alkaline-sub-alkaline boundary is
from Irvine and Baragar (1971).
Additional data for the volcanic rocks
from Abebe et al. 1998), Chernet
(1995) and Peccerillo et al. (1995).
The felsic volcanic rocks in the rift
margin
are
relatively
aphyric
compared to the basic rocks, and
generally show trachytic textures.
The phenocryst minerals (<15%)
are commonly plagioclase (albiteoligoclase) and anorthoclase, and
lesser sanidine, nepheline, alkali amphibole (rebeckite), and aegirine-augite (Chernet, 1995;
Abebe et al., 1998). Plagioclase and anorthoclase are the common phenocryst minerals in the rift
margin felsic volcanic rocks. Groundmass contains plagioclase, K-feldspars, opaques, augite,
and aegirine-augite. Some feldspar grains show melt inclusions along their margins, and rarely
amphiboles are altered.
The felsic rift center volcanic rocks contain similar type of minerals as their rift margin
equivalents. Majority of the samples are aphyric, but few samples containing up to 35%
phenocryst phases with both felsic and ferromagnesian minerals forming the phenocryst and
groundmass phases. Felsic minerals are usually sanidine, quartz, sodic plagioclase, and
anorthoclase, whereas the ferromagnesian minerals are aegirine-augite, sodic-amphibole, olivine,
8
and opaques. Both the lavas and ignimbrites contain the same minerals, but differ in their
textures. The lavas commonly show perlitic cracks/spherulitic groundmass textures, whereas
ignimbrites show eutaxitic textures with dominantly vitrophyric fiamme groundmasses.
4. GEOCHEMISTRY
4.1. Analytical Procedures
Out of 125 collected samples, 75 selected samples from each unit were analyzed for major and
trace elements and Sr-Nd isotopes (Table 1) in the Department of Geoscience, Shimane
University, Japan. Samples were crushed in a tungsten carbide ring mill (Roser et al., 1998), and
dried at 1100C for 24 hours. No significant Nb or Ta contamination was present in the carbide
ring mill compared to that ground in agate (Roser et al., 1998). The remaining XRF data will be
provided upon request. Major and trace element analyses were performed using glass beads
prepared either by fusing 0.7g of rock powder mixed with 3.5g of Li2B4O7 (Norrish and Hutton,
1969) or by mixing 1.8g sample with 3.6g alkali flux (LiBO2:LiB4O7 = 1:4) (Kimura and
Yamada, 1996). Analyses were made using a Rigaku RIX 2000 X-ray fluorescence spectrometer,
using conventional peak/background methods, with calibration against a suite of Geological
Survey of Japan (GSJ) and USGS rock standards. The reproducibility was monitored with
appropriate international standards JB01, JB02, and JB03, and was within +10% for all elements
with concentrations higher than 10 ppm.
Sr- and Nd- isotopic measurements and analyses of Sm and Nd (by isotope dilution) were carried
out using Finnigan MAT262 thermal ionization mass spectrometer, following the methods of
Iizumi et al. (1994, 1995). Measured
87
Sr/86Sr and
143
Nd/144Nd ratios were normalized to
87
Sr/88Sr = 0.1194 and 146Nd/144Nd = 0.7219. During the analysis of unknowns, measurements of
87
Sr/86Sr and 143Nd/144Nd in NBS 987 and JMC standard samples were 0.71026 ± 18 (2σ, n = 10)
and 0.51196 ± 10 (2σ, n = 10), respectively (Table.1)
4.2. Major Element Data
Whole rock analyses and Sr-Nd isotope data of representative samples are listed in Table 1. The
remaining data are available upon request. The basic (SiO2 < 53 wt.) rocks have MgO contents
between 4 and 14 wt.%, Mg-numbers (mg#) between 50 and 74, Ni ≤ 380 ppm and Cr ≤ 822
ppm. Most of these values suggest that very few rocks represent the primary mantle melts (Sato,
9
1977; Wilson, 1989). Majority of the samples having mg# between 44 and 50 suggest that they
do not represent primary magma compositions (Table.1).
15
MgO
20
10
15
10
Figure 3. Plots of selected
major elements (wt.%) against
differentiation index (SiO2) for
northern MER volcanic rocks.
Symbols as in Fig. 2.
5
5
Selected major and trace
Al2O3
0
P2O5
2
0
1.5
elements
plotted
against
SiO2 are shown in Figs. 3, 4
1.0
and 5. With the exception of
two intermediate samples,
Wt.% oxide
1
0.5
TiO2
all the felsic rocks have SiO2
> 57 wt.%, and an apparent
0
12
0.0
8
9
6
6
4
3
2
∑FeO and CaO generally
0
12
0
decrease
9
4
SiO2
gap
(53-57
wt.%)
exists between the basic and
felsic volcanic rocks. In all
volcanic sequences MgO,
FeOtot
Na2O
2
3
CaO
0
45
65
SiO2
75 45
55
65
SiO2
Na2O
contents
increase from about 2 wt.%
in the basic volcanic rocks
K2O
55
increasing
SiO2 and K2O (Fig. 3).
However,
6
with
0
75
to about 7 wt.% in the felsic
rocks (at 65 wt.% SiO2), and
then decreases to 4.2 wt.% increasing SiO2 (>70%) in most of the felsic samples. In the Boseti
unit, however, Na2O continues to increase in some of the felsic rocks. Al2O3 contents in basic
samples range from 13.0 to 21.3 wt.%, maintain similar values (13.3-17.5 wt.%) in felsic
samples through to 65 wt.% SiO2, and then decrease to values as low as 9.2 wt.% at 75 wt.%
10
SiO2. P2O5 and TiO2 contents also tend to increase from basic to intermediate compositions (up
to 55 wt.% SiO2), and then sharply decrease up to 65 wt.% SiO2 and remain flat (Fig. 3).
Figure 4. Plots of selected trace elements (ppm) against differentiation index (SiO2 wt.%) for northern
MER volcanic rocks. The lines on the Zr, Nb, Sr and Rb versus SiO2 illustrate the compositional contrasts
between the volcanic rocks in the rift center and margins. The shaded area on the Zr and Nb plots indicate
where samples from the Furi and Yerer Mountains overlap samples from the rift center. Symbols as in
Fig. 2.
11
Compositional differences exist between the rift margin and rift center volcanic rocks (Table 1;
Fig. 3). For example, CaO contents in the rift center basic rocks (9.4-11.3 wt.%) are generally
higher than most of the rift margin equivalents (7.7-10 wt.%), except in few clinopyroxenephyric samples from the Addis Ababa unit that are younger (10.9-12.5 wt.%). In contrast, Al2O3
contents of most rift margin basic volcanic rocks are higher (14.8-18.6 wt.%) compared to the
rift center equivalents (13-17.5 wt.%).
The contrasts in CaO and Al2O3 seen in the basic lavas of the two regions are also apparent in the
felsic volcanic rocks. Al2O3 contents are greater in most of the felsic rift margin volcanic rocks
than in the rift center, whereas CaO contents are generally greater in the rift center (Fig. 3). K2O
concentrations are generally greater and ∑FeO contents lesser in the rift margin felsic volcanic
rocks than in the rift center. However, the above compositional differences become less
pronounced in samples with SiO2 contents > 67 wt.%. Rift center felsic samples are generally
richer in SiO2 than most from the Wechecha (Pliocene) unit at the rift margin (Fig. 3)
4.3. Trace Element Data
In both the rift margin and rift center volcanic sequences, compatible elements such as Ni, Co,
and Cr decrease in abundance with increasing SiO2 (Table 1). This indicates the influence of
olivine, clinopyroxene, and Fe-Ti oxide fractionation. Zr, Nb, Y, and Rb, being incompatible
elements show increase in their concentration with fractionation within each rock series, albeit
with some scatter (Fig. 4). Sr and Ba contents in the felsic volcanic rocks generally decrease with
fractionation, representing compatible behavior, though they too show wide variations.
Most rift margin basic lavas have greater Zr and Nb contents than do the rift center basic lavas.
Moreover, at given SiO2 content Sr is much more enriched in rift margin basic lavas than in their
rift center equivalents (Fig. 4, Table 1). Zr, Nb, Sr, and Rb are also generally more enriched in
their rift margin felsic rocks than in Quaternary rift center felsic equivalents. However, Y and Ba
abundances mostly overlap, as does Sr at > 65 wt.% SiO2. Among the Wechecha felsic samples,
high Zr and Nb are observed in the Menagesha and Wechecha mountain samples (Table 1),
whereas relatively low Nb and Zr contents characterise samples from the Furi and Yerer
mountains (Fig. 4).
4.4. Nd- and Sr- Isotope Data
Twenty-eight samples spanning the compositional range from the least fractionated MgO-rich
rocks to the most fractionated rhyolite samples were analysed for Sr- and Nd- isotopes. Isotopic
12
ratios as shown in table 1 are plotted in figure 5, together with data for 15 samples from Chernet
(1995) and Abebe et al. (1995). Initial 143Nd/144Nd ratios in the basic lavas range from 0.512812
to 0.51289, and from 0.51237 to 0.51286 in the felsic volcanic rocks. Initial 87Sr/86Sr ratios range
from 0.70349 to 0.70456 in the basic lavas, and from 0.70446 to 0.70783 in the felsic volcanic
rocks. The rift margin basic volcanic rocks have higher Nd- and lower Sr- isotopic ratios than rift
center equivalents (Table 1; Fig. 5).
In contrast, felsic volcanic rocks from the rift margin have lower Nd- isotopic ratios than most of
their rift center equivalents. However, their Sr- isotopic compositions are equally variable. The
felsic volcanic rocks extend from the isotopic range of the basic volcanic rocks towards higher
87
Sr/86Sr (Fig. 5). Most of the felsic rift center volcanic rocks lie above the Debre Zeit field
(Gasparon et al. 1993), whereas the felsic rift margin volcanic rocks partly overlie it. Rift center
felsic volcanic rocks overlie the field defined by equally felsic lavas from northern Kenya
(Kabeto et al., 2001) (Fig.5).
Figure 5. Initial 87Sr/86Sr and
143
Nd/144Nd ratios for volcanic
rocks of the study area, compared
to west central (WC) Afar and
MER (Hart et al., 1989), Debre
Zeit (Gasparon et al., 1993), south
Ethiopian (Stewart and Rogers,
1996), Asela-Ziway & Chilalo
(Trua et al., 1999), northern
Kenyan volcanic rocks (Kabeto et
al., 2001b), and the East African
Carbonatite
Line
(EACL).
Measured 143Nd/144Nd ratio is used
for some samples (Table 1).
Symbols as in Fig. 2.
5. DISCUSSION
It is well established that compositional differences in parental basaltic magmas are reflected in
the compositions of felsic melts (Wilson et al., 1995; Panter et al., 1997). Fractional
crystallization of basaltic magmas with some crustal assimilation and partial melting of basic
13
lower crust/underplated igneous rocks were proposed as dominant processes for generation of
felsic melts in the MER rift center (Fig. 1; Gasparon et al., 1993; Abebe et al., 1998; Peccerillo et
al., 1995; Trua et al., 1999; Boccaletti et al., 1999). The influence of different parental magma
compositions and the processes involved in the genesis of felsic melts in the northern sector of
the MER are discussed below.
5.1. Influence of Parental Magma Compositions
The general, elemental contrasts between the rift margin and rift center volcanic rocks discussed
above (Figs. 3 and 4; Table 1) are clearly evident on CaO/Al2O3 vs SiO2, Zr vs Nb, and Zr/Nb vs
Zr plots (Fig. 6). At a given SiO2 content most of the basic rocks (SiO2 < 53 wt.%) from the rift
center are displaced towards higher CaO/Al2O3 ratios (Fig. 6a). Few basic rocks from the rift
margin have CaO/Al2O3 ratios comparable with rift center equivalents, and in both groups the
ratio decreases with fractionation. CaO/Al2O3 ratios remain higher in the felsic rift center rocks
(SiO2 > 53 wt.%) than in rift margin equivalents, but a few felsic samples from the Wachecha
(Furi and Yerer Mts.) unit overlap (Table 1). This may be due to similar fractionating phases
controlling their evolution, or indicate that they were derived from compositionally similar basic
parents.
Zr and Nb contents show a well defined linear correlation (Fig. 6b), and both being incompatible
elements increase with fractionation (Kamber and Collerson, 2000; Kabeto et al., 2001b).
Constancy of trace element ratios between basic and felsic melts (e.g., Zr/Nb; Fig. 6b and c) is
often cited as strong evidence that fractional crystallization has been the dominant process in
their evolution (Weaver, 1977; Wilson, 1989). At given Zr content most rift margin samples
show higher Nb contents than do the rift center samples. Kamber and Collerson (2000) have
indicated that Nb is more sensitive to variations in degrees of partial melting than Zr, and hence
can be used to decipher the influence of variable degrees of melting. In this regard, some rift
margin samples from the Furi and Yerer Mountains have Nb contents as low as the rift center
samples (Table 1; Figs. 4 and 6b). It is evident that Yerer and Furi mountains are
compositionally closer to rift center composition than the Wechecha and Menagesha Mountains.
Furthermore, samples from Yerer and Furi are younger and have a narrower age range (2.03-4.04
Ma) than the Wechacha and Mengasha Mountains samples (3.09-6.63 Ma) (Chernet et al., 1998)
(Fig.6).
14
Plotting Zr/Nb vs Zr (Fig. 6c) also shows that felsic volcanic rocks of the study area fall into two
clusters. Based on their Zr/Nb ratios, most samples from the rift margin plot along Zr/Nb ≤ 5,
whereas rift center volcanic rocks, and the few samples from the Yerer and Furi Mountains with
low Nb contents cluster along Zr/Nb ≥ 6. Similarly, Zr/Nb ratios in the basic rocks of the two
sequences also vary (Fig. 6c). The rift center basic rocks have Zr/Nb ratio ≥ 5, whereas most of
the basalts from the Addis Ababa area that are thought to be parental to the rift margin felsic
volcanic rocks, and a few rift center basic rocks; have Zr/Nb ≤ 5.
Zr/Nb ratios in volcanic rocks may also reflect crustal contamination,titanite fractionation or
variation in degree of partial melting (Wilson et al., 1995; Kamber and Collerson, 2000; Kabeto
et al., 2001b). For example, basaltic sample ET1201 from the Plateau unit (Table 1) has a Zr/Nb
ratio of 13.0. ET1201 has a very low initial Nd- isotopic ratio (0.51222 ± 8) compared to other
basic lavas with lower Zr/Nb ratios, which may indicate that the higher Zr/Nb ratio reflects
crustal involvement (Kabeto et al. 2001). Hence, we consider the rift center volcanic rocks
(Zr/Nb > 6), and those rift margin volcanic rocks with Zr/Nb ratios > 5, to reflect either crustal
input, titanite fractionation (Fig. 6c) or different sources.
The clear differences in crustal thickness, extensional tectonics, age of volcanic activity (Morton
et al., 1979; Makris and Ginzburg, 1987; Wolde, 1996; Abebe et al., 1998; Boccaletti et al.,
1999), and compositions of volcanic rocks in the two regions suggest that they were derived
from different parantal magmas. The role of parental basic magma compositions in the rift
margin and center felsic melts must be considered in the light of elevated or depleted absolute
trace and major element abundances and the degree of silica saturation. Furthermore, the effects
of crustal contamination must be accounted for. For example, the silica saturation in the felsic
volcanic rocks in the rift center could be produced by substantial contamination of the basic
magma that is parental to most of the Pliocene rift margin eruptives by silica-rich crust.
However, this is an unsuitable mechanism to produce the felsic volcanic rocks from the rift
center, because open-system behavior would produce higher incompatible trace element
abundances in the more contaminated series (DePaolo, 1981; Nelson and Davidson, 1993). This
is not observed here. Moreover, the Nd- isotopic compositions of the felsic rift center volcanic
rocks lie within the range of basic volcanic rocks in the region (Fig. 5). The lowest Nd- isotopic
ratios are noted only in those felsic volcanic rocks at the rift margin that show higher degree of
contamination.
15
Figure 6. (a) Plot of CaO/Al2O3 ratio
against SiO2, showing the two
apparent evolution trends for the
northern MER. Most rift margin
volcanic rocks plot at lower ratios. (b)
Linear correlation between Zr and Nb
contents. Samples generally plot along
Zr/Nb = 5 and Zr/Nb = 7, which may
indicate different sources or crustal
input (see text for discussion). (c)
Zr/Nb vs Zr plot for the volcanic
rocks. Arrows show assumed AFC and
differentiation trends (FC) for rift
margin and center sequences from
different basic parent. The shaded area
on (b) and (c) indicate samples from
Furi and Yerer Mountains overlapping
the samples from the rift center (see
text for discussion). Data for crustal
rocks from Davidson and Wilson
(1989). Symbols as in Fig. 2.
Alternatively, contamination of basic
magma, thought to be the parent for
rift center felsic products by bulk
assimilation of silica deficient crustal
material (amphibolitic or basic crust?)
must also be considered. This would
make the rift margin Wechecha unit
magmas more contaminated than rift
center felsic magmas (Nelson and
Davidson, 1993). We consider this to
be unlikely, as it cannot explain the
relative depletion of the compatible
elements MgO and CaO in the
Wechecha samples (Fig. 3).
It appears that differing degrees of
partial melting of the mantle source
16
provide the best explanation for the compositional differences seen between the two volcanic
sequences (Nelson and Davidson, 1993). A lower degree of melting, deeper partial melting
(Kushiro, 1968; O’Hara, 1968) or melting in the presence of CO2 (Davies and Macdonald, 1987)
could easily have produced incompatible element-enriched magma that differentiated to produce
the most felsic rift margin samples. In contrast, higher degrees of melting could produce
incompatible element depleted magmas that subsequently differentiated to produce the rift center
felsic volcanic rocks.
It has been suggested that basic lavas produced by low degrees of mantle partial melting have
high incompatible element contents (e.g., Zr, Nb, Y, K, and Rb), high Al2O3, and low SiO2 and
CaO (Tatsumi and Kimura, 1991; Wolde, 1996; Kabeto et al., 2001b). In contrast, basic lavas
that are produced by high degrees of partial melting have lower incompatible element and Al2O3
contents, but higher CaO and SiO2. Wolde (1996) has shown that alkali basalts produced by a
low degree of partial melting are common in the western part of the rift and the margin, and are
found only locally east of Debre Zeit (the rift center). In contrast, transitional basalts which
originate from higher degrees of partial melting are commonly found within the rift center,
where thinning of the crust has been identified from seismic refraction studies (Makris and
Ginzburg, 1987). Moreover, Abebe et al. (1995) have suggested that the degree of alkalinity in
basaltic melts increases away from the rift center.
In the northern MER, the felsic melts in the rift center generally have lower Al2O3, K2O, Zr, Nb,
Y, Rb, and Sr and higher SiO2 and CaO than most rift margin equivalents (Figs. 3 and 4). Hence,
the major and trace element contrasts in the felsic products between the two regions could
originate from compositional differences in their basaltic parents. Wilson et al. (1995)
demonstrated that compositional differences between silica-undersaturated and oversaturated
felsic melts in the continental magmatism of the Central Massif (France) were controlled by
subtle compositional differences between their respective basic magmas. In line with this
suggestion Kabeto et al. (2001) have considered that silica-saturated (basalt-trachyte) and silicaundersaturated (basanite-phonolite) lineages in the northern Kenyan rift sector originated from
compositional differences in their parental basic magmas, and both felsic by differentiation
combined with a little assimilation. This is also likely to be the case here.
17
5.2. Fractional Crystallization and/or Degree of Partial Melting
The general decrease of Ni, Cr, Sr, Ba, MgO, CaO, ∑FeO, TiO2, and P2O5 with increasing SiO2
(Figs. 3 and 4; Table 1) indicates that the geochemical evolution of these volcanic rocks was
governed by fractionation of olivine, clinopyroxene, Fe-Ti oxide, feldspars, and apatite. TiO2 and
P2O5 also show to decrease at the same SiO2 content in all rock suites, indicating simultaneous
apatite and Fe-Ti oxide fractionation. Moreover, the general increase in Zr, Nb, Rb, Y, and K
concentrations with increasing SiO2 is also consistent with fractional crystallization from a
similar basic parent, to produce the mugearites, benmoreites, trachytes, and rhyolites of both
regions. The well-defined linear correlation between Zr and Nb contents in the sequences (Fig.
6b and c) also suggests that fractional crystallization was a dominant process.
The Sr- and Nd- isotopic ratios of the felsic lavas (Fig. 5) also do not lie wholly within the
isotopic range of their associated basic lavas, indicating that fractional crystallization was not the
only process responsible for their genesis.
Genesis of the felsic MER volcanic rocks by anatexis of the upper continental crust must be
discarded on the basis of geochemical characteristics. The Afro-Arabian continental crust
displays a wide range of isotopic composition (Davidson and Wilson, 1989; Hegner and
Pallister, 1989; Möller et al., 1998) which is dissimilar to the felsic MER volcanic rocks (Table
1). The Nd- isotopic compositions of most felsic MER volcanic rocks are also more radiogenic
than mean upper crust. Hence, partial melting of the upper continental crust cannot be the source
for the felsic melts. Low degrees of partial melting of basic lower crust and/or underplated basic
magmas as possible source for the felsic melts can be tested by batch melting modeling
calculations (Shaw, 1970; Skjerlie and Johnston, 1993) performed on Rb vs Sr (not shown).
Based on the Sr- isotopic variations between the basic (0.70349 to 0.70456) and felsic (0.70446
to 0.70783) volcanic rocks (Table 1; Fig. 5), it is possible that fractional crystallization might not
be the only process responsible for the generation of felsic lavas. Even if the data appears to be
explained well by fractional crystallization from a basaltic parent, the fractionation stage could
be an open system, and hence some assimilation of crustal material is possible. The estimation of
the extent of contamination by crustal material is complicated by the diversity shown by AraboAfrican basement rocks (Hegner and Pallister, 1987; Davidson and Wilson, 1989; Möller et al.,
1998).
18
The Sr-Nd isotopic compositions of the felsic volcanic rocks from the rift margin also suggest
that assimilation of crustal material has occurred may be limited. Initial Nd- isotopic ratios,
which are insensitive to small degrees of contamination, are variable and range from 0.51276 to
0.51237. Hence, we favor fractional crystallization from different basic parents, combined with
assimilation of crustal materials, over a combined partial melt and fractionation origin for the
felsic volcanic rocks.
5.3. Assimilation and Fractional Crystallization (AFC)
A conventional way of identifying crustal contamination in a suite is to show that 87Sr/86Sr and
143
Nd/144Nd initial ratios vary systematically (increase and/or decrease) with increasing degree of
differentiation. Using SiO2 as a fractionation index, most of the felsic rocks here show an overall
increase in Sr-isotopic ratio (Fig. 7b). Most intermediate samples from the Wechecha and Boseti
units exhibit lower Nd- isotopic ratios than the highly felsic samples, suggesting higher rates of
contamination occurred in the intermediate
lithotypes (Kabeto et al., 2001). The felsic
volcanic samples from the study area
apparently plot along AFC trend (Fig. 7a),
with some scatter.
Figure 7. Plots of (a) initial Sr- and (b) Ndisotopic ratios against SiO2 for the northern
MER volcanic rocks. The felsic rocks plot
on two apparent AFC trends. Apparent AFC
and differentiation trends (FC) are also
shown. The Nd- isotopic ratios also show
variations with SiO2 (see text for
discussion). Symbols as in Fig. 2.
Few mineral aggregates and resorbed
feldspar with melt inclusions were observed
during
our
petrographic
investigation.
Although this could indicate magma mixing
or assimilation, complete mixing can be
excluded,
19
because
no
straight
line
relationships exist on simple binary plots such as SiO2-MgO (Fig. 3). Therefore, we consider that
magma mixing is a minor process in the genesis of the felsic volcanic rocks.
Figure 8. Graphical presentation
of the assimilation-fractional
crystallization (AFC) model
(DePaolo, 1981). (a) Plot of
initial 87Sr/86Sr vs Sr (ppm),
model AFC curves calculated
using an average Sudanese upper
crust as an assimilant (Davidson
and Wilson, 1989; 87Sr/86Sr =
143
0.727967,
Nd/144Nd
=
0.511367, Sr = 426, Nd = 9.6,
and Nb = 10 ppm) and starting
basic parent ET1602 (Table 1).
Symbols as in Fig. 2.
The elemental and Sr-Nd isotopic
compositions (e.g., Table 1; Figs.
4 , 5, 6b, 6c, and 7) of the felsic
rocks
provide
operation
of
evidence
both
of
fractional
crystallization and some crustal
assimilation
(AFC)
(DePaolo,
1981). AFC calculations were
carried out using sample ET1602
(143Nd/144Nd = 0.512894, and
87
Sr/86Sr = 0.70350; Table 1) as
starting
and
basaltic
using
compositions
DePaolo’s
(1981)
equations (Fig. 8). It is assumed
that the rock examined was little
modified by crustal contamination, and that crustal influence is of minor importance
(isotopically) in the basaltic sample. For example, sample ET1602 has higher
143
Nd/144Nd and
lower 87Sr/86Sr ratios than rhyolite ET1302A, which is most radiogenic (87Sr/86Sr = 0.70727 and
20
143
Nd/144Nd = 0.51273). In addition, the isotopic variability among the basalts is very small (Fig.
5; Table 1) compared to the diversity shown by Arabo-African basement rocks (Cohen et al.,
1984; Hegner and Pallister, 1987; Davidson and Wilson, 1989; Halliday et al., 1991; Möller et
al., 1998). Hence, the choice of initial compositions is therefore not critical for the AFC
calculations (Fig. 8).
Several calculations were made taking R (assimilation rate/crystallization rate ratio) to be 0.001,
0.01, 0.1, 0.3, 0.5, and 0.6. For each model, we assumed DSr = 3.2, DNb = 0.3, and DNd = 0.45.
The D values used are similar to those in the compilation of Rollinson (1995). An example of
AFC calculation performed from the average of the Sudanese upper crust (Davidson and Wilson,
1989, 87Sr/86Sr = 0.72797, Sr = 426 ppm, and 143Nd/144Nd = 0.51137, Nb = 9.6, Nd = 24 ppm) is
shown in Fig. 8a and b, with basalt sample ET1602 taken as the basic parent. Assimilation of the
assumed crustal rock by strongly differentiated trachytes and rhyolites at low mass assimilation
to mass crystallization rates (R = 0.001-0.6), and moderate to high F values (> 0.1 on AFC
curves) can produce the samples which exhibit high
143
87
87
Sr/86Sr, low Sr (< 185 ppm), low
Nd/144Nd and high Nb (> 200 ppm).
Sr/87Sr ratios plotted against Sr concentration (Fig. 8a) clearly show the effects of AFC. The
data describes a curved differentiation trend with a sharp inflection around the highly felsic
compositions, reflecting the strong influence of plagioclase fractionation with concomitant
decrease in Sr (Figs. 4 and 8). The hypothetical trend describes the general differentiation trend
among the felsic rocks, along which Ni, Cr, CaO, and MgO contents broadly decrease. Such
hypothetical curves were also tested in the Jebel Marra area of the Sudan (Davidson and Wilson,
1989) and were successfully applied to northern Kenyan felsic lavas (Kabeto et al., 2001), to
examine the evolution of basaltic to trachytic and phonolitic magmas.
Relatively higher rates of assimilation are calculated for intermediate rocks (Fig. 8a and b). This
may be explained by differentiation of intermediate magmas at deeper levels in the crust, where
higher ambient wallrock temperatures and presence of hot basic magmas would facilitate higher
rates of assimilation (Davidson and Wilson, 1989; Macdonald et al. et al., 1995; Panter et al.,
1997). Furthermore, high rates of assimilation (0.3-0.6) are evident for most felsic volcanic rocks
from the rift margin (Fig. 8b), indicating a greater degree of contamination than at the rift center.
This is also in agreement with the thicker crust at the rift margin than at the rift center (Makris
and Ginzburg, 1987).
21
6. CONCLUSIONS
Studies of the northern Main Ethiopian Rift (MER) volcanic rocks offer insight into the genetic
relations of basic and felsic volcanic rocks, and establish that compositional contrasts occur in
equivalent volcanic rocks at rift margin and center magma series within a single intraplate
continental setting. Higher Al2O3, K2O, Zr, Nb, Sr, and Rb and lower CaO, CaO/Al2O3, and
Zr/Nb concentrations in the rift margin felsic volcanic rocks erupted mostly in the Pliocene
compared to the Quaternary equivalents in the rift center reflect inheritance from their basic
parents. The spatial and temporal distinctions between the volcanic suites in the study area and
their markedly different geochemistry are explained by evolution along separate magma trends.
Hence, alkaline basaltic melts produced by lower degrees of partial melting are a possible source
for most of the felsic volcanic rocks at the rift margin. In contrast, transitional basaltic melts
produced at high degrees of partial melting are thought to be the parent for the felsic volcanic
rocks in the rift center.
Modeling of the geochemical variations suggests that crystal-liquid fractionation processes
within the shallow reservoirs were dominant during most trachyte-rhyolitic production in the rift
center, along with less well-developed AFC processes. AFC appears to play a greater role in the
genesis of intermediate rift center rocks and felsic rocks at the rift margin. We consider this to
partly be a function of depth of fractionation of the magmas, implying that intermediate and rift
margin magmas differentiated at deeper levels, whereas the more felsic trachytes and rhyolites of
the rift center originated at shallow crustal levels. This is also in agreement with known variation
in crustal thickness, variation as thinner crust is present in the rift center than at the rift margin.
7. ACKNOWLEDGMENTS
We thank Prof. S. Iizumi for his help with isotope analyses and Dr. T. Bary for AFC model
calculation and discussion. K. Kabeto acknowledges financial support from Japan Society for
Promotion of Science (JSPS) during post doctoral research at Shimane University.
8. REFERENCES
Abebe, T., Mazzarini, F., Innocenti, F.
& Manetti, P. 1998. The Yerer-Tullu Wellel
volcanotectonic lineament: a transtensional structure in central Ethiopia and the
associated magmatic activity. J. Afr. Earth Sci., 26:135-150.
22
Abebe, T. 1995. “L’allineamento vulcano-tettonica Yerer-Tullu Wellel (Etiopia centrale):
evoluzione petrologica e volcanologica di un sistema transversale al rift etiopico.”
University of Florence, Dipartimento di Scienze della Terra, PhD thesis.
Ayalew D., Barrey, P., Marty, B., Reisberg, L., Yirgu, G. & Pik. R. 2002. Source, genesis, and
timing of giant ignimbrite deposits associated with Ethiopian continental flood basalts
Geochemica et Cosmochimica Acta, 66: 1429-1448.
Boccaletti, M., Mazzuoli, R., Bonini, M., Tura, T., & Abebe, B. 1999. Plio-Quaternary
volcanotectonic activity in the northern sector of the Main Ethiopian Rift: relationships
with oblique rifting, J. Afr. Earth Sci., 29: 679-698.
Bigazzi, B., Bonadonna, F.P., DiPaola, G.M. & Giuliani, A. 1993. K-Ar and fission track ages of
the last volcano tectonic phase in the Ethiopian Rift Valley (Tullu Moye area). In:
Geology and mineral resources of Somalia and surrounding regions. Ist. Argon.
Oltremare, Firenze, Relaz. E Monogr., 113: 311-322.
Chernet, T., Hart, W.K., Arson, J.L. & Walter R.C. 1998. New age constraints on the timing of
volcanism and tectonism in the northern Main Ethiopian Rift-southern Afar Transition
zone (Ethiopia) Journal of Volcanological and Geothermal Research, 80: 267-280.
Chernet, T. 1995. Petrological, Geochemical and Geochemical and Geological
Investigation
of volcanism in the northern Main Ethiopian Rift-Southern Afar Transition Region.
University of Miami, Oxford, Ohio, PhD thesis.
Cohen, R.W. S., O’Nions, R. K. & Dawson, J. B. 1984. Isotope geochemistry of xenoliths from
East Africa: Implications for the development of mantle reservoirs and their interactions.
Earth Planet. Sci. Lett., 68: 209-220.
Davidson, J.P. & Wilson, I.R. 1989. Evolution of an alkali basalt-trachyte suite from Jebel Marra
volcano, Sudan, through assimilation and fractional crystallization. Earth Planet. Sci.
Lett., 95: 141-160.
Davies, G.R. & Macdonald, R. 1987. Crustal Influences in the petrogenesis of the Naivasha
Complex: Combined Trace Element and Sr-Nd-Pb Isotope constraints. J. Petrol., 28:
1009-1031.
DePaolo, D. J. 1981. Trace element and isotopic effects of combined wallrock assimilation and
fractional crystallization. Earth Planet. Sci. Lett., 53: 189-202.
23
Gasparon, M., Innocenti, F., Manetti, M., Peccerillo, P. & Abebe, T. 1993. Genesis of the liocene
to Recent bimodal mafic-felsic volcanism in Debre Zeit area, central Ethiopia:
volcanological and geochemical constraints. J. Afr. Earth Sci., 17: 145-165.
Halliday, A.N., Davidson, J.P., Hildreth, W. & Holden, P. 1991. Modeling the petrogenesis of
high Rb/Sr silicic magmas. Chem. Geol., 92: 107-114.
Hart, W.K., WoldeGabriel.G., Walter, R.C. & Mertzman, S.A. 1989. Basaltic volcanism in
Ethiopia. Constraints on continental rifting and mantle interactions. J. Geophys. Res., 94:
7731-7748.
Hegner, E. & Pallister, J.S. 1989. Lead, strontium and neodymium isotopic characteristics of
Tertiary Red Sea rift volcanics from the central Saudi Arabian coastal plain. J. Geophys.
Res., 94: 7749-7755.
Iizumi, S., Maehara, K., Morris, P. & Sawada, Y. 1994. Sr isotope data of GSJ reference
samples. Mem. Fac. Sci., Shimane Univ., 28 : 83-86.
Iizumi, S., Morris, P. & Sawada, Y. 1995. Nd isotope data of GSJ reference samples. Mem. Fac.
Sci., Shimane Univ. 29: 73-76.
Irvine, T.N. & Baragar, W. R. A. 1971. A guide to the chemical classification of the common
volcanic rocks. Canadian Journal Earth Sciences, 8: 523-548.
Kabeto, K., Sawada, Y. & Wakatsuki, T. 2001. Different Evolution trends in Alkaline evolved
lavas from the Northern Kenyan Rift. J. Afr. Earth Sci., 3: 419-433.
Kamber, B.S. & Collerson, K.D. 2000. Zr/Nb Systematics of Ocean Island Basalts Reassessed—
the Case for Binary Mixing. J. Petrol., 41: 1007-1021.
Kimura, J. & Yamada, Y. 1996. Evaluation of major and trace element XRF analyses using a
flux to sample ratio of two to one glass beads. J. Min. Petr. Econ. Geol., 91: 62-72.
Kushiro, I. 1968. Compositions of magma formed by partial zone melting of the Earth’s upper
mantle. J. Geophys. Res., 73 : 619-634.
Le Bas, M. J., LeMaitre, R. W., Streckeisen, A. & Zanettin, B. 1986. A chemical classification
volcanic rocks based on the total alkali-SiO2 diagram. J. Petrol., 27: 45-750.
Lippard, S.J. 1973. The petrology of phonolites from the Kenyan rift. Lithos, 6: 217-34.
Macdonald, R. 1974. Nomenclature and petrochemistry of the peralkaline oversaturated
extrusive rocks. Bull. Volcanology, 35: 78-83.
24
Macdonald, R. Davies, G.R., Upton, B.G. J., Dunkley, P.N., Smith, M. & Leat, P.T. 1995.
Petrogenesis of Silali volcano, Gregory Rift, Kenya. J. Geol. Soc. London, 152: 703-720.
Makris, J. & Ginzburg, A. 1987. The Afar depression: transition between continental rifting and
sea-floor spreading. Tectonophysics, 141: 199-214.
Mohr, P. 1992. Nature of the crust beneath magmatically active continental rifts.
Tectonophysics, 213: 269-284.
Möller, A., Mezger, K. & Schenk, V. 1998. Crustal Age Domains and the Evolution of the
Continental Crust in the Mozambique Belt of Tanzania: Combined Sm-Nd, Rb-Sr, and
Pb-Pb Isotopic Evidence. J. Petrol., 39: 749-783.
Morton, W.H., Rex, D.C., Mitchell, J.G. & Mohr, P. 1979. Rift-ward younging of volcanic units
in the Addis Ababa region. Ethiopian rift valley. Nature, 280: 284-288.
Nelson, S. & Davidson, J.P. 1993. Intractions Between Mantle-Derived Magmas and mafic
Crust, Henry Mountains, Utah. J. Geophys. Res., 98: 1837-1852.
Norrish, K. & Hutton, J.T. 1969. An accurate X-ray spectrographic method for the analysis of a
wide range of geochemical samples. Gochem. et Cosmochem. Acta, 33: 431-453.
O’Hara, M.H. 1968. The bearing of phase equilibria studies in synthetic and natural systems on
the origin and evolution of basic and ultrabasic rocks, Ann. Rev. Earth Planet. Sci. lett.,
4: 69-133.
Panter, S. K., Philip, R. K. & John, L. S. 1997. Petrogenesis of a Phonolite-Trachyte Succession
at Mount Sidley, Marie Byrd Land, Antarctica. J. Petrol., 38: 1225-1253.
Peccerillo, A., Yirgu, G. & Ayalew, D. 1995. Genesis of Acid volcanics along The Main
Ethiopian Rift: Case history of the Gedemsa Volcano. SINET, Ethiop. J. Sci., 18: 23-50.
Pik. R., Daneil C., Yirgu, G., Hofmann, C. & Ayalew D. 1998. The northwestern Ethiopian flood
basalts: Classification and spatial distribution of magma types Journal of Volcanological
and Geothermal Research, 81: 91-111.
Rollinson, H. R. 1995. Using Geochemical Data: evaluation, presentation and interpretation.
Longman, London, 352 pp.
Roser, B.P., Sawada, Y. & Kabeto, K. 1998. Crushing performance and contamination trials of a
tungsten carbide ring mill compared to agate grinding. Geosci. Rept. Shimane Univ.,
17:1-9.
25
Sato, H. 1977. Nickel content of basalt magmas; identification of primary magmas and a
measure of the degree of olivine fractionation. Lithos, 10: 113-120.
Shaw, D.M. 1970. Trace element fractionation during anatexis. Gochem. et Cosmochem. Acta,
34: 237-243.
Skjerlie, K.P. & Johnston, D. 1993. Fluid absent melting behavior of F-rich tonolitic gneisses. J.
Petrol., 34: 785-803.
Stewart, K. & Rogers, N.W. 1996. Mantle plume and lithosphere contributions to basalts from
southern Ethiopia. Earth Planet. Sci. Lett., 139: 195-211.
Tatsumi, Y. & Kimura, N. 1991. Secular variation of basalt chemistry in the Kenya Rift:
evidence for the pulsing of asthenospheric upwelling. Earth Planet. Sci. Lett., 104: 99113.
Trua, T., Deniel, C. & Mazzuoli, R. 1999. Crustal control in the genesis of Plio-Quaternary
Bimodal magmatism of the Main Ethiopian Rift (MER): geochemical and isotopic (Sr,
Nd, Pb) evidence. Chem. Geol., 155: 201-231.
Weaver, S. D. 1977. The Quaternary Caldera Volcano Emuruangogolak, Kenya Rift, and the
Petrology
of
a
Bimodal
Ferrobasalt-Pantelleritic-Trachyte
Association.
Bull.
Volcanology, 40: 209-230.
Wilson, M. 1989. Igneous Petrogenesis. Unwin Hyman, London, 466 pp.
Wilson, M., Downes, H. & Cebriá, J. 1995. Contrasting fractionation trends in coexisting
continental alkaline magma series; Cantal, Massif Central, France. J. Petrol., 36: 17291753.
Wolde, B. 1996. Spatial and temporal variations in the compositions of Upper Miocene to
Recent basic lavas in the northern main Ethiopian rift: implications for the causes of
Cenozoic magmatism in Ethiopia. Geol. Rundsch, 85: 380-389.
WoldeGabreil, G., Aronson, J.L. & Walter, R.C. 1990. Geology, geochronology and rift basin
development in the central sector of the Main Ethiopian Rift. Geol. Soc. Bull., 102: 439458.
26
Table 1. Chemical and Sr-Nd isotopic compositions of selected samples from the northern MER. Abbreviations are: nd = not detected; – = not
measured; M = Measured ratio; I = initial ratio, Tra = trachyte, rhy = rhyolite, benm = benmoreite, ha = hawaite, ba = basalt, igni = ignimbrite, and
obs/pst = obsidian/pitchstone. Major oxides in wt.% and trace elements in ppm. Major and trace element oxides are recalculated to 100% on an
hydrous basis. Ages from Morton et al. (1979) and Boccaleti et al. (1999).
UNIT
Nazreth
Boseti
Sample No.
ET14
ET
971602
ET
117
ET1607
ET1608
ET
1702A
ET1504
ET1507
ET1505
ET1501
ET1506
ET1605
ET1604b
ET1603b
Locality
Tafu
Kone
Barko
Harbona
Kimbo
Boku
Boseti
Boseti
Boseti
Wolenchiti
Boseti
Hada
Hada
Hada
Rock type
trachyte
rhyolite
rhyolite
rhyolite
rhyolite
rhyolite
benmoreite
trachyte
trachyte
trachyte
rhyolite
trachyte
rhyolite
rhyolite
Age (Ma)
0.6
0.3
0.5
0.6
0.83
0.5
0.5
0.5
0.5
0.5
0.5
0.5
0.5
0.5
SiO2
67.98
68.84
71.72
69.35
69.67
73.78
56.24
65.44
66.82
66.90
69.66
66.61
67.19
68.00
TiO2
0.59
0.41
0.40
0.38
0.61
0.31
2.13
0.63
0.41
0.47
0.32
0.45
0.41
0.41
Al2O3
14.07
13.02
11.00
14.94
15.35
10.85
15.38
12.07
13.71
14.50
9.84
15.42
15.18
15.04
Fe2O3
0.85
0.93
0.92
0.59
0.42
0.82
1.52
1.48
1.02
0.91
1.22
0.84
0.82
0.80
FeO
4.34
4.73
4.74
2.99
2.14
4.16
8.33
7.54
5.20
4.66
6.24
4.30
4.17
4.06
MnO
0.21
0.19
0.21
0.16
0.15
0.15
0.31
0.30
0.22
0.23
0.26
0.16
0.16
0.16
MgO
0.32
0.06
0.03
0.22
0.44
0.02
2.60
0.48
0.01
0.09
-
0.22
0.14
0.13
CaO
1.68
0.67
0.44
0.58
0.85
0.27
5.14
1.04
0.72
1.10
0.29
1.64
1.48
1.47
Na2O
5.86
6.82
5.77
6.57
6.16
5.02
5.56
7.49
7.89
6.38
8.04
6.38
6.22
5.90
K2O
3.99
4.30
4.77
4.17
4.12
4.62
1.99
3.47
3.98
4.72
4.12
3.92
4.20
3.99
P2O5
0.09
0.03
0.01
0.04
0.09
-
0.80
0.05
0.02
0.03
0.01
0.06
0.04
0.04
LOI
1.47
-
-
1.48
2.10
0.20
0.20
2.45
0.29
1.07
0.73
1.08
1.98
0.45
Alkali total
9.85
11.12
10.53
10.74
10.28
9.64
7.55
10.96
11.87
11.10
12.16
10.29
10.42
9.89
Agpaitic index
0.99
1.22
1.33
1.03
0.95
1.22
0.73
1.33
1.26
1.08
1.80
0.96
0.97
0.93
Mg number
11.75
2.08
1.05
11.70
26.95
1.04
35.76
10.22
0.47
3.23
-
8.47
5.64
5.36
27
(Continued)
UNIT
Plateau
Addis Ababa
Sample No.
ET1201
ET1303
Locality
Rock type
Age (Ma)
Intoto
basalt
22.8
Wechecha
basalt
Wechecha
ET1602
Addis
basalt
ET1403
ET1203B
ET1405
ET1404A
ET1203A
ET1302
ET1301
ET1606
ET1508
ET1507b
Akaki
Menagesha
Yerer
Yerer
Menagesha
Wechecha
Wechecha
Wonji
Sodorei
Boku
benmoreite
benmoreite
benmoreite
trachyte
trachyte
basalt
basalt
4
3.09
3.09
basalt
7.5
7
Rift center mafic
7
3.6
rhyolite
4
4.6
0.5
0.5
hawaiite
0.5
SiO2
47.20
46.32
45.69
46.71
57.02
60.46
62.59
65.47
68.20
68.50
47.81
48.33
52.16
TiO2
1.72
2.24
2.63
2.21
2.08
1.14
0.92
0.27
0.21
0.22
1.92
2.32
2.20
Al2O3
21.17
14.75
15.52
15.21
15.48
16.44
16.53
17.79
15.85
16.00
14.12
15.14
15.56
Fe2O3
1.55
1.76
1.89
1.72
1.39
1.26
1.10
0.63
0.62
0.61
1.77
1.85
1.76
FeO
7.93
8.97
9.65
8.76
7.50
6.41
5.63
3.20
3.17
3.10
9.04
9.42
8.99
MnO
0.14
0.18
0.16
0.17
0.33
0.13
0.13
0.11
0.08
0.07
0.17
0.17
0.21
MgO
4.24
10.81
9.77
9.57
2.19
0.98
0.75
0.06
0.02
0.01
10.42
7.73
4.94
CaO
12.51
11.26
9.66
11.62
4.37
3.30
2.01
0.31
0.36
0.22
11.29
10.55
8.02
Na2O
2.67
2.50
3.41
2.84
5.55
4.62
5.50
7.02
6.51
6.21
2.43
2.93
4.11
K2O
0.63
0.79
1.07
0.80
3.14
4.86
4.56
5.12
4.97
5.03
0.66
0.92
1.31
P2O5
0.25
0.42
0.56
0.40
0.93
0.41
0.28
0.02
0.01
0.03
0.37
0.64
0.73
LOI
1.06
1.61
0.20
1.47
1.48
0.19
0.74
1.47
1.57
1.07
1.60
0.54
0.76
Alkali total
Agpaitic
index
3.29
3.28
4.48
3.64
8.69
9.48
10.06
12.14
11.48
11.23
3.09
3.85
5.42
0.24
0.34
0.44
0.36
0.81
0.78
0.85
0.96
1.02
0.98
0.33
0.38
0.53
Mg number
48.83
68.23
64.34
66.08
34.25
21.39
19.14
3.31
0.95
0.81
67.26
59.39
49.51
(Continued)
28
UNIT
Sample No.
Rift center mafic
ET
ET
941503A
941502
Gedemsa
ET
1706
ET
941503B
Nazreth
ET60
ET54
ET38
ET 9865
ET
9840
Gedemsa
Koka
Gedemsa
Gedemsa
Bulbula
rhyolite
rhyolite
obsidian
ignimbrite
ignimbrite
0.5
0.5
Locality
D.Zeit
D.Zeit
Mojo
D.Zeit
Sololuya
Gedemsa
Alula
Rock type
basalt
basalt
basalt
basalt
rhyolite
rhyolite
obsidian
Age (Ma)
0.6
0.6
1
0.5
SiO2
49.09
46.38
49.45
49.08
TiO2
1.35
1.65
1.95
2.24
Al2O3
16.16
13.00
17.45
15.88
Fe2O3
1.51
1.73
1.76
1.92
FeO
7.70
8.81
8.96
MnO
0.14
0.16
MgO
9.28
CaO
0.5
71.88
0.6
0.6
ET06
ET103
ET94
0.5
0.6
0.5
72.19
72.69
73.59
72.91
72.34
73.47
67.14
0.32
0.31
0.27
0.35
0.31
0.33
0.69
12.85
10.86
9.57
9.89
12.61
9.15
14.26
0.62
0.62
0.89
1.07
0.94
0.61
1.00
0.91
9.80
3.21
3.20
4.59
5.51
4.81
3.13
5.16
4.66
0.14
0.18
0.13
0.13
0.17
0.23
0.21
0.12
0.22
0.21
13.93
6.27
6.36
0.13
0.12
0.06
0.11
0.05
0.06
0.05
0.52
11.00
11.15
9.43
9.42
0.70
0.65
0.41
0.30
0.35
0.55
0.85
1.78
Na2O
2.34
2.26
3.30
3.43
5.13
5.23
5.65
4.83
5.94
5.16
5.29
5.75
K2O
1.17
0.64
0.91
1.06
4.95
4.67
4.30
4.51
4.55
5.09
4.48
3.97
P2O5
0.25
0.30
0.38
0.64
0.01
0.01
0.06
0.01
0.00
0.00
0.13
LOI
0.11
0.17
1.12
0.86
0.22
0.18
1.17
Alkali total
3.51
2.91
4.21
4.49
9.90
9.96
9.34
10.49
10.24
9.77
9.72
Agpaitic index
0.32
0.09
0.37
0.43
1.07
1.06
1.28
1.34
1.49
1.11
1.48
0.96
Mg number
68.23
73.81
55.51
53.65
6.61
6.33
2.30
3.52
1.86
3.55
1.67
16.57
0.33
12.92
10.07
29
-
(Continued)
UNIT
Plateau
Addis
Ababa
Sample No.
ET1201
ET1303
Locality
Rock type
Intoto
basalt
Wechecha
ET1602
Wechecha
Addis
basalt
basalt
Rift center mafic
ET1403
ET1203B
ET1405
ET1404A
ET1203A
ET1302
ET1301
ET1606
ET1508
ET1507b
Akaki
Menagesha
Yerer
Yerer
Menagesha
Wechecha
Wechecha
Wonji
Sodorei
Boku
benmoreite
benmoreite
benmoreite
trachyte
trachyte
basalt
rhyolite
basalt
hawaiite
Cr
32
672
331
491
5
20
17
19
16
Ni
15
231
185
154
8
2
3
7
8
Ba
160
361
301
446
1,001
89
1,267
48
143
Nb
15.7
33.1
47.1
31.4
106.5
108.0
104.6
231.7
192.2
481.2
774.8
745.0
1,104
965.4
818.4
117.7
174.9
157.1
65.0
66.2
82.0
74.9
204.8
44.3
21.2
16.3
28.0
540.0
313.0
276.1
5.3
5.6
6.6
401.6
481.4
466.9
79.0
119.3
120.1
189.5
159.3
151.1
12.0
16.0
16.9
Zr
110.5
125.6
232.6
Y
15.1
23.7
23.8
Sr
792.6
675.8
643.7
122.4
20.7
629.0
19
basalt
2
151
166.0
558
245
36
208
107
52
274
467
1,456
20.4
28.5
26.1
Rb
19.2
16.6
16.6
16.2
La
12.9
24.4
24.9
21.7
93.3
104.2
175.3
694.2
17.8
17.0
24.3
28.9
Ce
28.7
47.8
55.9
45.4
191.5
183.0
152.1
229.0
74.9
37.3
52.9
63.3
Pr
3.7
5.9
6.9
5.5
22.1
22.3
25.0
25.0
4.3
4.8
6.6
8.3
Nd
15.8
24.9
28.3
23.5
84.5
83.3
99.8
100.0
16.0
20.7
26.9
Sm
3.6
5.2
5.8
5.0
15.9
15.5
14.6
30.0
4.0
4.7
5.6
Eu
1.3
1.8
1.9
1.7
3.8
3.7
0.8
5.0
0.8
1.6
2.1
3.5
Gd
3.5
5.2
5.5
4.8
13.7
13.9
10.4
20.0
4.8
4.6
5.7
7.1
Tb
0.5
0.8
0.8
0.7
2.1
2.1
1.8
5.0
1.0
0.7
0.8
1.0
Dy
3.1
4.4
4.5
4.2
12.9
13.3
12.0
20.0
7.9
4.2
4.7
5.7
Ho
0.6
0.8
0.8
0.8
2.5
2.7
2.6
6.7
1.8
0.8
0.9
1.1
30
(Continued)
UNIT
Plateau
Addis
Ababa
Sample No.
ET1201
ET1303
Locality
Rock type
Intoto
basalt
Wechecha
ET1602
Wechecha
Addis
basalt
basalt
Rift center mafic
ET1403
ET1203B
ET1405
ET1404A
ET1203A
ET1302
ET1301
ET1606
ET1508
ET1507b
Akaki
Menagesha
Yerer
Yerer
Menagesha
Wechecha
Wechecha
Wonji
Sodorei
Boku
benmoreite
benmoreite
benmoreite
trachyte
trachyte
basalt
rhyolite
basalt
basalt
hawaiite
Er
1.6
2.3
1.9
2.2
7.2
8.0
8.1
17.7
5.9
2.2
2.2
2.9
Tm
0.2
0.3
0.3
0.3
1.0
1.1
1.3
2.4
0.9
0.3
0.3
0.4
Yb
1.3
1.9
1.8
1.8
6.9
7.4
8.8
15.6
6.6
1.9
2.0
2.5
Lu
0.2
0.3
0.2
0.3
1.0
1.1
1.4
2.2
1.0
0.3
0.3
0.4
Hf
2.6
3.0
5.2
3.0
17.6
17.3
23.2
21.7
18.5
2.9
3.5
3.8
Ta
1.0
1.9
5.7
1.9
6.1
5.8
13.7
10.6
9.1
1.2
2.5
1.5
Th
1.2
2.2
2.4
2.2
14.5
14.2
31.9
24.2
20.9
1.5
2.0
2.3
U
0.3
0.5
0.6
0.5
3.4
3.4
3.7
3.4
2.8
0.4
0.6
0.4
Zr/Nb
7.04
3.79
4.94
3.90
4.52
7.17
7.12
4.76
5.02
4.93
5.77
6.15
6.02
0.70376
0.70383
0.70350
0.70368
0.70387
0.70489
0.70490
0.70504
0.70727
0.70624
0.70427
0.70441
0.70456
0.51222
0.51281
0.51289
0.51288
0.51264
0.51262
0.51264
0.51276
0.51273
0.51237
0.51276
0.51279
0.51270
Isotope
(87Sr/86Sr) I
(143Nd/144Nd)
I
31
(Continued)
UNIT
Rift center mafic
Sample No.
ET
ET
941503A 941502
Locality
D.Zeit
D.Zeit
Rock type
basalt
basalt
Age (Ma)
0.6
0.6
Cr
298
822
Ni
148
380
Ba
331
278
Nb
24.1
26.6
Zr
106.1
109.7
Y
17.6
19.0
Sr
466.6
461.6
Rb
20.3
15.7
La
18.7
18.5
Ce
37.8
38.1
Pr
4.5
4.7
Nd
18.2
19.2
Sm
3.8
4.1
Eu
1.4
1.4
Gd
3.7
4.1
Tb
0.6
0.6
Dy
3.4
3.8
Ho
0.7
0.7
ET
1706
Mojo
basalt
1
Gedemsa
ET
ET60
ET54
ET38
ET 9865
941503B
D.Zeit Sololuya Gedemsa Alula
Gedemsa
basalt
rhyolite
rhyolite
obsidian
rhyolite
0.5
0.5
0.6
0.6
0.5
139
14
9
28
13
Nazreth
ET94
ET 9840
ET06
ET103
Koka
rhyolite
0.5
12
Gedemsa Gedemsa
Bulbula
obsidian
ignimbrite ignimbrite
0.5
0.6
0.5
16
13
15
114
73
28
28
17
28
27
25
28
27
518
491
492
183
128
53
397
66
1,246
32.2
104.1
103.7
100.9
140.0
128.6
109.8
113.9
75.4
165.5
761.2
764.1
791.3
1,077
918
813.8
668.8
503.1
28.1
74.8
74.6
93.5
60.0
94.3
82.2
62.7
48.3
635.9
21.5
20.2
10.9
9.5
3.5
16.0
11.7
179.3
17.3
115.2
113.6
111.2
140.2
120.3
129.9
128.6
82.9
29.6
93.9
87.7
112.8
109.5
107.3
96.0
59.2
62.6
185.6
185.6
226.0
279.8
224.2
196.1
123.3
7.7
22.0
20.6
25.0
25.0
25.0
22.3
14.2
32.1
81.2
76.6
100.0
100.0
95.8
82.6
54.8
6.6
15.8
15.2
20.5
21.2
19.3
16.6
10.9
2.2
2.0
2.0
3.1
3.0
2.2
2.1
3.1
6.3
13.9
13.6
18.2
17.9
17.7
14.8
9.8
0.9
2.3
2.2
3.0
2.9
2.9
2.5
1.6
5.5
14.5
14.1
18.6
17.5
18.1
15.6
9.7
1.1
2.9
2.8
3.7
3.3
3.7
3.1
1.9
46
320
26.4
143.8
24.4
555.0
15.1
25.4
48.8
6.5
27.2
5.6
1.9
5.4
0.8
4.9
0.9
32
(Continued)
UNIT
Rift center mafic
Sample No.
ET
ET
941503A 941502
Locality
D.Zeit
D.Zeit
Rock type
basalt
basalt
Age (Ma)
0.6
0.6
Er
1.9
2.0
Tm
0.3
0.3
Yb
1.7
1.8
Lu
0.3
0.3
Hf
2.5
2.6
Ta
1.4
1.6
Th
2.1
2.1
U
0.6
0.5
Zr/Nb
4.40
4.12
Isotope
(87Sr/86Sr)
I
(143Nd/144
Nd) I
ET
1706
Mojo
basalt
1
Gedemsa
ET
ET60
ET54
ET38
ET 9865
941503B
D.Zeit Sololuya Gedemsa Alula
Gedemsa
basalt
rhyolite
rhyolite
obsidian
rhyolite
0.5
0.5
0.6
0.6
0.5
3.0
8.4
8.4
10.6
9.3
ET 9840
ET06
ET103
Nazreth
ET94
Koka
Gedemsa Gedemsa
Bulbula
rhyolite obsidian
ignimbrite ignimbrite
0.5
0.5
0.6
0.5
10.7
9.2
5.5
2.6
0.4
1.2
1.2
1.5
1.4
1.6
1.3
0.8
2.7
8.2
8.2
10.0
9.6
10.2
8.9
5.4
0.4
1.2
1.2
1.4
1.4
1.5
1.3
0.8
3.9
18.7
18.5
19.2
26.4
22.3
19.9
12.4
1.8
6.1
6.1
5.9
8.1
7.5
6.7
4.3
2.2
16.0
15.7
15.2
20.4
16.9
16.7
9.8
0.5
3.6
3.7
3.6
1.2
4.0
3.9
1.9
5.14
7.31
7.37
7.84
7.69
7.14
7.41
5.87
0.70476
0.70469
0.70783
0.70463
0.70459
0.70589
0.51278
0.51273
0.51278
0.51281
0.4
2.3
0.3
3.5
1.5
2.1
0.5
5.45
0.5127
8
0.51282
33
6.67
(Continued)
UNIT
Nazreth
Boseti
Sample No.
ET14
ET
971602
Locality
Tafu
Kone
Barko
Harbona
Kimbo
Boku
Rock type
trachyte
rhyolite
rhyolite
rhyolite
rhyolite
rhyolite
Cr
10
18
11
17
20
13
-
25
20
16
17
16
20
17
Ni
28
1
28
1
2
26
10
2
1
2
4
2
2
3
Ba
1,290
602
95
974
857
67
657
1,974
821
461
394
968
695
973
126.7
111.9
114.2
Nb
81.7
ET 117
ET1607
ET1608
ET1702A
ET1504
ET1507
ET1505
ET1501
ET1506
ET1605
ET1604b
ET1603b
Boseti
Boseti
Boseti
Wolenchiti
Boseti
Hada
Hada
Hada
benmoreite
trachyte
trachyte
trachyte
rhyolite
trachyte
rhyolite
rhyolite
92.1
129.2
33.8
85.1
109.1
92.4
Zr
539.7
579.0
772.3
756.4
686.9
1,024
161.4
492.9
706.7
528.0
Y
55.5
91.0
56.7
58.2
70.5
80.3
22.6
68.5
80.9
62.8
Sr
156.3
32.2
3.0
21.8
114.9
3.9
578.4
28.3
12.8
6.7
160.8
1,065
129.5
5.4
88.0
92.3
91.3
645.6
680.0
679.4
65.8
69.2
65.8
110.4
83.2
100.8
Rb
85.0
111.8
122.1
95.0
110.2
148.9
21.7
56.9
97.3
78.6
142.9
97.5
103.4
103.8
La
65.7
91.6
93.6
82.5
85.3
106.3
27.4
61.2
82.4
65.4
129.8
73.1
75.9
74.5
Ce
135.0
185.3
168.7
165.8
175.4
215.2
55.7
130.0
169.3
136.5
267.0
148.8
153.2
152.9
Pr
15.9
21.1
23.3
19.8
20.3
24.8
6.6
25.0
19.6
16.0
15.6
17.0
17.7
17.4
Nd
60.5
80.0
87.5
75.0
77.0
91.4
26.5
100.0
75.6
62.0
62.4
64.2
66.3
66.5
Sm
12.2
16.3
17.0
14.5
14.9
17.9
5.4
24.7
15.6
12.7
13.2
12.8
13.3
13.2
Eu
3.4
3.3
2.4
2.9
3.4
2.4
1.8
5.0
3.6
2.2
4.2
3.1
2.9
3.0
Gd
11.0
15.3
14.2
12.3
13.1
15.9
5.1
20.0
14.8
11.7
12.9
11.9
12.3
12.1
Tb
1.8
2.6
2.3
2.0
2.1
2.6
0.8
4.0
2.5
1.9
2.1
2.0
2.0
2.0
Dy
11.0
16.8
13.9
12.5
13.1
15.9
4.5
20.0
15.5
12.0
13.4
12.5
13.0
12.6
Ho
2.2
3.5
2.7
2.5
2.6
3.1
0.9
5.0
3.1
2.4
2.7
2.5
2.6
2.6
34
(Continued)
UNIT
Nazreth
Boseti
Sample No.
ET14
ET
971602
Locality
Tafu
Kone
Barko
Harbona
Kimbo
Boku
Rock type
trachyte
rhyolite
rhyolite
rhyolite
rhyolite
rhyolite
Er
6.3
10.5
7.7
7.1
7.7
8.8
2.4
14.7
9.2
7.2
7.8
7.5
7.8
7.6
Tm
0.9
1.5
1.2
1.1
1.1
1.3
0.3
2.1
1.3
1.1
1.1
1.1
1.1
1.1
Yb
6.1
9.9
8.0
7.7
7.5
8.6
2.1
13.6
8.6
7.2
7.2
7.3
7.6
7.3
Lu
0.9
1.5
1.2
1.2
1.1
1.3
0.3
2.0
1.3
1.1
1.1
1.1
1.1
1.1
Hf
13.1
19.2
18.7
17.8
16.3
22.9
3.7
25.8
16.5
12.9
11.8
15.2
15.9
15.9
Ta
4.7
7.4
6.3
6.5
4.6
7.2
1.9
9.4
6.3
5.0
4.9
5.0
5.3
5.3
Th
10.8
15.0
15.9
12.6
13.4
14.8
2.5
19.2
12.4
9.7
7.6
12.2
12.7
12.5
U
1.9
6.1
2.1
1.8
3.0
3.3
0.6
4.8
3.2
2.3
2.1
1.9
2.8
1.7
Zr/Nb
6.61
4.57
6.90
6.62
7.46
7.92
4.78
5.79
6.48
5.71
6.62
7.33
7.37
7.44
0.70502
0.70577
0.70477
0.70486
0.70628
0.70478
0.70553
0.70450
0.70538
0.70416
0.70447
0.70446
0.70474
0.51277
0.51285
0.51266
0.51281
0.51279
0.51262
0.51285
0.51286
0.51284
0.51277
0.51261
0.51280
0.51274
ET 117
ET1607
ET1608
ET1702A
ET1504
ET1507
ET1505
ET1501
ET1506
ET1605
ET1604b
ET1603b
Boseti
Boseti
Boseti
Wolenchiti
Boseti
Hada
Hada
Hada
benmoreite
trachyte
trachyte
trachyte
rhyolite
trachyte
rhyolite
rhyolite
Isotope
(87Sr/86Sr) I
(143Nd/144Nd)
I
35
Groundwater Suitability for Irrigation: a Case Study from Debre Kidane
Watershed, Eastern Tigray, Ethiopia
*Nata Tadesse1, K. Bheemalingeswara1 and Asmelash Berhane2
1
Department of Earth Science, College of Natural and Computational Sciences, Mekelle
University, P.O. Box 1604, Mekelle, Ethiopia (* [email protected])
2
Department of LaRMEP, College of Dry Land Agriculture and Natural Resource Management,
Mekelle University, Mekelle, Ethiopia
ABSTRACT
The present paper tries to assess groundwater suitability for irrigation purpose in Debre Kidane
Watershed (45.09 km2), northern Ethiopia. The goal was to evaluate the suitability of the
groundwater for irrigation, examine the water types and to investigate possible long and short
term impact on groundwater quality. Thirty six water samples each were collected in rainy and
dry (irrigation) seasons from the active shallow hand dug wells. The well selection for detailed
study and water sampling was done using stratified followed by random sampling technique. The
number of wells selected for study is accounting 10% of the total available hand dug wells in the
area. Parameters such as electrical conductivity (EC), pH, Total Dissolved Solids (TDS),
temperature and other physical properties were recorded in the field. Major anions and cations
(Ca2+, Mg2+, K+, Na+, CO32-, HCO3-, Cl-, SO42- and NO3-) were analyzed in the laboratory.
Chemical data suggests that the water type in the area though varies from Mg-HCO3, Ca-HCO3,
Mg-Cl, Na-HCO3 to K- HCO3, predominantly it is mixed cation-HCO3 type in the rainy season;
and Ca-HCO3 (Piper diagram) and Na-HCO3 (Stiff diagram) in irrigation season. Dominance of
Na in irrigation season and increase in Sodium Absorption Ratio (SAR) are primarily due to
possible increase of Na absorption in the soil replacing Ca. According to the guidelines for
irrigation water quality, the groundwater is suitable for irrigation with some minor exceptions. In
rainy season, 89% of the samples fall under the water class “good” and 11% “permissible”
whereas in irrigation season only 30% are classified as “good” and 70% under “permissible”
class. Toxicity in terms of SAR and salinity though presently under control and there is no
specific toxicity effect on vegetables and field crops, however, there are indications of their
increase in due course of time if proper measures are not taken.
Keywords: Groundwater, Irrigation, Salinity, SAR, Toxicity, Water quality, Hand dug well,
Ethiopia.
1. INTRODUCTION
Water occurring either as surface or subsurface water is the mankind’s most vital and versatile
natural resources. Surface water exists in rivers, lakes, ponds, and oceans. Surface water mainly
rainwater by percolating through soil and weathered rock horizons occupies the subsurface
permeable layers and develops into a groundwater. Due to its interaction for longer period of
time with soil, weathered and fresh rock materials which in turn vary in composition,
groundwater compared to surface water provides a wide range of variation in its composition.
Thus, the utility of groundwater depends not only on its abundance but also on its quality.
Groundwater utilization for irrigation and domestic purposes has been at a maximum compared
others, like livestock and, industrial use etc., particularly due to food security problems and
global climatic changes. Thus groundwater has become an important source for irrigation and it
has become an integral part of the irrigation strategy “to overcome food scarcity” in many
developing countries including in African continent.
Such efforts have helped to expand
irrigation with time particularly from 1950’s the expansion has been rapidly increasing
(Rosegrant et al., 1999). Currently, it accounts for about 72% of global water withdrawals and
about 90% in the case of developing countries.
According to the Ministry of Water Resources (2002), Ethiopia is endowed with huge natural
water resources, which include 122 billion m3 annual surface runoff and 2.9 billion m3 of
groundwater. However, the county’s water resource has contributed little to the country’s socioeconomic development, because major part of the surface runoff is not utilized and at the same
utilization of groundwater for irrigation is still in infantile stage. In the case of the Tigray region
the main economic means is rain fed agriculture. The region is characterized by undulating
topography and experiences arid to semi-arid conditions with highly erratic and unreliable
rainfall. Thus the region is not in a position to cover the annual food requirement of its people.
To alleviate the challenges of food insecurity in the country promotion of irrigated agriculture
was given priority (Mekuria, 2003). Thus, hand dug well construction has become one of the
activities both by the individuals as well as the government on a sustainable basis.
Implementation of this technology does not need high investment and skilled work force instead
a household can easily adopt and practice it. However, the issue of sustainability in terms of
quality and quantity which changes with time and type of practices in the field, demands
attention.
The changing conditions have provided impetus to develop irrigation using groundwater. Debre
Kidane Watershed in Tigray is one such area where about 360 shallow hand dug wells were
constructed primarily to overcome the moisture scarcity. The households benefited from these
dug wells and started cultivating and producing different high value crops even two times a year.
Availability of water by itself is not a guaranty for sustainable development, but, its fitness to
37
specific purpose like irrigation, domestic and industries is important. Since economic advantage
alone cannot sustain the practice, the issues like ecological and social are to be taken in to
account in the development and management of the irrigation schemes. Quality of water should
be made part of such studies. Ultimately the knowledge of irrigation water quality is critical to
understand what management changes are necessary for long-term and short term productivity
particularly for crops that are sensitive to changes in quality (Bohn et al., 1985; FAO, 1985;
Brady, 2002).
The present paper examines the quality of groundwater used for irrigation in Debre Kidane
watershed (Fig.1) and tries to a) determines suitability of groundwater for irrigation purpose; b)
investigate the water types; c) record water salinity, sodicity if any; d) determine the possible
sources for such variations; and, e) examine possible future trends in the toxicity.
1.1 Description of the Study Area
Debre Kidane watershed is located in the eastern Tigray, North Ethiopia. Geographically, it is
found bounded between 545000 – 553000m E and 1535000 – 1544000m N, and has an aerial
coverage of 45.09 km2 (Fig. 1) with a mean altitude of 2200 meters above sea level. The mean
annual rainfall in the area is about 524.08 mm (Nata, 2003). Monthly rainfall is concentrated
mostly in the mid of June to the mid of September. The drainage is well developed and originates
from the elevated areas in the northeast and flow towards south west (Fig. 1). Northeastern
elevated areas form the main recharge source for groundwater in the area. The mean annual
temperature is 18.1ºC and the yearly average maximum and minimum temperatures are 25.1ºC
and 10.8ºC, respectively. The annual range of temperature is 3.7ºC (Nata, 2003). The dominant
crops grown in the area are barley (Hordeum vulgare), wheat (Tritcum asteivum), teff
(Eragrostive teff), maize (Zea mays), finger millet (Pennisetum americanum), beans (Phaseolus
vulgaris) and peas (Pistum sativum).
Soil texture observed in the field are sandy, light clay, silty, loamy sand, and loam. Dominant
soil texture being silt loam (Nata et al., 2008). Major soil of the watershed from along the inlet
and outlet is Vertisols, Cambisol, Leptosol, and Alluvisol.
In the study area the types of aquifers are localized, confined and unconfined types. Confined
aquifers are located in the discharge areas of the basins and the unconfined types are found in the
recharge areas. Out of the observed hand dug wells, 72.5% are located in unconfined aquifers
and the remaining are in the confined aquifers (Nata et al., 2008).
38
15
N
Tigray
14
Afar
Amhara
13
0
18 Miles
12
38
37
39
40
(A)
1544000
N
Streams
0
1 Kilometer
1535000
545000
553000
(B)
Figure 1. Location and drainage map of the watershed (Nata et al., 2007).
(Source for A: Disaster Prevention and Preparedness Agency (DPPA), 2006)
39
2. GEOLOGY AND AQUIFER CONDITION OF THE STUDY AREA
The area consists of three types of rock units, the low-grade metamorphosed Proterozoic
basement rocks and overlying sedimentary rocks comprising sandstone and tillite of Paleozoic
age and recent alluvial deposits (Fig. 2) (Nata et. al., 2007). A brief account of the rock types and
the aquifer characteristics is given below.
2.1 Geology
2.1.1 Basement rocks
These are the dominating rock type in the area and their outcrops are found in northern,
northeastern, eastern, southeastern, central, southwestern, and northwestern parts of the study
area. They occupy both low in the southern parts and high elevated areas in the northern parts
and show flat to steep slopes. They cover 22.8% of the area (Nata et. al., 2007).
The basement rocks are composed of metavolcanics and metasediments. The most conspicuous
foliation trend is north-south with deviations to northeast and northwest. Apart from foliations,
fractures are also present and are irregularly spaced in the rock. In some places they show light to
dark grey, light to dark brown and reddish color. Dominant minerals present in these low grade
greenschist facies rocks include: mafic minerals, chlorite, quartz, feldspars and opaque in
metavolcanic rocks and quartz, chlorite, feldspar, muscovite, pyrite, limonite in metasediments
(Nata et. al., 2007).
2.1.2 Sandstone
This rock unit accounts for about 8.4% of the total study area (Nata et. al., 2007). The outcrops
of these rocks are found in the southern and southeastern parts of the study area. They occupy
higher elevations in the topography, show steep slopes, cliff forming, and overlie the basement
rocks. The rock out crops is very limited as they are covered by alluvium. However, as it was
observed from the geological logs of the hand dug wells, it extends from southeast to the central,
southwestern, western and northwestern parts of the study area. Its maximum thickness which is
observed at the southeastern periphery of the watershed is of about 160 meters (Nata et. al.,
2007).
It is composed of white, coarse grained, cross-bedded, calcareous sandstone containing lenses of
siltstone, grit and polymict conglomerate with sub-rounded to well-rounded pebbles, cobbles and
boulders. Clasts of granite, gneisses are scattered erratically and are seen in many places. The
lower part of the sandstone is white in color with medium grained angular grains and at places
40
strongly cemented by clay minerals. Upper part is the glacial unit, shows features of a glaciofluvial deposit, angular, poorly sorted grains of quartz and lenses of conglomerate. The rock is
porous, highly weathered and show presence of fractures, joints and cross bedding. Among
structures, joints are prominent and show presence of two sets of joints in sandstone which lies
N15ºE and N25ºW (Nata et al., 2007).
The tillite unit is not found as an outcrop in the study area. However, as it was observed from the
lithologs of the hand dug wells, this rock is found underlying the alluvium in the northern,
northeastern, eastern, southeastern, southern southwestern parts of the watershed (Nata et al.,
2007). It is characterized by dark grey color, poorly sorted texture, the presence of big boulder
and its unstratified nature. On the basis of the logs of the hand dug wells, its maximum thickness
is about 4.5m. The rock is dominated by quartz with limited amount of feldspar and varying
amounts of clay, silt and ferruginous and calcareous cementing material (Nata et al., 2007).
2.1.3 Alluvial Deposits
In the watershed alluvial deposits are found covering wide area both in the highlands and the
lowlands. They are also found as thin strips along the margins of the major rivers and their
tributaries. The alluvium is accounting for about 68.8% of the area. The relative abundances and
stratigraphic relations of the sediments, however, are generally not uniform through out the
watershed. Toward the mountains front that is NW and E and topographically high areas in the
central parts of the study area, where river gradients are high due to steep slopes, the alluvial
sediments, in general, are dominated by sub-angular to sub-rounded coarse grained quartz. In the
western parts of the watershed, where the gradient of the rivers decreases down slope, the
alluvial deposit is characterized by the presence of medium to fine grained sand with variable
content of silt and clay. The maximum thickness of the alluvium is about 3.50 m.
2.2 Hydrogeology
2.2.1 Aquifer Types and Characteristics
Different rock types and unconsolidated sediments which host water and act as aquifers have
been classified into two based on the type of porosity and permeability and their extent.
1. Localized aquifers with intergranular porosity and permeability (unconsolidated
sediments: alluvial sediments along the margins of the major river and its tributaries);
and,
41
2. Less extensive aquifers with intergranular and fractured porosity and permeability
(consolidated sediments and basement rocks: sandstone, tillite and basement rocks).
A brief account of the hydrogeological characteristics of the different rocks and unconsolidated
sediments of the watershed is given below with a particular reference to their water storage and
transmission capacities (Fig. 3) (Nata et al., 2008).
1544000
N
Legend
Basement rocks
Sandstone
Alluvium
Hand Dug Well
0
1535000
545000
1 Kilometer
553000
Figure 2. Geological and Hand dug well location map, Debre Kidane, eastern Tigray
(Nata et al., 2007).
2.2.1.1 Basement rocks
These rocks are generally impervious, nevertheless local pockets of groundwater reservoir occurs
in the weathered layer and fracture zones. Geological and hydrogeological logs of nine hand dug
wells in the western and central parts of the watershed indicate that the weathered and fractured
zones are the main sources of groundwater supply in the basement rocks with discharges of hand
dug wells range from 0.4 to 1.5 l/s.
42
2.2.1.2 Sandstone
The rock is characterized by highly weathered and fractured nature. These weathered and
fractured zones as a whole reaches up to a maximum depth of about 4 m. However, in this rock
secondary porosity is significant due to fracturing. Fracturing has increased the void space as
well as its capacity for water transmission and enhanced its usefulness to the water supply.
However, at places it is reduced due to high degree of cementation and in turn permeability and
productivity. This rock is considered to host a highly potential aquifer in the flat areas (northern
and northeastern parts of the watershed) compared to the steeply to gentle dipping rocks which
act as a conduit rather than being an aquifer in the southern and southeastern parts of the area.
In the case of tillite presence of considerable amounts of silt and clay as minerals and as cement
reduces the intergranular permeabilities drastically and makes the rock impervious and hamper
water movement. At places, however, the fractures provide secondary porosity and permeability
and hence make the rock permeable for groundwater flow and storage,
2.2.1.3 Alluvial Deposits
These recent deposits cover major parts of the area overlying both the basement and sedimentary
rocks. They vary in thickness as well as in composition in the plateau areas and along the
margins of the rivers and tributaries. The alluvial deposits mainly comprise of clay, silt, sand and
gravel sized particles in different proportion. These deposits have an average thickness 3.50 m.
In the western part of the study area, these deposits can be considered as potential water bearing
formation due to their primary porosity and location whereas due to their location in the southern
parts the alluvium act as a conduit rather than being an aquifer (Fig. 3).
3. METHODOLOGY
Initially inventory of all hand dug wells that are functional for irrigation purpose was carried out.
During inventory in situ measurement of electrical conductivity, temperature of the groundwater
and air temperature for each well were carried out. Since the electrical conductivity values were
measured in situ at a temperature different from the standard 25 °C, an adjustment of the
electrical conductivity values of the water was made by multiplying the respective measured
electrical conductivity value by the factor corresponding to the temperature at which the
measurement was made.
43
1544000
N
Legend
High Productive Area
Medium Productive Area
Poor Productive Area
0
1 Kilometer
1535000
545000
553000
Figure 3. Groundwater potential in the study area (After Nata et al., 2008).
To determine the number of water samples for chemical analyses stratified and random sampling
techniques were utilized. The in situ measured and corrected electrical conductivity values of the
groundwater were grouped into different water classes based on Quality Classification of Water
for Irrigation (Wilcox, 1955). Then after, 36 water samples were selected randomly from the
different water classes for chemical analyses. Since majority of the wells in the area are present
in the alluvium in the river valley, the wells selected for sampling are also restricted to this part
of the area. Thirty six water samples each were collected from hand dug wells in both rainy
season and irrigation (dry) season. Apart from temperature, other physical parameters like
turbidity, colure, taste and odor were also measured at each site. For the purpose of chemical
analysis the water samples were collected in one liter plastic bottles after thorough wash. All the
water sample locations are shown in the figure 2.
The water samples were analyzed in the Geochemistry Laboratory of Applied Geology
Department, Mekelle University, Mekelle. The samples were analyzed for calcium (Ca2+),
magnesium (Mg2+), potassium (K+), sodium (Na+), bicarbonate (HCO3-), chloride (Cl-), sulfate
44
(SO42-), and nitrate (NO3-). Besides, pH and electrical conductivity in µS/cm at 25 °C were also
measured.
The cations were analyzed using Atomic Absorption Spectrophotometer and the anions using
UV-Spectrophotometer (Varian Spectra, 50B). The data are presented in the form of graphs.
Piper, Stiff and Box and Whisker diagrams were used for representing and comparing water
quality analyses
In this study FAO (1985 and 1989) guidelines are used to evaluate the suitability of the
groundwater of the watershed for irrigation. The guidelines were proposed for evaluating the
potential of an irrigation water to create soil or crop problems.
Various thematic maps such as location, digital terrain model of the watershed, drainage,
geological, and hydrogeological were produced by using ArcView 3.3 and CorelDRAW 12
software’s.
4. RESULTS
4.1 Electrical Conductivity, pH and Total Dissolved Solid (TDS)
EC values were recorded in 36 functional hand dug wells indicate that electrical conductivity
values range from 0.05-2.0 dSm-1 with a mean of 0.945 dSm-1. The values measured for 36
groundwater samples vary from 0.27 dSm-1 to 0.96 dSm-1 with an average of 0.53 dSm-1 at 25oC
for rainy season. The same during irrigation season range from 0.399 to 1.696 dSm-1 at 25oC with
a mean of 0.877 dSm-1 at 25oC. The EC values in groundwater are much higher during irrigation
season up to 65% compared to the same of the rainy season (Fig. 4).
Hydrogen ion concentrations (pH) range from 7.70 to 8.60 during rainy season and 7.22 to 8.39
during irrigation season. The average pH value in rainy and irrigation season is 8.12 and 8.01,
respectively. Total dissolved solid (TDS) values in both the seasons for groundwater range from
274.34 to 1092.2 mg/l. In general, TDS values for groundwater in both rainy and irrigation
seasons are less than 1000 mg/l.
4.2 Cations
Among cations, calcium is the most abundant element and varies from 8 to 60 mg/l during rainy
season with a mean concentration of 35.03 mg/l. During irrigation season, the most abundant
cation is Ca2+ followed by Na+, K+ and Mg2+. The highest measured calcium concentration is 91
mg/l and the lowest is 30 mg/l, and the mean concentration is 64.92 mg/l. In comparing the
45
concentration of calcium in the rainy season to that of irrigation season, it is found that the
concentration of calcium increased by about 50% in the irrigation season. In the case of
magnesium the concentrations during the rainy season range from 6 to 42 mg/l with an average
value of 20.15 mg/l. While in dry season the concentrations of magnesium (Mg2+) range from 8
to 41 mg/l with an average value of 22.25 mg/l. In general, without exception Ca2+ is dominant
over Mg2+ and magnesium do not show any significant change in concentration.
During rainy season Na+ values range from 8 to 140 with a mean concentration of 43.33 mg/l
(Fig.5). In dry season the values range from 18 to 193 with a mean concentration of 54.56 mg/l.
In the irrigation (dry) season, the concentration of sodium is increased by about 25% compared
to the rainy season. K+ values very low and do not show any variation and hence not considered
here.
Figure 4. EC of groundwater samples in rainfall and irrigation season
(RF- rainy season; IRR- Irrigation (dry) season).
250.0
Na concntration (mg/l)
150.0
100.0
50.0
W
16
W
39
W
34
W
18
7
W
27
1
W
13
0
W
49
W
15
6
W
22
7
W
16
7
W
26
0
W
32
9
W
23
6
W
31
4
W
15
2
W
31
8
W
11
0
0.0
W
10
3
Na(mg/l)
200.0
Rain season
Irrigation
Wells code
Figure 5. Na concentration in the rainy and irrigation seasons.
46
4.3 Anions
4.3.1 Bicarbonate (HCO-3) and Carbonate (CO32-)
In the study area, carbonate is found mainly in the form of bicarbonates and is the most abundant
ion among anions. During rainy season, the concentrations of bicarbonate (HCO-3) range from
82.88 to 477.50 mg/l with an average value of 230.28 mg/l. The same during dry season range
from 170.80 to 547.00 mg/l with a mean value of 342.38 mg/l. The mean concentration of
bicarbonate in irrigation season is increased by about 12% compared to the rainy season. In both
seasons the carbonate (CO32-) concentrations are very low and all are below 6.7 mg/l.
4.3.2 Chloride (Cl-)
In the study area chloride is the second abundant anion next to bicarbonate. Chloride values
range from 9.5 to 133.4 mg/l in the rainy season with a mean concentration of 50.71 mg/l. In the
irrigation season, the concentrations of chloride (Cl-) range from 6.5 to 156.1 mg/l with an
average value of 45.41 mg/l. Chloride ion do not show any significant seasonal variation (Fig. 6).
Water (RF)
Chloride concntration
Water(IRR)
W110
W318
W152
W314
W236
W329
Wells Code
W260
W167
W227
W156
W49
W130
W271
W187
W34
W39
W16
W103
Cl- (mg/L)
180
160
140
120
100
80
60
40
20
0
Figure 6. Chloride ion concentration in the rainy and irrigation seasons
(RF- Rainy season; IRR- Irrigation (dry) Season)).
4.3.3 Sulphate (SO4-2)
Sulphate is the third most abundant anion next to bicarbonate and chloride in the study area. The
values range from 7.2 to 51.80 mg/l in the rainy season with a mean concentration of 25.03 mg/l.
In the irrigation season, the concentrations of sulphate (SO42-) range from 15.12 to 86.09 mg/l
with an average value of 36.35 mg/l.
47
4.3.4 Nitrate (NO3-)
The measured nitrate concentrations range from 0.5 to 43.9 mg/l in the rainy season with a mean
concentration of 9.74 mg/l and range from 3.00 to 57.98 mg/l for dry season with an average
value of 16.47 mg/l (Fig.7).
Figure 7. Nitrate values in water from hand dug wells (RF- rainy season; IRR- dry season).
4.4 Water Type
To understand the chemical type of water, different diagrams such as Piper, Stiff, and Box and
Whisker diagrams were used to plot the data. The purpose of plotting data in various diagrams is
to see how much effective they are for irrigation data and what additional information they
provide compared to Piper diagram. Before plotting the proportions of both the anions and
cations they were converted from ppm to epm. Samples collected in both rainy and irrigation
season were plotted in these Piper, Stiff and Box and Whisker diagrams. They are shown in
figure 8 (Piper); figure 9 (Stiff) and figure 10 (Box and Whisker). According to these plots
groundwater type in the study area varies from Mg-HCO3, Ca-HCO3, Mg-Cl, Na-HCO3 to KHCO3 type. During rainy season, the dominant water type is Ca-HCO3 and is indicated by 42%
of the samples, followed by Mg- HCO3 by 30%, Na- HCO3 by 22%, Mg-Cl by 3% and K-HCO3
by 3% samples. In irrigation season the Ca-HCO3 is the dominant type indicated by 69% of the
samples and is followed by Na-HCO3 type by 24% and Mg- HCO3 type by 7% samples.
Interestingly, Piper diagram shows HCO3 and mixed cation type for rainy season and Ca-NaHCO3 type for dry season. Whereas Stiff diagram shows Ca-HCO3 type for rainy season and NaHCO3 for dry season and Box and Whisker plot indicates Ca-HCO3 type for both the seasons.
48
Figure 8. Piper diagram for the water samples A. rainy and B. irrigation season.
Figure 9. Stiff diagrams (A) Ca -HCO3 water type and (B) Na-Ca-Mg-HCO3-Cl type.
Figure 10. Box and Whisker plot for water samples, (A) rainy and (B) irrigation season.
49
4.5 Sodium Absorption Ratio (SAR)
SAR (Todd, 1980) was calculated using Aqua Chem 4.0 software. The values range from 0.35 to
6.51 with an average value of 1.51 for rainy season. The mean, minimum and maximum SAR
during the irrigation season is 1.44, 0.49 and 5.63, respectively.
5. DISCUSSION
5.1 Physical and Chemical Characteristics
Groundwater being subsurface water derives its contents from different sources like soil,
weathered rock, fresh rock and fertilizer with which it will be in contact during infiltration or
percolation. This will be aided by the climate (ex. Temperature); CO2 either free or dissolved
state in soil and time duration. Variations in the content of groundwater often associated with
irrigation practices, temperature, CO2 and time as the main lithological source generally remain
constant. Such changes are prominent during rainy and dry seasons where both quality and
quantity will be affected. During rainy season supply of water being more its ability to carry
suspended material will be more compared to ions due to limited amount time duration. The data
collected on color, taste etc in the field, shows that about 12% of the samples are muddy or
turbid in rainy season with red/brownish color due to the presence excessive amount of
suspended matter. Whereas it was found that the water was transparent in dry season. Regarding
other parameters, there was no problem of odder or taste or aroma in samples in both the
seasons.
Furthermore, a major part of the water in rainy season leaves the catchment as run off and the
remaining part will percolate through irrigated soil, alluvial cover and weathered rock to
recharge groundwater. So the shallow hand dug wells get recharge prominently from two
different routes, 1) drainage channels with sand dominated channel sediments, and 2) irrigated
soil, non-irrigated soil and weathered rock. Thus sheet flow and stream water supply dominate
rainy season while soil infiltration during dry season. This is well reflected in the compositional
variations observed in the water samples collected during rainy and dry seasons. Such as Ca,
HCO3, NO3 and SO4 are dominant in the dry compared to the rainy season. The reason for such
variations is partly natural and partly anthropogenic.
EC, pH and TDS data suggest though shows variation in their values the majority of the values
lie within the acceptable range for irrigation purpose (FAO, 1989). According to FAO (1989)
50
irrigation water quality guidelines, normal pH range is 6.5 to 8.4. Out of the total analyzed
groundwater samples in both sampling seasons, 97% of them are within this normal range.
In the case of carbonate and bicarbonate ions, which are derived from the dissolution of
carbonate minerals and available carbonic acid in the soil horizons, are within the acceptable
range according to FAO (1985) water quality guidelines for irrigation purpose. The acceptable
range for bicarbonate acceptable range is 1-10 meq/l for bicarbonate and 0-0.1 meq/l for
carbonate. Therefore, the groundwater of the watershed has no restriction in use for irrigation
purposes.
Chloride values being <4 meq/l in the samples collected in rainy season and also the same values
in all the samples except one for dry season indicating that if the groundwater of the watershed is
considered for surface irrigation use, no chloride toxicity will rise. However, a groundwater from
one well needs slight to moderate restriction on use.
According to FAO (1985), water with a sulphate concentration of 0-20 meq/l is considered as a
usual range in irrigation water. In the study area, the concentration is not above 1.0 meq/liter.
Therefore, it is recommended for use for irrigation purpose without restriction.
Unlike most other elements in groundwater, nitrate is not derived primarily from the minerals
(rocks) that make up the groundwater reservoir. Instead, nitrate enters groundwater as part of
nitrogen cycle in the earth’s hydrosphere and biosphere. In groundwater natural nitrate
concentrations generally range from 0.1 to 10 mg/l (Davis and Dewiest, 1966). Nitrate - nitrogen
(NO3 – N) content greater than 30 mg/l is considered harmful and is severely restricted to use for
irrigation purposes (FAO, 1989). Out of the total analyzed groundwater samples (Fig. 8), 31% in
rainy season and 3% in irrigation season have a concentration of nitrate - nitrogen (NO3 – N)
below 5 mg/l. Groundwater from these wells is not hazardous and needs no restriction on use
(FAO, 1989). 3% in rainy season and 8% in irrigation season have a concentration of nitrate nitrogen (NO3 – N) greater than 30 mg/l. Groundwater from these wells is hazardous and needs
severe restriction on use. In the remaining 66% in rainy season and 89 % in irrigation season, the
groundwater of the watershed have a concentration of nitrate - nitrogen (NO3 – N) between 5
and 30 mg/l. Groundwater from these wells needs slight to moderate degree of restriction on use.
The source for this ion seems to be anthropogenic.
In the case of Ca, Mg and Na the concentration being in the range of 0-20 meq/l for Ca2+, 0-5
meq/l for Mg2+ and 18-193 ppm for Na. These values though are within the acceptable range for
51
irrigation (FAO, 1985) purpose, some of the samples showing higher values for Na in dry season
which is well reflected in the Stiff diagram (Fig. 9), seems is an indicator towards increasing Na
with time and in turn SAR. Sodium ion is mainly derived from the breakdown of silicate
minerals particularly feldspars (albite) and generally get mobilized in water in dissolved state
and hence will be lower its content in the stream and shallow hand dug well as well in the rainy
season. It is also well reflected in its concentrations during rainy season. But during irrigation
(dry) season the values are higher because the well gets its recharge mainly from the soil and
rock sources in addition from drainage channels. The reason for such increase of Na content may
be because 1) it gets more time to interact with the source material soil and rock; 2) the irrigated
soil is becoming sodium-rich gradually due to its absorption replacing Ca; 3) availability of more
minerals in the weathered front and alluvium for interaction and dissolution. Though the reasons
are common for other cations as well in the case of Na, absorption process seems to be
prominent.
5.2 Salinity Hazard
In the rainy season, out of the analyzed thirty six groundwater samples, twenty eight samples
have an electrical conductivity values below 0.7 dS/m and the remaining eight samples have an
electrical conductivity ranging from 0.72 to 0.96 dS/m (Fig. 4). In irrigation (dry) season, out of
the analyzed thirty six groundwater samples, eleven samples have an electrical conductivity
values below 0.7 dS/m and the remaining twenty five samples have a range from 0.71 to 1.70
dS/m. Therefore, based on electrical conductivity values, two types of groundwater are
recognized in the study area- groundwater which is not hazardous and needs no restriction on use
and groundwater which needs slight to moderate degree of restriction on use. The first type
groundwater can be used for irrigation for almost all crops and for almost all kinds of soils. No
soil or cropping problems will rise. Very little salinity may develop which may require slight
leaching; but it is permissible under normal irrigation practices except in soils of extremely low
permeabilities. To achieve a full yield potential using the second type, gradually increasing care
in selection of crop and management alternatives are required.
5.3 Water Infiltration Rate (Sodicity)
An infiltration problem related to water quality occurs when the normal infiltration rate for the
applied water or rainfall is appreciably reduced and water remains on the soil surface too long or
infiltrates too slowly to supply the crop with sufficient water to maintain acceptable yields.
52
Although the infiltration rate of water into soil varies widely and can be greatly influenced by the
quality of the irrigation water, soil factors such as structure, degree of compaction, organic
matter content and chemical make-up can also greatly influence the intake rate. The two most
common water quality factors which influence the normal infiltration rate are the salinity of the
water and its sodium content relative to the calcium and magnesium content. High salinity water
will increase infiltration. Low salinity water or water with high sodium to calcium and
magnesium ratio will decrease infiltration. Both factors may operate at the same time. The
infiltration rate generally increases with increasing salinity and decreases with either decreasing
salinity or increasing sodium content relative to calcium and magnesium - the sodium adsorption
ratio (SAR). Therefore, the two factors, salinity and SAR provide information on its ultimate
effect on water infiltration rate.
5.3.1 Rainy season
Out of the analyzed thirty six groundwater samples, 1) five samples show SAR values between 3
and 6 and electrical conductivity between 1.2 and 0.3 dS/m at 25 ºC; 2) twenty three samples
have a computed SAR ranging from 0 to 3 and an electrical conductivity value ranging from 0. 7
to 0.2 dS/m at 25 ºC; and 3) seven samples have a computed SAR values ranging from 0 - 3 and
a corresponding electrical conductivity greater than 0.7 dS/m at 25 ºC. In this season in general
two types of water are recognized: groundwater which needs slight to moderate degree of
restriction on use (i.e., groundwater from both the five and twenty three samples) and a
groundwater which is not hazardous and needs no restriction on use, which is a groundwater
from the last seven samples (FAO, 1989).
5.3.2 Irrigation season
Out of the analyzed thirty six groundwater samples, 1) eleven samples have SAR values ranging
from 3 to 6. However, these eleven samples are classified in to two groups based on their
respective electrical conductivity values. Of the eleven samples two of them have an electrical
conductivity greater than 1.2 dSm-1 at 25 ºC, indicating no restriction on use and in the nine
samples their respective electrical conductivity values are in the range of 1.2-0.3 dSm-1 at 25 ºC,
indicating that the groundwater from these hand dug wells needs slight to moderate restriction on
use; 2) the remaining twenty five samples have a computed SAR values ranging from 0 to 3.
However, these twenty five samples are classified in to two groups based on their respective
electrical conductivity values. Of the twenty five samples, eighteen samples have an electrical
53
conductivity value greater than 0.7 dSm-1 at 25 ºC, indicating no restriction in using this water,
and the remaining seven samples have an electrical conductivity value ranging from 0.7 to 0.2
dSm-1 at 25 ºC, indicating that the water needs slight to moderate degree of restriction on use.
In addition, Wilcox diagram (Wilcox, 1955) was also used to evaluate the potential of the
groundwater to create hazard of salinity and sodicity. Accordingly, three classes of water are
recognized in the rainy season: S1-C1 (low Sodium/Alkali and low salinity hazards), S1-C2 (low
Sodium/Alkali and medium salinity hazards) and S1-C3 (low Sodium /Alkali and high salinity
hazards). As shown in figure 11, out of the analyzed samples 14% of them are in the class of S1C1 (low Sodium /Alkali and low salinity hazards), 80% in the class of S1-C2 (low Sodium /Alkali
and medium salinity hazards) and the remaining 6% in S1-C3 (low Sodium /Alkali and high
salinity hazards).
In the irrigation season two classes of water are recognized: S1-C2 and S1-C3. As shown in figure
12, out of the analyzed samples 30 % of them are in the class of S1-C2 and the remaining 70 %
are in the class of S1-C3.
C1
C2
250
750
C3
2250 C4
32
SodiumHazard(SAR)
Legend
K
26
S4
19
4
13
S3
I
6
0
I
I I I II II IIIIIII
III IIIIIIIIII I
100
I
3
I
S2
S1
Default
Sodium (Alkali) hazard:
S1: Low
S2: Medium
S3: High
S4: Very high
Salinity hazard:
C1: Low
C2: Medium
C3: High
C4: Very high
1000
Salinity Hazard (Cond)
Figure 11. Irrigation water classes using Wilcox diagram (Rainy season).
54
C1 250 C2 750 C3 2250
C4
32
Legend
Sodium Hazard (SAR)
K
26
Sample data
19
S4
Sodium (Alkali) hazard:
S1: Low
S2: Medium
S3: High
S4: Very high
S3
Salinity hazard:
C1: Low
C2: Medium
C3: High
C4: Very high
13
6
K
KK
K
K
K
K
K
K
K
K
K
K
KKKK
K KK
K
KK
K
K
K
K
0
100
Default
K
K
S2
S1
1000
Salinity Hazard (Cond)
Figure 12. Irrigation water classes using Wilcox diagram (Irrigation season).
5.4 Toxicity Problems
5.4.1 Rainy season
With the exception of five samples, in all the remaining thirty one samples the SAR values are
well below 3, indicating no sodium toxicity will rise by using the groundwater from these hand
dug wells for surface irrigation (FAO, 1989). In the five samples the SAR values lie within the
range from 3 to 9, suggesting the necessity of slight to moderate degree of restriction on use of
the groundwater from these five hand dug wells for surface irrigation. The likelihood of sodium
toxicity hazards is high if the groundwater from these five hand dug wells is considered for use.
In the case of SAR, the values below 9 indicate low or little danger for sodium and use on
sodium sensitive crops must be cautioned; between 10 and 17 medium hazard; 18 and 25 highly
hazard and above ≥ 26 very high hazard (Grattan, 2002).
5.4.2 Irrigation season
Out of the analyzed thirty six groundwater samples, in the twenty five samples the SAR values
are well below 3, indicating no sodium toxicity will rise by using the groundwater from these
hand dug wells for surface irrigation (FAO, 1989). In the remaining eleven samples the
computed SAR values lay within the range from 3 to 9, suggesting the necessity of slight to
moderate degree of restriction on use of the groundwater from these eleven hand dug wells for
55
surface irrigation. The likelihood of sodium toxicity hazards is high if the groundwater from
these eleven hand dug wells is considered for use.
5.5 Water Classes
The suitability of the groundwater for irrigation is also examined based on Wilcox (1955)
classification scheme. Accordingly, in the rainy season the groundwater is categorized into two
groups: good and permissible. Out of the total analyzed samples, 89% of the groundwater is in
good class and the remaining 11% is in permissible class. In the irrigation season, out of the total
analyzed samples, 30% of them are in good class and 70% is in the permissible class.
Table 1. Quality classification of water for irrigation (Irrigation season) (after Wilcox, 1955).
EC
Mean
STDEV
(dS/m)
No.
Water Class
Degree of restriction
Wells
(Todd, 1980)
(FAO, 1985)
%
0.25-0.75
0.59
0.11
11
Class 2, Good
None
30%
0.75-2.00
0.99
0.22
25
Class 3, Permissible
Slight - moderate
70%
The data generated in this study are limited. The effect of lithologies on the chemistry of water
and also seasonal variation could not be understood because of non-exposure of the rock and the
wells studied are mainly confined to very shallow hand dug wells not deeper wells. So the data is
discussed mainly in terms of its suitability for irrigation purposes and possible toxic effects if
any.
6. CONCLUSIONS
In general, in most part of the study area the groundwater quality is suitable for irrigation. If the
groundwater is considered for irrigation use, no serious soil or cropping problems will arise. In
few areas, however, the groundwater needs slight to moderate degree of restriction on use. In
these areas farmers has to be advised on how they should manage and utilize the groundwater for
their irrigation activities. Gradual increase of Na due to absorption may cause sodicity problem
hence necessary measures are to be taken to control it.
To maximize the opportunity of the utilization hand dug wells for irrigation, the following
recommendations are suggested:
56
● Detail groundwater potential of the watershed should be investigated to overcome the
overexploitation of groundwater and to know the maximum limit of abstraction;
● Soil quality analysis should be conducted to see the effect of groundwater utilization for
irrigation on soil quality;
● Recharging measures should be given due attention by the community to maintain and
maximize the discharge; and,
● An impact assessment study is suggested to see the significance of hand dug wells
utilization on livelihoods of the households.
7. ACKNOWLEDGEMENTS
We thank the Department of Earth Sciences, Mekelle University, for providing geochemistry
laboratory facilities to carry out the water analysis work. Thanks are also due to all those friends
who helped during field and lab work and for going through the manuscript many times and
providing many constructive criticisms.
8. REFERENCES
Bohn, H. L., Brain L. M. & George A. O’Connor. 1985. Soil chemistry, Second Edition. John
Wiley & Sons, New York, pp. 234-248.
Brady, N. C. 2002. The Nature and properties of soil, Upper Addle River, 13th Edition.
Macmillan, New Jersey, pp. 413-436.
Davis, S.N. & DeWiest, R. J. M., 1966. Hydrogeology. John Wiley & Sons, New York, pp. 463.
FAO. 1985. Water quality for agriculture. Food and Agricultural Organization (FAO) of the
United Nations. FAO, Irrigation and Drainage Paper 29, Rome.
FAO. 1989. Water quality for agriculture. FAO, Rome, pp. 163.
Grattan, Stephen R. 2002. Irrigation water salinity and crop production. University of California.
(http://www.avocadosource.com/links/ Salinity/GrattanStephen2002_ABS.htm -7k –)
(Accessed on August 12, 2005).
Mekuria Tafesse. 2003. Small-scale irrigation for food security in sub-Saharan Africa. Report
and recommendations of a CTA study visit Ethiopia, 20–29 January 2003, CTA
Working Document Number 8031, The ACP-EU Technical Centre for Agricultural
and Rural Cooperation (CTA), The Netherlands.
57
Ministry of Water Resources. 2002. Water Sector Development Program 2002-2016 Report. The
Federal Democratic Republic of Ethiopia, Ministry of Water Resources, Addis Abeba,
Ethiopia.
Nata Tadesse. 2003. Hydrogeological investigation and environmentally sound plans for the
development of groundwater in the Weri River Basin, Tigray, Ethiopia. Ph.D. diss.,
Institutes of Applied Geology, University of Natural Resources and Applied Life
Sciences (BOKU) Vienna, Vienna.
Nata Tadesse, Ahmed Mohammed & Essayas Tsegaye. 2007. Comparison on Aquifer
Characterization of Abrha-Weatsbha Area and Debre Kidane Watershed, Tigray.
Unpublished Report, Department of Applied Geology, Mekelle University, Mekelle.
Nata Tadesse, Asmelash Berhane & Bheemalingeswara, K. 2008. Initiatives, Opportunities and
Challenges in Shallow Groundwater Utilization: a Case Study from Debrekidane
Watershed, Hawzien Woreda, Tigray Region, Northern Ethiopia. Agricultural
Engineering International: the CIGR Ejournal. Manuscript LW 08 008, Vol. X, p. 22.
Rosegrant, M.W. & Clauidi Ringler. 1999. Impact food security and rural development of
reallocating water from agriculture, environment and production technology division,
International Food Policy Research Institute (IFPRI); EPTD Discussion Paper No.47,
USA.
Todd, D. K. 1980. Groundwater hydrology. John Wiley & Sons, New York, pp. 525.
Wilcox, L. V. 1955. Classification and use of irrigation waters, U.S. Dept. Agric. Circ. 969,
Washington, D.C., pp. 19.
58
The Positive Effect of Micro-Dams for Groundwater Enhancement: a Case
Study around Tsinkanet and Rubafeleg Area, Tigray, Northern Ethiopia)
*Dessie Nedaw 1 and Kristine Walraevens 2
1
Mekelle University, Ethiopia, College of Natural and Computational Sciences, Department of
Applied Geology, P.O.Box 231, Mekelle (*[email protected])
2
Ghent University, Belgium, Laboratory for Applied Geology, Department of Geology and Soil
Science, Krijgslaan 281 – 58 – B – 9000 Gent.
ABSTRACT
The government of Tigray Regional State, Northern Ethiopia has been conducting a massive
construction of micro-dams (small reservoirs) in order to decrease the rainfall dependency and
alleviate food insecurity in drought prone areas of the region. Tsenkanet and Rubafeleg
reservoirs are examples of this endeavor. The purpose of this investigation is to evaluate the role
of these micro-dams in enhancing the surrounding groundwater by artificial recharge. In this
study the physical characteristics of the reservoir are discussed including the groundwater
surface water relationship with respect to water level elevation and also hydro-chemical
composition. Six water samples have been analyzed. Samples are taken after rinsing the plastic
bottles with distilled water and the sample to be taken. One surface water sample from each
reservoir and one groundwater samples from each well next to each reservoir has been collected.
Moreover, one sample from spring and one from river has been analyzed. The chemical
composition of the reservoir water and the surrounding groundwater seem to suggest a similar
origin and to be resulting from similar hydro-chemical processes. The chemical composition of
all samples is found to be of the CaHCO3 type, which could be explained as resulting from
precipitation water in which the mineral calcite has been dissolved. Both the topographic
evidence and the water level monitoring data has confirmed the feeding of the reservoir to the
near by shallow groundwater system. Quantifying the amount of groundwater accretion by the
use of modeling and water balance method is recommended.
Keywords: Groundwater, Surface water, Topography, Water Level, Type, Hardness, Class.
1. INTRODUCTION
Water scarcity and unwise management is a global concern especially in sub-Saharan countries.
Global fresh water demand is alarmingly increasing with increase in population and civilization.
As industrial, agricultural and domestic pollution threaten existing supplies – water becomes
increasingly precious resource.
Ethiopia surface water and groundwater resources have been regarded as high giving a name to
the country as the water tower of east Africa (Said, 1993). This is factually true when
considering half of the country, particularly the western and south western part of the country.
59
The endowment can be used for productive purpose that can transform the countries socio
economy (Selashi, 2007). Unfortunately its uneven distribution in space and time coupled with
poor management and development of the resource lead the country to a repeated famine
resulting from drought (FAO, 2005).
The Tigray Regional State is in the northern part of the country, where drought and subsequent
famine was common in the past. Geba catchment, the Project area is characterized by
intermittent rivers which are dry 8 to 9 months with arid and semi arid climatic condition. The
main socio-economic activity in the area is rain fed agriculture which directly linked to the
erratic rainfall. To alleviate this rainfall dependency which is full of uncertainty the regional
government has devised a mechanism by which it can supplement the rain fed agriculture with
irrigation by constructing micro-dams. This task was mainly assigned for a governmental
organization named Co-SEART (Commission for sustainable agriculture and environmental
rehabilitation of Tigray). The activity of the organization has been stopped after constructing 60
micro-dams because it has been believed that they are less effective in supporting the rain fed
agriculture through small scale irrigation schemes for institutional, technical and socio economic
reasons. Therefore the purpose of this investigation was to evaluate the role of these seemingly
failed micro-dams in enhancing the surrounding groundwater system by artificial recharge. The
specific objective of this study were to investigate first the interaction of these reservoir with the
surrounding groundwater and if possible to quantify the total amount of recharge from the
reservoir to the surrounding groundwater system.
2. METHODOLOGY
To achieve the above mentioned objective three wells labeled as Tsinkanet Well 1 (TSW1),
Tsinkanet Well 2 (TSW2) and Rubafeleg well (RFW) and two micro dams labeled as Rubafeleg
Dam (RFL) and Tsinkanet Dam (TSL) has been closely examined. The water level fluctuation
(both in the dams and wells) has been monitored for nearly two years with the average interval of
2 hours, in order to see change of the water level in response to rainfall and recharge. As the
general purpose is to monitor the interaction between the reservoirs and the groundwater system
the impact of irrigation was not considered in this investigation. The water level both in the well
and the reservoir is monitored by TD DIVERS, by which a continuous record of temperature and
water depth is recorded from November 2004 to December 2006 with an interval of 2 hours. In
60
order to be accurate on the elevation difference between measuring points of the diver altimetry
leveling has been conducted.
Water samples collected from Tsinkanet and Rubafeleg on Jan 6, 2006 and March 14, 2006 has
been analyzed in Belgium and Ethiopia. Four water samples were analyzed in the Laboratory of
Applied Geology, Department of Geology and Soil Sciences, University of Ghent, Belgium and
two samples were analyzed at department of Earth Science, Mekelle University. For the Ca++,
Mg++, Na+, K+, Fe, Mn, and Si Varian AAS has been used and for HCO3 titration method is
used and NO3, SO4, Cl, and PO4 UV spectrophotometer is used. The accuracy of the analysis
has been tasted using ionic balance (Appelo, 1996). In all but one the balance is found to be less
than 5%. This indicates that the analysis is sufficiently accurate. In order to interpret and classify
the water a new classification scheme developed by Stuyfzand is adopted (Stuyfzand, 1986). The
classification starts with main type based on Chloride concentration, then each main type is
subdivided to type based on total hardness, then each type is further subdivided into sub type
based on the proportional share of main constituents in the sum of cations and anions in meq/l,
finally each sub type is further classified to classes by using the following formula (Na + K +
Mg) corrected = (Na + K + Mg) measured – 1.061 Cl (meq/l)
3. LOCATION AND ACCESSIBILITY
Geba-catchment has an area 5133 km2. The elevation ranges from 955 m a.m.s.l to 3295 m
a.m.s.l. The mean elevation has been found to be 2146 m, a.m.s.l. Entire Geba catchment is
found between 39°30’ to 40°00’ E (longitude.) and 13°45’ to14°15 N (latitude) figure 1.
Tsinkanet and Rubafeleg area are found in the upper catchment of Geba along the west and east
boundaries respectively (figure 1). In both areas there is a reservoir on upstream side and shallow
hand dug wells in the command areas. The UTM location of the reservoirs and the shallow hand
dug wells where monitoring is conducted is shown on table 1.As one can see from the table the
Monitoring well at Tsinkanet is 400 meter west of the reservoir where as the monitoring well at
Rubafeleg is more than 1 km North of the reservoirs. Both reservoirs are not more than 1 km2
area, and there depth is not more than 10 meter. Tsinkanet Area is found between (561500 –
555500 UTME and 1552100 – 1546000 UTMN) and Rubafeleg area is found between (581700 –
575700 UTME and 1543000 – 1537500 UTMN).
61
The study area is accessible by all weather roads running from Hawzen to Senkata and from
Atsbi to Dera.
Table 1. UTM location of water points used for monitoring.
Name
point
of
Water
Tsinkanet
Tsinkanet (center
Rubafeleg
Tsinkanet well well 2
of
(center
1(monitoring
(TSL)
(TSW2)
dam
well
dam Axis)
(RFW)
RFL
1549080
1541634
1540500
558723
578715
Well) (TSW1)
Location UTMN 1549139
UTME 558303
Axis) Rubafeleg
578700
Tsinkanet
Rubafeleg
Figure1. Location map of Geba catchment, Tsinkanet and Rubafeleg Area (USGS, 2004).
62
of
4. GEOLOGY AND HYDROGEOLOGY OF THE STUDY AREA
The study area is mainly characterized by basement complex of Precambrian age, composed of
meta-volcanics and meta-sediments (Kazmin, 1978). These are exposed in most parts of the
mapped area.
Figure 2. Geological Map of Northern Geba Catchment
(Modified from Tesfamicheal Gebreyohannes et al., 2009)
63
At the eastern most part and along the northwestern part Paleozoic sandstone is exposed. In the
southwestern part a younger granite intrusion is mapped cross cutting the basement complex as
shown in figure 3. Since the area is mainly characterized by crystalline metamorphic rocks, the
groundwater availability and movement is mainly controlled by fractures and weathering zones.
The main water supply for drinking and small scale irrigation is the groundwater tapped from
boreholes to a maximum depth of 50 m and hand dug well with a maximum depth of 10 meter.
In Tsinkanet area the reservoir is mainly characterized by Enticho sandstone, with a thickness
not more than a couple of meters. At the command area the sandstone thins out and the
underlying metavolcanic rock is exposed. In Rubafeleg area both the reservoir and the command
area are characterized by metamorphic rocks mainly metavolcanic and Metasediments.
5. RESULTS AND DISCUSSION
5.1 Geomorphologic analysis from Topographic and Shuttle Radar Topographic Mission
(SRTM, 2004) data
Topographic position is one of the governing factors that control the direction of groundwater
flow. In the area the geographic location of the reservoir and the aquifer system is in a suitable
location enabling the reservoir to feed the groundwater. In addition the high degree of fracturing
and weathering also encourage the down gradient flow of the seepage water. Accordingly the
subsurface water downstream to the reservoir is under favorable location for groundwater
recharge figures 3 and 4.
5.2. Water level Analysis
Water level both in the reservoir and the surrounding groundwater was being monitored since
late 2004 in Tsinkanet and Rubafeleg area. Monitoring has been conducted for more than two
years till December 2006. As it can be seen on the following graphs (Figure 6 & 7) the following
observations are seen.
–
The head in the micro-dam is always higher than that of the wells during the
recorded period.
–
Moreover the general trend of water level change both in the reservoir and in the
well with respect to time is more or less similar.
64
Figure 3. Tsinkanet Area topography.
Figure 4. Rubafeleg area Topography
65
–
The water level in the groundwater is generally found to rise, indicating a
continuous accretion of groundwater from the reservoir even though the water in
the wells is being continuously used for irrigation particularly during the dry
season in Tsinkanet area. As the scope of this paper is to indicate the unseen role
of the reservoirs as a groundwater recharging mechanism the focus is in the
general trend not on quantifying the volume of water recharged.
–
Marshy areas are developed downstream of reservoir.
The above observation indicates that the surface water in the reservoir to interact with the
surrounding groundwater. Moreover the surface water is feeding the groundwater down stream
enhancing the groundwater system as indicated by development of wet area down stream figure
5.
Figure 5. Seepage downstream of
Rubafeleg dam (Top, Rubafeleg
dam and Bottom,
south of the Dam, in the command
area)
66
Figure 6. General Water Level rise on Tsinkanet Well 1 (TSW1).
(From Nov. 2004 to Dec. 2006)
Figure 7. Tsinkanet well 1 (TSW1) being recharged by Tsinkanet micro dam (TSL).
(Evidence from water level data from June 2005 to November 2006)
67
Table 2. Chemical Characteristics of the samples from Rubafeleg and Tsinkanet.
Sample Id
Location (UTM)
TSW2
TSL
RFW
RFL
RFR*
RFS*
558649
558723
578715
578700
578615
578660
1548515
1549080
1541634
1540500
1442465
1541918
ph
7.2
7.4
7.4
7.5
EC (µs/cm)
202
250
454
181
Na (ppm)
13.65 9.9
17.38
5.92
20
18
K (ppm)
1.51
0.46
1.66
4
3
Ca (ppm)
23.28 26.8
63.83
26.5
22
13
Mg (ppm)
6.08
9.45
10.45
5.38
14
15
Fe (ppm)
0.29
0.21
0.18
0.11
1
1
Mn (ppm)
nil
0.01
nil
0.01
0
0
NH4 (ppm)
0.36
0.21
nil
0.17
Zn (ppm)
0.02
0.02
nil
0.01
0
0
Cl (ppm)
8.39
13.92
17.89
7.95
39.48
9.69
SO4 (ppm)
14.82 33.11
21.63
9.28
26.03
79.28
NO3 (ppm)
0.94
0.38
12
0.2
NO2 (ppm)
0.02
0.14
0.02
0.02
HCO3 (ppm)
95.16 128
230.58
99.43
170.32
152.37
PO4 (ppm)
0.03
2.89
2.89
2.89
4.545
5.01
Si (ppm)
8.03
<7
11.74
<7
3.75
* RFR is Rubafeleg River, analyzed at Mekelle University, Earth Science Lab.
* RFS is Rubafeleg spring, analyzed at Mekelle University, Earth Science Lab.
5.3. Hydro-chemical analysis
5.3.1. General chemical characteristics of the water samples in the area:
The general chemical composition of the water samples has been described with the help of table
and graph here. Four samples have been analyzed in Applied Geology Laboratory, Department
of Geology and Soil Sciences, university of Ghent, Belgium for major and minor ions.
68
Additional two samples were analyzed in the Department of Earth Science, Mekelle University.
The results of the analysis are shown on table 2 and Figure 8. The samples are believed to be
sufficient considering the size of the target area and the scope of the investigation. Two samples
are from Rubafeleg, from the dam and borehole and two samples are from Tsinkanet, again from
dam and hand-dug well. The other two samples are from river and spring in Rubafeleg area. As it
has been depicted in table 2, the general characteristics of the water sample is more or less
similar indicating similar origin and trend.
5.3.1.2. Stuyfzand Classification of the water samples:
To further verify similarities of the water samples from the dam and the groundwater Stuyfzand
classification method has been used (Stuyfzand, 1986). Accordingly they are divided into the
following.
Figure 8. Chemical composition of the samples collected from water points.
1. Main type: it is a function of Cl- content.
The chloride concentration of the samples ranges from 7.95 mg/l to 17.9 mg/l. This indicates all
the samples to have a Cl concentration of less than 150 mg/l, accordingly they are grouped as
fresh (F) (Table 3).
Table 3. Main type classification of water samples
69
Sample Id
RFL
RFW
TSL
TSW2
Cl- concentration (mg/l)
7.95
17.89
13.92 8.39
RFR
RFS
39.48
9.69
2. Type (Hardness Code): This is derived from the total hardness calculated from the
following formula.
TH = 2.5 (Ca) + 4.1 (Mg)
Where TH is total hardness in mg/l and (Ca) and (Mg) are also in mg/l
Accordingly the hardness of the samples is tabulated table 4 and found that they range from soft
to hard water.
Table 4. Type classification of the water samples.
Sample Id
Ca (ppm)
Mg (ppm)
Hardness in mg/L
Hardness in mmol/l
Hardness name
TSW2
23.28
6.08
83.128
0.83
Soft
TSL
26.8
9.45
105.745
1.06
Hard
Hardness code
0
2
RFW
63.83
10.45
RFL
26.5
5.38
88.308
0.88
Soft
202.42
2.02
Moderately
hard
1
0
RFR*
22
14
RFS*
13
15
112.4
1.12
Moderately
hard
1
94
0.94
soft
0
Table 5. Milliequivalents and the proportions for major cations and anions.
Sample id
RFL
RFW
TSL
TSW2
RFR
RFS
Units
Ions
Ca +Mg
(Na+K)+
NH4
(Al+H)+(F
e+Mn)
HCO3+CO3
SO4+(NO3+
NO2)
Cl
meq
1.76
0.31
0
1.63
0.19
0.22
%
85.02
14.98
0
79.9
9.3
10.8
meq
4.05
0.77
0.01
3.78
0.64
0.51
%
83.9
15.9
0.2
76.7
13
10.3
meq
2.12
0.54
0.01
2.11
0.7
0.39
%
79.4
20.2
0.4
65.9
21.9
12.2
meq
1.66
0.65
0.01
1.56
0.32
0.24
%
71.6
28
0.4
73.6
15.1
11.3
meq
2.25
0.97
0.036
2.8
0.54
1.11
%
69.1
29.8
1.1
62.93
12.13
24.94
meq
1.88
0.856
0.036
2.5
1.65
0.27
%
67.82
30.88
1.3
56.56
37.33
6.11
70
2. Sub Type: This is quite important to recognize the processes that have determined the
water quality.
The milliequivalent concentration has been converted to proportion as Cations group and anions
group. Accordingly all the samples have been found to be CaHCO3 type (Table 5).
3. Class (Cation Exchange Code)
The code indicates whether the cation exchange has taken place or not and also the nature of
the exchange. The sum of Na, K and Mg (meq/l) is corrected for the seawater contribution,
determined from the Cl- content.
(Na + K + Mg) corrected = (Na + K + Mg) measured – 1.061 Cl (meq/l) (see Table 6)
Table 6. Class classification of the water samples.
Sample Measured
1.061*(Cl) Corrected
- 1 (Cl )
2
Id
(meq)
(meq)
(Na+K+Mg)
(Na+K+Mg)
RFL
0.74
0.23
0.51
-0.33
RFW
1.63
0.54
1.09
-0.5
TSL
1.31
0.41
0.9
-0.44
TSW2
1.13
0.25
0.86
-0.34
RFR
2.12
1.18
0.94
-0.74
RFS
2.086
0.29
1.76
-0.38
Name
(Na +
surplus
(Na +
surplus
(Na +
surplus
(Na +
surplus
(Na +
surplus
(Na +
surplus
Code
K+ Mg) - +
K+ Mg) - +
K+ Mg) - +
K+ Mg) - +
K+ Mg) - +
K+ Mg) - +
The Stuyfzand classification has been used to investigate see water intrusion but as the Author
suggested it can also be used to determine natural waters which has undergone different
processes (Stuyfzand, 1986). In general all the water samples fall in more or less similar group in
which among cations Calcium is dominant and among anions HCO3 is dominant. Rubafeleg
Well has relatively higher Electrical conductivity indicating that it is truly groundwater as
compared to the others. Higher NO3 is observed in this well that could be explained with its
relative position with respect to the toilet near by. From the above classification one can see that
in both Rubafeleg and Tsinkanet area the water from different sources to fall in the same group
71
with more or less similar chemical composition indicating a relationship between the dam water
and the groundwater.
The chemical composition of the water could be explained by dissociation of silicate minerals
mainly plagioclase feldspar and/or K feldspars as the groundwater flows through the surrounding
area or as base flow water into the micro-dam. But this should be further verified by making
detailed investigation on the geochemistry of the surrounding rocks.
6. CONCLUSION
The study tries to justify the interaction between the surface water in Tsinaknet and Rubafeleg
reservoir, from physiographic, water level and hydro-chemical respect it has been found that the
surface water to interact with the surrounding groundwater. In addition this reservoirs
constructed for the purpose of surface water harvesting are found to enhance the groundwater
system downstream encouraging the use of small hand-dug wells managed by a single family for
their subsistence farming.
The study further recommend to make an investigation to quantify the quantitative recharge
calculation to be made to further understand the role of this reservoir as groundwater enhancing
structure using modeling approach. More over a detailed Geochemical investigation will further
consolidate the findings of the research from hydrochemical point of view.
7. REFERENCES
Appelo, C.A.J. & Postma, D. 1996. Geochemistry, Groundwater and Pollution. A.A Balkema,
Rotterdam,
FAO. 2005. FAO’s information system for Water and Agriculture (AQUASTAT).
http://www.fao.org/nr/water/aquastat/countries/ethiopia/index.stm
Kazmin, V. 1978. Geology of Ethiopian basement and possible relation between the
Mozambique and the Red Sea Belts. Egyptian Journal Geology, 22: 73-86.
Said, R. 1993. The River Nile, Geology, Hydrology and Utilisation. Pergamon Press, England,
pp. 282.
Seleshi, B. 2007. Water challenges, innovations and interventions for Ethiopia, Think Tank
paper on water resource, pp. 10
72
Stuyfzand, P.J. 1986. A new hydrochemical classification of water types; principles and
application to the coastal dunes aquifer system of the Netherlands, 9th salt water intrusion
meeting, Delft, proceedings, 641 – 655.
Tesfamichael, G, F., De Smidt., Hagos,M., Amare, K., Kabeto, K., Hussein, A., Jan Nyssen,
Bauer, H., Moeyersons, J., Deckers, J. & Taha. N. 2009. Tigray Livelihood, Large scale
geological mapping of the Geba basin, Northern Ethiopia.
USGS, 2004. Shuttle Radar Topographic Mission, srtm.usgs.gov.
73
Silicon Status and its Relationship with Major Physico-Chemical Properties of
Vertisols of Northern Highlands of Ethiopia
Fassil Kebede
Department of LaRMEP, College of Dry Land Agriculture and Natural Resource Management,
Mekelle University, P.O.Box-231, Mekelle, Ethiopia ([email protected])
ABSTRACT
Silicon is an essential mineral element for higher silicophile plant species and animals. Literature
states that one of the most important functions of Si is the stimulation of plant’s defense
mechanism against abiotic and biotic stresses. In addition, Si fertilization has a more positive
effect than liming on the chemical and physical properties of the soil. There are only a few
studies in Ethiopia that present silicon status of some Ethiopian soils. Hence this study was
initiated with the objective of understanding Si distribution in the soils of the northern highlands
of Ethiopia. To this effect, 32 soil samples were collected from five agro ecological zones of
Tigray. Si in all the samples was determined in the Laboratory of Ethiopian Geological Survey.
This study revealed that Si contents ranged from 79.8 to 87.5 g Si kg-1 in the cultivated Vertisols
of Adigudom, from 97.7 to 115.2 g Si kg-1in Axum, from 113.7 to 117.2 g Si kg-1in Maychew,
from 130.0 to 133.9 g Si kg-1 in Shire and from 137.3 to 166.3 g Si kg-1in Wukro. The highest
concentration was hence found in Wukro where the sand content amounted to 50% whereas the
lowest level was obtained from soils of Adigudom where the clay content exceeded 60%. The Si
contents in all the studied soils are lower than the documented ranges of 200 and 300 g Si kg-1.
Significant correlation was found between silicon status and organic carbon 0.84*(p<0.05), silt 0.84*(p<0.05) and clay 0.84*(p<0.05). Finally, based on the enormous role of Si in plant
physiology and agriculture, this study suggests commencing Si fertilization in these soils for an
expected increase in crop productivity.
Keywords: Essential nutrient, Silica, Silicophile plant species and animals, Si Fertilization
1. INTRODUCTION
Since, crop production, is the major source of food for the population and hence it is the prime
contributing sector to food security in Ethiopia. In addition, it plays a vital role in generating surplus
capital to speed up the overall socio-economic conditions of the farmers. Ethiopia is believed to have a
considerable land resource for agriculture. About 73.6 million hectares (66%) of the country’s land area is
estimated to be potentially suitable for agricultural production (Fassil, 2002). It is generally accepted that
this land resource can support a large population by providing enough food and other agricultural
products required for the development of other sectors. However, the country has remained unable to feed
its people for many years due to archaic agricultural practices and climate variability.
74
The farming systems in Ethiopia can be classified into five major categories namely the highland mixed
farming system, the lowland mixed agriculture, the pastoral system, shifting cultivation and commercial
agriculture. The highland areas are inhabited by four-fifths of the population and also support about 70%
of the livestock population. According to Fassil (2002), crop production is estimated to contribute on
average about 60%, livestock accounts around 27% and forestry and other sub-sectors around 13% of the
total agricultural value. Small-scale farmers who have been adopting low input and low output rainfed
mixed farming with traditional technologies dominate the agriculture sector.
As the highland’s population continues to grow rapidly, its agricultural land is becoming
increasingly degraded. Farmers in the highlands are intensifying land use to meet food needs
without proper management practices and cost effective inputs. The resulting depletion of
nutrients from soils has caused crop production to stagnate or decline in the highlands. Unless a
concerted effort is placed in confronting the problems of nutrient depletion, deteriorating
agricultural productivity will seriously undermine the foundation of sustainable economic growth
of the nation at large.
Silicon is the second most abundant element in the earth’s crust constituting approximately 20
atomic % of the lithosphere (Iler, 1979). In soil solutions, the prevailing form is monosilicic acid,
Si(OH)4, with a solubility in water (at 25 ˚C) of about 2 mM (equivalent to 120 mg SiO2 per
liter). On average, the concentrations in soil solutions are 30-40 mg SiO2 per liter range between
about 7 and 80 mg) with a tendency to lower concentrations at high pH (>7) and when large
amounts of sesquioxides are present in soils and anion adsorption is dominant (Jones and
Handreck, 1967).
Beginning in 1840, numerous laboratory, greenhouse and field experiments showed sustainable
benefits of Si fertilization for rice (Oryza sativa L.), barley (Hordeum vulgare L.), wheat
(Triticum vulgare Vil), maize (Zea mays L.), sugarcane (Saccharum officinarum), cucumber
(Cucumus sativa L), tomato (Lycopersicon esculentum Mill), citrus (Citrus taitentis Risso) and
other crops (Matichenkov and Calvert, 2002). Unfortunately, the present opinion about Si being
an inert element is prevalent in plant physiology and agriculture despite the fact that Si is a
biogeochemically active element and that Si fertilization has significant effects on crop
production, soil fertility and environmental quality (Epstein, 1999; Matichenkov and
Bocharnikova, 2000).
Higher plants differ characteristically in their capacity to take up silicon (Marschner, 1996).
Depending on their SiO2 content (expressed as a percentage of shoot dry weight), they can be
75
divided into three major groups: wetland Gramineae, such as wetland rice or horsetails
(Equisetum), 10-15%; dryland Gramineae, such as sugarcane and most of the cereal species, and
a few dicotyledons, 1-3%; and most dicotyledons, especially legumes, <0.5% (Takahashi and
Miyake, 1977).
Generally, silicon has more benefits such as the mobilization of soil phosphorus (Marschner,
1996), reduced water loss by cuticular transpiration, and increased resistance against lodging and
pests, deserve more attention in the future in crops other than rice and sugarcane (Marschner,
1996). Ethiopia is a center of biodiversity where several landraces are under productivity.
Reasons for poor crop production in Ethiopia are manifold. One of these may be nutrient
depletion, elements like Si. Hence, the present study was carried out with the objective of
auditing the current Si status in agriculturally important soils of the northern highlands of
Ethiopia and suggest way forward with Si fertilization in crop production.
2. MATERIALS AND METHODS
2.1. Soil sampling
Soil samples were taken from five agroecological zones of Tigray, northern Ethiopia. Five
profile pits were opened in representative sites of Adigudom, Axum, Maychew, Shire and
Wukuro. Soil samples were taken from 0-15, 15-30, 30-45, 45-60, 60-90, 90-140 and 140-160
cm depths. A total of 32 disturbed soil samples were collected. The disturbed soil samples were
air dried and sieved to pass through 2 mm mesh prior to analysis.
2.2. Soil analysis
Organic carbon was analyzed according to Walkley and Black (1934) method, while total N was
analyzed by Kjeldahl method (Bremner and Mulvaney, 1982) in the National Soil Research
Laboratory in Addis Abeba. The silicon composition of the soil samples was determined
following digestion with aqua regia and perchloric acid and finally was determined using Atomic
Absorption Spectrophotometer. The silicon distribution in the soil samples was determined in
Geochemistry Laboratory of the National Institute of Geological Survey in Addis Abeba. In
addition, the silicon distribution was determined by multiplying a factor 0.47 with the SiO2
content that was determined in the Laboratory. Finally, the data were analyzed using Statistical
Packages for Social Sciences (SPSS) 13 for windows.
76
3. RESULTS
As presented in tables 1-5 the Si contents range in the cultivated Vertisols of Adigudom from
79.8 to 87.5 g Si kg-1, in Axum from 97.7 to 115.2 g Si kg-1, in Maychew from 113.7 to 117.2 g
Si kg-1, in Shire from 130.0 to 133.9 g Si kg-1 and finally in Wukro from 137.3 to 166.3 g Si kg-1.
The highest concentration (i.e., 137.3 to 166.3 g Si kg-1) was found in Wukro where the sand
content amounted to 50% whereas the lowest level (81.4 to 95 g Si kg-1) was obtained from soils
of Adigudom where the clay content exceeds 60%.
The Pearson’s correlation was used to find a correlation between silicon status and selected soil
physical and chemical properties as given in table 6. Accordingly, significant correlation was
found between silicon status and silt -0.85* (p<0.05) and clay 0.85* (p<0.05) for Vertisols in
Adigudom; between silicon and total nitrogen 0.77*(p<0.05) for Vertisols in Axum; between
silicon and sand -0.88* (p<0.05) (Isn’t sand positively correlated with Si under normal
circumstances-check your results!) and clay -0.88* (p<0.05) for the Vertisols in Shire and
between silicon and clay -0.85 (p<0.05). The test showed, in most cases, that there is nonsignificant correlation between Si content and pH and soil organic carbon (Table 6). These
properties in turn determine silicon’s effect on soil fertility and plants.
Table 1. Physical and chemical properties of Vertisols of Adigudom.
Depth
pH
cm
Sand
Silt
Clay
TN
OC
Si
-1
%
%
%
%
%
g kg soils
0-15
8.4
16
22
62
0.116
1.5
87.1
15-30
8.4
18
22
60
0.106
1.4
87.4
30-45
8.3
14
23
63
0.084
1.0
87.5
45-60
8.4
16
20
64
0.082
1.4
84.6
60-90
8.6
12
26
62
0.081
1.4
82.3
90-140
8.0
18
64
22
0.095
1.8
74.8
140-160
8.4
14
28
54
0.070
1.0
79.1
77
Table 2. Physical and chemical properties of Vertisols of Axum.
Depth
pH
Sand
cm
Silt
Clay
TN
OC
Si
-1
%
%
%
%
%
g kg soils
0-15
8.4
14
34
52
0.148
1.2
111.3
15-30
8.5
12
30
58
0.127
1.1
114.1
30-45
9.0
12
24
64
0.120
1.0
114.8
45-60
9.0
12
26
62
0.111
0.9
112.7
60-90
9.1
16
24
60
0.116
0.8
115.2
90-140
8.7
16
28
56
0.062
0.8
109.1
140-160
8.3
18
26
56
0.051
0.6
97.7
Table 3. Physical and chemical properties of Vertisols of Maychew.
Depth
pH
cm
Sand
Silt
Clay
TN
OC
Si
%
%
%
%
%
g kg-1 soils
0-15
7.4
20
20
60
0.106
0.7
113.7
15-30
7.8
20
32
60
0.069
0.6
113.8
30-45
8.0
20
32
48
0.067
0.5
114.7
45-60
8.1
18
30
48
0.041
0.3
114.5
60-90
8.4
14
30
52
0.040
0.3
114.8
90-140
6.4
14
30
56
0.038
0.3
117.2
Table 4. Physical and chemical properties of Vertisols of Shire.
Depth
pH
cm
Sand
Silt
Clay
TN
OC
Si
%
%
%
%
%
g kg-1 soils
0-15
6.9
10
32
58
0.057
0.7
133.1
15-45
7.0
10
32
58
0.069
0.6
133.9
45-60
7.3
12
28
60
0.073
0.6
130.9
60-90
7.5
12
28
60
0.050
0.5
131.9
90-140
7.5
12
28
60
0.043
0.7
130.0
78
Table 5. Physical and chemical properties of Vertisols of Wukro.
Depth
pH
cm
Sand
Silt
Clay
TN
OC
Si
%
%
%
%
%
g kg-1 soils
0-15
8.2
50
20
30
0.078
1.0
166.3
15-30
8.2
40
20
40
0.088
1.4
153.8
30-45
8.5
33
23
44
0.090
1.1
141.1
45-60
8.3
33
23
44
0.092
1.1
137.3
60-90
8.5
31
20
49
0.060
1.0
141.3
90-140
8.7
13
38
49
0.066
1.0
144.9
140-160
8.5
36
17
47
0.063
0.9
146.5
Table 6. Correlation between silicon and major soil physical and chemical properties in Vertisols
of Tigray.
Soil properties Adigudom
Axum
Maychew
Shire
Wukro
pH
0.53
0.67
‐0.67
‐0.81
‐0.56
Sand
‐0.08
‐0.73
‐0.75
‐0.88(*)
0.65
Silt
‐0.85(*)
‐0.01
0.30
0.88(*)
‐0.22
Clay
0.85(*)
0.45
‐0.16
‐0.88(*)
‐0.85(*)
Total N
0.39
0.77(*)
‐0.65
0.43
0.00
Soil organic C
‐0.15
0.67
‐0.73
0.70
‐0.05
* Correlation is significant at the 0.05 level (2-tailed).
4. DISCUSSION
The Si values in the Vertisols of Tigray are less than from the previously reported Si content for
clay soils. As silicon is the most abundant element in the earth’s crust after oxygen, it ranges
from 200 to 300 g Si kg-1 in clay soils and 450 to 480 g Si kg-1 in sandy soils (Kovda, 1973;
Matichenkov and Calvert, 2002). Although the Si contents in Vertisols of Debre Zeit and Sheno
in the Central Highlands of Ethiopia fall within this range (Fisseha Itanna, 1992), the clay soils in
Tigray Region have much lower Si. The comparatively lower values for Si in these soils can be
justified due to; first, severe and frequent soil erosion and sediment transportation in the study
areas. Secondly, usually plants absorb Si more than other elements (Savant et al., 1997); hence,
79
much Si should be replaced to ensure sustainable yields. Thirdly, the nature of parent materials
in the study areas could be the cause for the lowered Si levels in these soils. Virtually, the main
portions of soil Si-rich compounds are represented by quartz or crystalline silicates, which are
confirmed in this current studies where the highest Si content was found in sand dominated soils.
The physically and chemically active Si substances in the soils are represented by soluble and
weakly adsorbed monosilicic acids, polysilicic acids, and organosilicon compounds
(Matichenkov and Ammosova, 1996). These forms are interchangeable with each other as well
as with other crystalline minerals and living organisms (soil microorganisms and plants).
Monosilicic acid is the center of these interactions and transformations. Monosilicic acids are the
product of Si-rich mineral dissolution (Lindsay, 1979). Plants and microorganisms can absorb
only monosilicic acid (Yoshida, 1975). Polysilicic acid has a significant effect on soil texture,
water holding capacity, adsorption capacity, and soil erosion stability (Matichenkov et al., 1995).
The study concludes that quantifying monosilicic and polysilicic acid contents and conduct
applied research so as to elaborate optimum Si rate and best time and methods of its application
are imperative. It is also advisable to consider additional Si extractants to those used in this
study. Plant uptake studies may also be useful to justify whether it contributes to the low levels
of Si in Tigray Region.
5. ACKNOWLEDGEMENT
The researcher duly acknowledges the Ethiopian Institute of Agricultural Research for funding
the project and technicians of the Department of Land Resource Management and Environmental
Protection of Mekelle University for sample collection.
6. REFERENCES
Bremner, J.M. & C.S. Mulvaney. 1982. Total Nitrogen. In: Page, A.L. Miller, R.H. and Keeney
D.R. (eds.), Methods of soil analysis, Part 2 - Chemical and microbiological properties.
Agronomy, Am. Soc. of Agron., Madison, Wisconsin, 9 (2): 595-624.
Epstein, E. 1999. The discovery of the essential elements. In: Kung S.D. and Yang S.F. (eds.),
Discoveries in plant biology, Volume 3, World Scientific Publishing, Singapore.
Fassil Kebede, 2002. Analysis of yield gap for wheat cultivation in the highlands of north
Ethiopia. PhD thesis, Gent University, Belgium.
80
Fisseha Itanna. 1992.
Micro- and macronutrient distributions in Ethiopian Vertisol landscapes. PhD Thesis.
Hohenheimer Bodenkundliche Hefte (2), University of Hohenheim, Stuttgart, Germany.
Iler, R. K. 1979. The chemistry of silica – solubility, polymerization, colloid and surface
properties, and biochemistry. Wiley & Sons, New York, pp. 866.
Jones & Handreck. 1967. Silica in soils, plants and animals. Adv. Agron. 19: 107-149.
Kovda, V.A. 1973. The bases of learning about soils. Volume 2, Nauka, Moscow.
Lindsay, W.L.1979. Chemical Equilibria in Soil. John Wiley & Sons, New York.
Marschner, H. 1996. Mineral Nutrition of Higher Plants. Academic Press INC San Diego, CA
92101.
Matichenkov, V.V., Pinsky, D.L. & Bocharnikova, E.A. 1995. Influence of mechanical
compaction of soils on the state and form of available silicon. Eurasian Soil Science 27
(12):58-67.
Matichenkov, V.V. & Ammosova. 1996. Effect of amorphous silica on soil properties of a sodpodzolic soil. Eurasian Soil Science 28 (10):87-99.
Matichenkov, V.V. & Bocharnikova, E.A. 2000. Comparison study of soil silicon status in sandy
soils of south Florida. Soil Crop Sci. Florida Proc. 59:132-137.
Matichenkov, V.V. & Calvert D.V. 2002. Silicon as beneficial element for sugarcane. J.
American Society of Sugarcane Technologists, Volume 22, 2002.
Savant, N.K., Korndorfer, G.H. & Datnoff, L.E. 1997. Silicon management and sustainable rice
production. Advances in Agronomy, 58:151-199.
Takahashi & Miyake, 1977. Silica and plant growth. Proc. Int. Seminar, Soil Environ. Fert.
Manage. Intensive Agriculture 1977, 603-611.
Walkley A. & Black, I.A. 1934. An examination of the Degtijareff method for determining soil
organic matter and proposed chromic and titration method. Soil Science, 37:29-38.
Yoshida, S. 1975. The physiology of silicon in rice. Bulletin. No.25, Food FERT. Tech. Centre,
Taipei, Taiwan.
81
Tillage Frequency, Soil Compaction and N-Fertilizer Rate Effects on Yield of
Teff (Eragrostis Tef (Zucc) Trotter) in Central Zone of Tigray, Northern
Ethiopia
Haftamu Gebretsadik1, Mitiku Haile2 and *Charles F. Yamoah2
1
Tigray Agricultural Research Institute, Axum Agricultural Research Centre, P.O. Box 230,
Axum, Tigray, Ethiopia.
2
Department of LaRMEP, College of Dry Land Agriculture and Natural Resource Management,
Mekelle University, P.O. Box 231, Mekelle, Ethiopia (*[email protected]).
ABSTRACT
In Ethiopia, teff is grown for its grain and straw. There is a dearth of information with respect to
plowing, compaction and fertilization on vertisols in central zone of Tigray. Therefore, this study
was conducted to determine the effects of plowing frequency, soil compaction and N on teff
yields. The experimental design was a split-split plot where main plot was plowing frequency
with three levels (once, twice and thrice); sub-plot was compaction with two levels (with and
without compaction) and sub-sub plot was N-fertilizer rate with four levels (0kgN/ha, 46KgN/ha,
69KgN/ha and 92KgN/ha). There were three replications. Results showed that plowing
frequency had no significant effect on most of the yield components except on tillering when the
soil is compacted. Maximum average number of tillers per plant (2.75) was obtained from
compacted plots plowed two times. Compaction affected almost all yield and yield components
significantly. Higher number of tillers per plant (2.64) from non compacted plots and higher
stand cover (about 94%) from compacted plots were found. In addition, maximum biomass
(4210.617kg/ha) and grain (1221.98kg/ha) yields were obtained from compacted plots due to
enhanced soil to seed contact resulting in increased plant population. Nitrogen fertilizer
significantly increased grain yield and yield components. Maximum stand cover (94.78%), plant
height (92.16cm), panicle length (37.75cm), biomass yield (4724kg/ha) and grain yield
(1387.9kg/ha) were found from plots receiving 92kgN/ha. Partial budget analysis of N fertilizer
rates indicted that higher marginal rate of return (525%) were found by applying 69kgN/ha. It is
recommended that farmers use 69kgN/ha so as to get economically feasible returns and yield.
Keywords: Compaction, Fertilizer N, Plowing frequency, Teff, Vertisol.
1. INTRODUCTION
Tef (Eragrostic teff (zucc.) Trotter) is annual C4 grass that belongs to the family Poacea
(Kebede, et al., 1989). The crop exhibits high variability within regions of cultivation and
between plants of the same accession (Tadesse, 1993). It is primarily self pollinated and provides
naturally inbred lines, though intermittent cross pollination introduces new genetic material to
82
existing teff population (Nicole, 1999). Teff is a grown both for its grain and straw.staple cereal
crop in Ethiopia.
In Ethiopia, teff performs well in ‘Weina dega’ agro-ecological zones or medium altitude (17002400 m above sea level) (Nicole, 1999; Deckers et al., 2001). According to Deckers et al. (2001),
the mean temperature and optimum rainfall during the growing season range from 10oC to 27 oC
and 450 to 550mm, respectively. The length of growing period (LGP) or the number of days to
maturity of teff, considering rainfall and evapo-transpiration of 2-6 mm/day ranges from 60 to
180 days (depending on variety and altitude) with an optimum of 90 to 130 days (Deckers et al.,
2001).
The first plowing for teff production in most part of the country is done as soon as the previous
crop is harvested. In less weed prone areas, it is done after the onset of the small (belg) or main
(kiremt) rainy seasons (Fufa et al., 2001). Teff needs high tillage frequencies as compared to
other cereal crops in Ethiopia. Also, it requires firm, level seedbed, free from clods and stumps
(Deckers et al., 2001). With respect to teff cultivation on Vertisols, several plowings are
necessary, occasionally as much as 12 times, relative to Nitosols (Deckers et al., 2001).
According to Kenea et al. (2001), the tillage frequency for teff in Ethiopia ranges from 3 times in
Nazareth to 12-times in western Wellega. Though research results indicated teff grain yield
increased with increasing number of plowings (IAR, 1998). Others recommended tillage
frequency for teff to be 3-5 times (Melesse, 2007), 5-9 times especially in high rainfall areas
(Tarekegn et al., 1996) and 4-times (Nyssen et al., 2000). Generally, the tillage frequency is not
consistent from region to region, from soil type to soil type and from farmer to farmer. This
suggests further research on tillage frequency for teff.
Due to shrink-swell characteristics of Vertisols, teff seeds broadcasted on them during sowing
need moderate soil compaction to enhance their attachment with the soil. Most teff growing
farmers on Vertisols trample their teff seedbed either before planting or soon after; in both
occasions using cattle, sheep, goats and/or donkeys (Tadesse, 1969). According to TARI (2007),
trampling makes seedbed firm and flat, prevents the soil surface from quick drying, provides a
thin coverage of broadcasted seeds, prevents seeds from desiccation and consequently enhances
good germination and seedling establishment ( TARI, 2007; Fufa et al., 2001). In addition, the
results of comparative studies conducted for three years at Debre Zeit on two soil types
(Inceptisol and Vertisol) also revealed that the effect of seedbed trampling by humans and oxen
83
on grain yield was not significant except at Chefe Donsa where a significant yield increase was
obtained (Fufa et al., 2001). Melesse (2007) reported the necessity of teff seedbed trampling and
recommended the use of subsoilers to break hard pans. According to TARI (2007), some farmers
in central and North-Western zones of Tigray trample their teff seedbed using human feet when
the land is small and there is less availability of animals.
Major factors affecting teff fertilizer recommendation are water logging, seasons of planting,
cropping history, lodging and weed growth (Kenea et al., 2001). The actual rate of fertilizer used
by farmers is below the blanket recommendation i.e. 100kg DAP/ha and 100kg urea/ha set by the
Ministry of Agriculture and Rural Development (Kenea et al., 2001). Generally, the
recommended rate of fertilizer for teff is 25 to 40 kgN/ha and 30 to 40kgP2O5/ha on light soils
such as Nitosols, Luvisols and Cambisols, and 50-60kgN/ha and 30-35 kgP2O5/ha for heavy soils
such as Vertisols (Deckers et al., 2001). Compaction narrows soil pores and reduces water
infiltration which result in water logging problem in vertisols. Such water logging causes N
losses by denitrification. Therefore, determining the optimum rate of N under those agronomic
practices is necessary.
2. MATERIALS AND METHODS
2.1. Site description
The research was conducted in northern Ethiopia, central zone of Tigray, Wereda La’elay
Maichew and Tabia Hatsebo. It is 5 km east of Axum town (38034’ and 39025’ east, and 13015’
and 14039’ north). Altitude is of 2050 m and classified as sub humid agro-ecology where most of
the middle altitude crops such as teff, wheat, Fababean are commonly grown. The rainfall ranges
from 300 to 800mm/annum. The major arable crop in the area is teff that is mostly widely
distributed soil type in the area.
2.2. The experimental design
A split-split plot design with three replications was used. The main factor was three plowing
frequencies (once, twice and thrice); the sub factor was two compaction levels (with and without
compaction) and the sub-sub factor was four rates of N-fertilizer (0kgN/ha (control), 46kgN/ha,
69kgN/ha and 92kgN/ha). The teff variety DZ-01-974 (Dukem) was used and sown at a rate of
30kg/ha. Treble Super Phosphate (TSP) was used as a source of P and applied uniformly to all
plots at rate of 60 kgP2O5/ha (a national recommendation of P for teff grown on vertisols).
84
Those plots, which received three plowing frequencies, were plowed first at the beginning of the
cropping season; second two months after and the third at sowing. The plots, which received two
plowing frequencies, were plowed first, at the beginning of the season and at sowing. The plots,
which received one plowing, were plowed only at sowing.
Sub plots were compacted by trampling with human labor soon after sowing and broadcasting
urea and TSP fertilizers. The N fertilizer rates were applied to the sub-sub plots of each
compacted and non-compacted plots, half at sowing and the rest at top dressed at time of the crop
tillering (i.e. one month after sowing). Penetrometer resistance was measured for each sub plots
at sowing (0-30 cm) soil depth. Penetrometer readings were recorded at 5cm interval along the
soil depth. Bulk density was determined by the core sample approach.
Stand percentage, which refers to the percentage of the plot area covered by teff seedlings, was
estimated when the teff seedlings reach three to four leaves age. Plant height, panicle length and
tillering potential were measured by randomly selecting 10 plants per plot. Both biomass and
grain yields were measured after sun drying for one week. Harvest index (HI) was calculated by
the formula defined by Fleischer et al., (1989):
HI = Grain yield (kg/ha) x 100
Total biomass yield (kg/ha)
2.3. Partial Budget Analysis
Variable cost of N fertilizer was largely used for partial budget analysis. Price fluctuations
during the production season was considered. Marginal Rate of Return, which refers to net
income obtained by incurring a unit cost of fertilizer, was calculated by dividing the net increase
in yield of teff due to the application of each rate to the total cost of N fertilizer applied at each
rate. This enables us to identify the optimum rate of N fertilizer for teff production.
3. RESULTS AND DISCUSSION
3.1. Penetration Resistance and Bulk Density
There is no significant difference in penetration resistance between compacted and noncompacted plots as well as within each soil depth in both plots. The average penetration
resistance (MPa) in both compacted and non-compacted plots is 0.83 and 0.77, respectively.
Maximum penetrometer resistance is observed at soil depth of 10-15cm in compacted and 20-
85
25cm in non-compacted plots. Ehlers et al., (1983) indicated that a penetrometer resistance of
3.6MPa and above under tilled field could impede root growth of oats which is similar to root
system of teff. In addition, any penetrometer resistance over 2.0MPa can significantly reduce
root growth and development (Ishaq et al., 2001; Oussible et al., 1992). Therefore, the average
penetration penetrometer from the compacted plots of 0.83MPa is lower than the values
mentioned in the literature. This means compaction of the plots does not affect the root
penetration of teff. However, there is slight change in bulk density of both plots, 0.79g/cm3 in
non-compacted plots against 0.98g/cm3 in compacted plots.
Table 1. Main effects of plowing frequency, compaction and N-fertilizer rates on tillering
potential and stand cover.
Factors
Tillering potential
(tiller/plant)
Stand percentage
Plowing frequency
P1=once
2.33
90.75
P2=twice
2.67
91.25
P3=thrice
2.38
92.208
SE
0.116
1.500
LSD
ns
Ns
CV (%)
18.77
3.55
compaction
Non compacted
2.64a*
88.56b
b
Compacted
2.28
94.25a
SE
0.094
0.520
LSD
0.259
5.374
CV (%)
18.77
3.55
N-fertilizer rate
0kgN/ha
2.28
83.06c
46kgN/ha
2.50
92.72b
69kgN/ha
2.67
95.06a
92kgN/ha
2.40
94.78ab
SE
0.134
0.735
LSD
ns
3.441
CV (%)
18.77
3.55
*Means not connected by the same letters are significantly different at alpha 0.05.
86
3.2. Yield and yield component
3.2.1. Stand cover and Tillering
Plowing frequency did not show any significant effect on tillering potential, and stand cover of
teff. However, compaction has significant effect on these yield attributes. This may be due to its
influence on absorption of nutrients and moisture by the crop. Relatively, higher stand cover
were found from compacted plots but higher average number of tiller per plant was found from
non compacted plots. The higher stand cover from the compacted plots is due to the good
attachment of teff seeds with the soil which might have resulted in efficient use of nutrients and
moisture as well as improved germination. Except in stand cover, the three N-fertilizer rates did
not show significant difference from the control in number of tillers per plant. Highest stand
cover (95%) was obtained from application of 69 kg N/ha.
Interaction of plowing frequency and compaction significantly affect tillering potential of teff
(Table.1). Relatively higher number of tillers per plant (~3 tillers/plant) was found from plots
plowed two times and compacted than the others (Table 2). In addition, interaction of
compaction with N rates significantly affect tillering because compaction enables teff plants to
properly utilize the applied N fertilizer (Table 3).
Table 2. Interaction effect of plowing frequency and compaction on tillering potential of teff.
Compaction
Plowing frequencies
P1
P2
P3
C0
2.67ab*
2.58ab
2.67ab
C1
2.00b
2.75a
2.08ab
*Means not connected by the same letters are significantly different at alpha 0.05.
Where: P1 = plowing first, P2 = plowing twice, P3 = plowing thrice, C0 = non-compacted plots
and C1 = compacted plots, LSD0.05 = 0.575, SE = 0.163, CV (%) = 18.8
87
Table 3. Interaction effect of compaction and N-fertilizer rate on tillering potential of teff
N-rate (kgN/ha)
Compaction
Non Compacted
Compacted
0
2.56ab*
2.00b
46
2.44ab
2.56ab
69
2.22b
3.11a
92
2.44ab
2.33ab
*Means not connected by the same letters are significantly different at alpha 0.05.
Where, LSD0.05 = 0.426, SE = 0.190, CV (%) = 18.8
Interaction effect of compaction and N fertilizer on ground cover was significant. Therefore,
compaction promotes emergence of teff seedlings through efficient utilization of soil nutrients
i.e. the applied N and moisture. Relatively, greater ground cover was obtained from compacted
plots receiving higher N fertilizer rates (Table 4).
Table 4. Interaction effect of compaction and N-fertilizer rate on stand cover of teff.
N-rate (kgN/ha)
Compaction
Non Compacted
Compacted
0
78.33d
87.78c
46
90.00bc
95.44a*
69
93.33ab
96.78a
92
92.56ab
97.00a
*Means not connected by the same letters are significantly different at alpha 0.05.
3.2.2. Plant Height and Panicle Length
Only N fertilizer rate caused significant effect in yield attributes. Both plowing frequency and
compaction had not any significant effect on plant height and panicle length (Table 5).
Teff plants with higher plant height (92cm) and panicle length (38cm) were found by applying
high amount N fertilizer (92kgN/ha) (Table 5). This is because high nitrogen usually favors
vegetative growth of teff which results in taller teff plants heights having relatively greater
panicle length.
88
3.2.3. Biomass Yield, Grain Yield and Harvest Index (HI)
There were no significant differences among plowing frequencies in biomass yield, grain yield
and harvest index of teff (Table 6). Though IAR (1998) as cited in Fufa et al. (2001), reported an
increase in grain yield of teff with an increase plowing frequency, such report contradicts our
present findings.
Table 5. Main effects of plowing frequency, compaction and N-fertilizer rates on plant height
and panicle length.
Factors
Plant height (cm)
Panicle length (cm)
Plowing frequency
0nce
79.63
34.27
Twice
78.75
33.98
Thrice
80.77
34.38
SE
0.973
0.58
LSD
ns
ns
CV (%)
5.39
7.46
Non compacted
79.05
34.85
Compacted
80.38
33.68
SE
0.793
0.47
LSD
ns
ns
CV (%)
5.39
7.46
0kgN/ha
57.73d
27.42d
46kgN/ha
81.82c
35.10c
69kgN/ha
87.16b
36.80b
92kgN/ha
92.16a*
37.75a
SE
1.12
0.67
LSD
1.58
0.95
CV (%)
5.39
7.46
compaction
N-fertilizer rate
*Means not connected by the same letters are significantly different at alpha 0.05.
89
Table 6. Main effects of plowing frequency, compaction and N-fertilizer rates on biomass yield,
grain yield and harvest index.
Factors
Biomass yield (kg/ha) Grain
yield Harvest
Index
(kg/ha)
(HI)
Plowing frequency
Once
3726.67
1095.09
0.291
twice
3753.80
1098.24
0.289
thrice
4011.48
1206.30
0.300
SE
108.64
37.14
0.004
LSD
ns
ns
ns
CV (%)
13.23
15.15
6.00
compaction
Non compacted
3450.679b
1044.4b
0.297
a*
a
Compacted
4210.617
1222
0.289
SE
88.70
30.33
0.003
LSD
125.40
42.89
ns
CV (%)
13.23
15.15
6.00
N-fertilizer rate
0kgN/ha
2368.27d
665.8d
0.279b
46kgN/ha
3758.77c
1120.6c
0.298a
b
a
69kgN/ha
4471.48
1358.5
0.304a
a
a
1387.9
0.293ab
92kgN/ha
4724.07
SE
125.44
42.89
0.005
LSD
177.40
60.70
0.007
CV (%)
13.23
15.15
6.00
*Means not connected by the same letters are significantly different at alpha 0.05.
There were significant differences between compacted and non-compacted plots in both biomass
(P = 0.0007) and grain yields (P = 0.0035) but not of harvest index (Table-6). This agrees with
research results from Chefe Donsa that a significant grain yield of teff was obtained from
seedbed compaction (Fufa et al., 2001). Greater biomass (4211kg/ha) and grain (1222kg/ha)
yields were obtained from compacted plots than non-compacted plots. The reason for the high
yield from compacted plots was due to early emergence and seedling growth as well as high teff
ground cover.
Nitrogen fertilizer significantly (P = 0.0001) improved biomass and grain yield in the control
plots and harvest index (P = 0.001) (Table 6). The biomass (4,724 kg/ha) and grain (1388kg/ha)
yields were obtained by applying 92 kgN/ha. However, greater harvest index (0.304) was
obtained by applying 69kgN/ha which means that more grain yield per unit biomass was
obtained from plots receiving this rate.
90
3.3. Partial Budget Analysis of N-Fertilizer rate
The partial budget analysis indicates that high marginal rate of return (525%) was obtained by
applying 69kgN/ha (Table 7). This means that the income obtained by applying 69kg N/ha for
teff was more than 5 times a unit total N-fertilizer cost. It is also fair to farmers to use 46kg N/ha
for it can bring comparable income which is more than 5 times total N-fertilizer cost. The latter
seems to support the nationally recommended N-fertilizer rate for teff. This analysis is done by
considering only grain yield of teff. If we add the value of the straw, the return will become more
than the already estimated income.
Table 7. Partial budget analysis for N-fertilizer rates.
N-fertilizer rates (Ni)
Total cost and total income
0kgN/ha 46kgN/ha
69kgN/ha
92kgN/ha
Cost
1. Labor cost
•
Fertilizer weighing and taking to field
0.00
30.00
30.00
30.00
•
Fertilizer Application
0.00
30.00
30.00
30.00
2. Fertilizer cost
0.00
400.00
600.00
800.00
Total cost (Tc)
0.00
460.00
660.00
860.00
Economic (grain) yield (kg/ha)
666
1121
1359
1388
Income (birr/ha)
3330
5605
6795
6940
Change in cost (∆C)
0.00
460
660
860
Change in income (∆I)
0.00
2275
3465
3610
Marginal Rate of Return (MRR =
0.00
494.6%
525%
419.8%
Income
(∆I)/(∆C)*100)
Where,
1. ∆C = cost of each rate
subtracted
from cost of the control
2. ∆I = income of each rate (Birr/ha)
subtracted from the income of
control
3. Labor costs were calculated by assuming 15birr/labor/day
91
4. Incomes were calculated by assuming 500birr/100kg of teff.
4. CONCLUSION AND RECOMMENDATION
Plowing frequency did not affect most of the yield and yield attributes of teff. However, it
affected tillering of teff when combined with compaction. Incidentally, this factor alone did not
increase the number of tillers per plant. Generally, frequent plowing is not a major factor to
increase teff productivity on Vertisols, but it is important to control weeds.
Compaction resulted in low tillering but high ground cover. This means compaction did not
favor tillering of teff but it encourages ground cover. Interaction between compaction and N also
influenced tillering and ground cover of teff. Optimum returns were found from biomass and
grain yields on compacted plots. The overall outcome of compaction is that it affects teff
productivity on vertisols.
Nitrogen encourages ground cover with the application of 69kgN/ha. High rate of nitrogen
(92kgN/ha) resulted in taller plants with relatively longer panicles and greater average biomass
and grain yields. The partial budget analysis shows that application of 69kgN/ha can bring an
income which is more than five times the cost of N-fertilizer. Farmers should not plow their land
more than three times as more frequent plowing do not enhance teff productivity on Vertisols.
Plowing three times is appropriate for weed control. Compaction contributes to teff productivity
by facilitating good seed-soil contact on Vertisols hence, farmers should be encouraged to
practice it.
Application of 69kgN/ha gave optimum biomass and grain yields of teff and hence farmers have
to use this rate for teff production on Vertisols. As an alternative, farmers can also use 46kgN/ha
to get comparable income from the crop on Vertisols. Generally, this implies that farmers can
apply N-fertilizers at a rate ranging from 46-69kgN/ha to get optimum teff yield on vertisols.
5. REFERENCES
Deckers J.A., Nachtergaele, F.O. & Spaargaren, O.C. (eds.). 1998. World Reference Base for
Soil Resource: Introduction. Acco Leuven/Amersfoort. Belgium. pp.165.
DZARC (Debre Zeit Agricultural Research Centre). 1989. Annual research progress report for
1988/89. Debre Zeit, Ethiopia.
92
Ehlers, W., Popke, V., Hesse, F. & Bohm, W. 1983. Penetration Resistance and Root Growth of
Oats in Tilled and Untilled Loam Soil. Soil Tillage Res., 3:261-275.
Fufa Hundera., Tesfaye Bogale., Hailu Tefera., Kebebew Asefa., Tiruneh Kefyalew., Abbera
Debelo & Seifu Ketema. 2001. Agronomy Research in Teff. In: Hailu Tefera, Getachew
Belay & M.Sorrels (eds.), Narrowing the Rift: Teff Research and Development. Ethiopian
Agricultural Research Organization (EARO). Addis Ababa, Ethiopia, pp.167-176.
IAR. 1998. Holleta Agricultural Research Center, progress report for the period April 1997 to
March 1998. IAR, Addis Ababa, Ethiopia.
Ishaq M., Ibrahim, M., Hassan, A., Saeed M. & Lal. R. 2001. Sub-soil Compaction Effects on
Crops in Punjab Pakistan. II. Root Growth and Nutrient Uptake of Wheat and Sorghum.
Soil Tillage Res.
Kebede H., Johnson, R.C. & Frris, D.M. 1989. Photohsynthetic Response of Eragrostis tef. to
Temperature. Physiol. Plant. 77:262-266.
Kenea Yadeta, Getachew Ayele. & Workneh Negatu. 2001. Farming Research on Teff: Small
Holders Production Practices. In: Hailu Tefera, Getachew Belay and M.Sorrels (eds.).
Narrowing the Rift: Teff Research and Development. Ethiopian Agricultural Research
Organization (EARO). Addis Ababa, Ethiopia, pp.9-23.
Melesse Temesgen. 2007. Conservation Tillage Systems and Water Productivity Implications for
Small Holder Farmers in Semi-Arid Ethiopia. PhD dissertation. Delft University of
Technology. The Netherlands.
Nicole D.J. 1999. Teff and Fingermillet: Archaeobotanical Studies of two Indigenous East
African Cereals. Master thesis. Simon Fraser University.
Nyssen J., Poesen, J., Mitiku Haile., Moeyersons, J. & Deckers, J. 2000. Tillage erosion on
slopes with soil conservation structures in the Ethiopian highlands. In: soil and tillage
research: incorporating
soil technology. The International Soil Tillage Research
Organization (ISTRO). Elsevier Science B.V. 57:115-127.
Oussible M., Crookstone, P.K. & Larson, W.E. 1992. Sub-surface Compaction Reduces the Root
and Shoot Growth and Grain of Wheat. Agron. J. 84:34-38.
Taddesse Ebba. 1969. Tef (Eragrostis tef): the cultivation, usage and some of the Known
Diseases and Insect Pests. Part-2. Debre Zeit Agricultural Research Station. Bulletin no.
60, Haile Slassie I University, College of Agriculture, Dire Dawa, Ethiopia.
93
Tadesse D. 1993. Study on Genetic Variation of Landraces of Teff (Eragrostis tef(Zucc.) Trotter)
in Ethiopia. Genetic Resources and Crop Ewfutim 40: 10 1-104.
Tarekegne, A., Gebre, A., Tanner, D. G. & Mandefro, C. 1996. Effect of Tillage Systems and
Fertilizer Levels on Continous Wheat Production in Central Ethiopia. In: Tanner, D. G.,
Payne, T. S. & Abdalla, O. S. (eds.), The Ninth Regional Wheat Workshop for Eastern,
Central and Southern Africa. Addis Ababa, Ethiopia, CIMMYT, pp56-63.
Tigray Agricultural Research Institute (TARI). 2007. Axum Agricultural Research Center.
Progress report for the period from February 2006 to December 2007.
94
Multidimensional and Multi-Parameter Fortran-Based Curve Fitting Tools
Daniel Tsegay and *Alem Mebrahtu
Department of Physics, College of Natural and Computational Sciences, Mekelle University,
P.O. Box. 3044, Mekelle, Ethiopia (*[email protected])
ABSTRACT
The Levenberg-Marquardt algorithm has become a popular method in nonlinear curve fitting
works. In this paper, following the steps of Levenberg-Marquardt algorithm, we extend the
framework of the algorithm to two and three dimensional real and complex functions. This work
briefly describes the mathematics behind the algorithm, and also elaborates how to implement it
using FORTRAN 95 programming language. The advantage of this algorithm, when it is
extended to surfaces and complex functions, is that it makes researchers to have a better trust
during fitting. It also improves the generalization and predictive performance of 2D and 3D real
and complex functions.
Keywords: Levenberg-Marquardt algorithm, Nonlinear curve fitting and Least square fitting
technique.
1. INTRODUCTION
Levenberg-Marquardt (LM) algorithm is an iterative technique (Levenberg, 1944; Kelley, 1999;
Avriel, 2003; Marquardt, 1963; Bates & Watts, 1988; Box, et al., 1969; and Gill, et al., 1981)
which helps in locating the discrepancy between a given model and the corresponding data. Such
functions are usually expressible as sum of squares of nonlinear functions. The LM algorithm
has become a standard technique for nonlinear least-square problems (Lourakis, 2005; Lampton,
1997; Arumugam, 2003; Coope, 1993; and Madsen, et al., 2004) and can be thought of as a
combination of steepest descent and the Gauss-Newton methods. The paper is presented as
follows: In section one, we present a brief introduction about the LM algorithm. In section two
we discuss about the least square fitting technique. Section three elaborates Vanilla Gradient
descent method. In the fourth section we present Newton’s method. A more detailed discussion
of LG algorithm is presented in section five. Section six discusses about the implementation of
the LM algorithm. In the last section we present a brief summary of the paper.
95
1.1. Least-Square’ Fitting Technique
Suppose we have a set of N experimental data points {x i , yi , L , f i , σ i } , where i = 1 ,…, N for
which we need to make a fitting. Here X i ≡ ( xi , yi ,...) are the data coordinates, f i is the data
value and σ i is the data error bar. Next we take a model which can estimate the values of f as a
function of X i ≡ ( xi , yi ,...) and a set of internal variable parameters P ≡ ( p1 , p2 ,..., pM ) : f ( X , P ) .
Let us construct the chi-square function:
χ (P) ≡
2
N
∑
i =1
⎛ fi − f (X i , P)
⎜⎜
σi
⎝
where ri ( Pc ) =
f i − f ( X i , Pc )
σi
⎞
⎟⎟
⎠
2
=
N
∑
i =1
ri 2 ( P )
(1)
is called residue function.
The goal of the least square
method is to determine the parameters P of the regression function f ( X , P ) so as to minimize
the squared deviations between f i and f ( X i , P ) for all data points: i = 1L N . If we assume
that all measured values of f i are normally distributed with standard deviations given
by σ i , then ‘statistically-the-best’ match would correspond to the minimal value of χ
2
. Thus,
the suitable model is essentially the one which gives the minimum value of the chi-square with
respect to the parameters. That is why the method itself is called the ‘least-square’ technique. Of
course, the error bars are determined not only by a statistical noise, but also by systematic
inaccuracies, which are very difficult to estimate and are not normally distributed. However, to
move on, we assume that they are some how accounted for by the values σ i . Other approaches
that are useful in determining the best-fit parameters for non-linear functions f ( X , P ) by
minimizing χ 2 iteratively include Newton’s method and Gradient descent method.
1.2. Vanilla Gradient Descent Method
The Gradient descent method is simply an instinctive moving in the ‘steepest descent’ direction,
which is apparently determined by the minus-gradient:
βk
1 ∂ χ 2 ( Pc )
=
≡ −
2
∂pk
N
∑
i =1
ri ( Pc )
∂ ri ( Pc )
=
∂pk
or
96
N
∑
i =1
f i − f ( X i , , Pc ) ∂ f
( X i , Pc )
σ 12
∂pk
⎡ β1
⎢
⎢
⎢ β2
⎢
⎢
⎢ M
⎢β
⎣ M
⎡ ∂ r1 ( P c )
⎤
⎢ ∂p
⎥
1
⎢
⎥
r
(
P
∂
c )
1
⎢
⎥
=
⎢
∂p2
⎥
⎢
⎥
M
⎢ ∂r (P )
⎥
c
1
⎢
⎥
⎦
⎣⎢ ∂ p M
∂ r2 ( Pc )
∂ p1
∂ r2 ( Pc )
∂p2
M
∂ r2 ( Pc )
∂pM
L
L
O
L
∂ rM ( Pc
∂ p1
∂ rM ( Pc
∂p2
M
∂ rM ( Pc
∂pM
)⎤
⎥
⎥
)⎥
⎥
⎥
)⎥
⎥
⎦⎥
⎡ r1 ( P c )
⎢
⎢
⎢ r2 ( Pc )
⎢
⎢
M
⎢
⎢r
⎣ M ( Pc )
⎤
⎥
⎥
⎥ .
⎥
⎥
⎥
⎥
⎦
(2)
In compact form β = − 1 ∇ χ 2 = [J ]T r ( P ) ,
2
Where J is called Jacobian matrix of the residue ri ( Pc ) which is defined in Eqn. 1. The onehalf coefficient is put to simplify the formulas. To improve the fit, we can shift the parameters
p kc → p kc + δp k , where δp k = cons tan t × β k
⎡ β1
⎡ δp1 ⎤
⎢
⎢ δp ⎥
⎢ 2 ⎥ = cons tan t × ⎢ β 2
⎢ M
⎢ M ⎥
⎢
⎢
⎥
⎣β M
⎣δpM ⎦
⎤
⎥
⎥ .
⎥
⎥
⎦
(3)
The steepest descent strategy is justified, when one is far from the minimum, but suffers from
slow convergence in the plateau close to the minimum, especially in the multi-parameter space.
Logically we would like large steps down the gradient at locations where the gradient (slope) is
small (near the plateau) and small steps when the gradient is large not to rattle out of the
minimum. Moreover, it has no information about the scale or the value of the constant and one
can see that δpk = cons tan t × β k has a problem with the unit dimensions.
1.3. Newton’s Method
Newton's method is an algorithm used for finding roots of equations in one or more dimensions.
Let
us
Pc ≡
expand
∇ χ 2(P )
( p 1 c , p 2 c ,...
using
a
Taylor’s
series
around
the
current
points,
p Mc ) , we get
∇ χ 2 ( P ) = ∇ χ 2 ( Pc ) + [δ P ]T ⋅ ∇ 2 χ 2 ( P c ) + higher order terms
97
(4)
⎡ α 11
⎢α
⎡ ∂χ 2 ( Pc ) ∂χ 2 ( Pc )
2
∂χ 2 ( Pc ) ⎤
⎢ 21
∇χ (P) = ⎢
,
, L,
⎥ + ⎢ M
∂p2
∂pM ⎦
⎣ ∂p1
⎢
⎣α M 1
α 12
α 22
M
αM2
L
L
α 1 M ⎤ ⎡ δp1 ⎤
α 2 M ⎥⎥ ⎢⎢ δp2 ⎥⎥
+ higher
O
M ⎥ ⎢ M ⎥
⎥ ⎢ ⎥
L α MM ⎦ ⎣δpM ⎦
order terms,
where
δp k = p k − p kc , δP T = [δp1 , δp 2 , L , δp M ] and
⎡ ∂2χ 2
⎢
⎢ ∂ p1∂ p1
⎢ ∂2χ 2
⎢
= ⎢ ∂ p 2 ∂ p1
M
⎢
⎢ ∂2χ 2
⎢
⎣ ∂ p M ∂ p1
α
∂2χ 2
∂ p1∂ p 2
∂2χ 2
∂p 2∂p 2
M
2
∂ χ2
∂p M ∂p 2
∂2χ 2
∂ p1∂ p M
∂2χ 2
∂p 2∂p M
M
2
∂ χ2
∂p M ∂p M
L
L
O
L
2
Note that ∂ χ ( Pc ) is the gradient vector of χ
2
∂pk
α
kl
⎤
⎥
⎥
⎥
⎥
⎥ .
⎥
⎥
⎥
⎦
with respect to pk evaluated at Pc and
2
2
1 ∂ χ ( Pc )
is the second order gradient vector of χ
2
∂p x ∂p l
≡
2
(is called Hessian
matrix) evaluated at P c .
Near the current points Pc , we can approximate the value of χ 2 ( P ) up to the second order, as
∇ χ 2 ( P ) = ∇ χ 2 ( Pc ) + [δ P
]T
⋅∇
2
χ
2
( Pc ) .
Assuming the chi-square function is quadratic around Pc and solving for the minimum values of
the parameters P by setting ∇ χ 2 ( P ) = 0 , we get the update rule (the next iteration point) for
Newton’s methods:
[α ][δ P ]
T
[δ P ]T
= −∇ χ ( Pc )
= − [α
2
M
⇔
∑α
l =1
]− 1 ∇ χ 2 ( Pc )
kl
δp l = β k
(5)
2
⇒ P = Pc − [α ] ∇ χ ( Pc ) .
−1
(6)
The chi function (which is quadratic) to be minimized has almost parabolic shape. The Hessian
matrix, which is proportional to the curvature of χ 2 , is given by
98
α
kl
≡
1 ∂ 2 χ 2 ( Pc )
=
2 ∂p k ∂pl
∂ 2 f ( X i , Pc ) ⎫
1 ⎧ ∂f ( X i , Pc ) ∂f ( X i , Pc )
[
]
f
f
(
X
,
P
)
−
−
⎬
⎨
∑
i
i
c
2
∂p k
∂p1
∂p k ∂ p l ⎭
i =1 σ i ⎩
N
(7)
(the one-half here is also added for the sake of simplicity). The components α kl of the Hessian
matrix in Eqn. (7) depends both on the first derivative,
∂
2
f ( X i , Pc )
∂pk∂pl
∂ f ( X i , Pc )
∂p k
, and second derivative,
, of the basic function with respect to their parameters. The Second derivative can
be ignored when it is zero, or small enough to be negligible when compared to the term
involving the first derivative. In practice, this is quite often small enough to neglect. If one looks
at Eqn. (7) carefully, the second derivative is multiplied by [ f i − f ( X i , Pc ) ] . For the successful
model, this term should just be the random measurement error of each point. This error can have
either sign, and should in general be uncorrelated with the model. Therefore, the second
derivative terms tend to cancel out when summed over time i . Inclusion of second derivative
term can in fact be destabilizing if the model fits badly or is contaminated by outlier points that
are unlikely to be offset by compensating points of opposite sign. So, instead of Eqn. (7) we shall
N
define the α-matrix simply as: α kl ≡ ∑ 1 ∂f ( X i , Pc ) ∂f ( X i , Pc ) which is equivalent to
2
i =1
α = [J ]T J =
⎡ N 1 ∂ f ( X i , Pc ) ∂ f ( X i , Pc )
⎢∑ σ 2
∂ p1
∂ p1
⎢ i N= 1 i
1 ∂ f ( X i , Pc ) ∂ f ( X i , Pc )
⎢
2
⎢∑
∂p 2
∂ p1
i =1 σ i
⎢
M
⎢ N
⎢ ∑ 1 ∂ f ( X i , Pc ) ∂ f ( X i , Pc )
⎢⎣ i = 1 σ i2
∂p M
∂ p1
σi
∂p k
∂p l
1 ∂ f ( X i , Pc ) ∂ f ( X i , Pc )
2
∂ p1
∂p 2
i =1
i
N
1 ∂ f ( X i , Pc ) ∂ f ( X i , Pc )
∑
2
∂p 2
∂p 2
i =1 σ i
M
N
1 ∂ f ( X i , Pc ) ∂ f ( X i , Pc )
∑
2
σ
∂p M
∂p 2
i =1
i
N
∑σ
1 ∂ f ( X i , Pc ) ∂ f ( X i , Pc ) ⎤
⎥
2
∂ p1
∂p M
i
⎥
1 ∂ f ( X i , Pc ) ∂ f ( X i , Pc ) ⎥ .
∑
2
⎥
∂p 2
∂p M
i =1 σ i
⎥
M
⎥
N
1 ∂ f ( X i , Pc ) ∂ f ( X i , Pc ) ⎥
∑
2
⎥⎦
∂p M
∂p M
i =1 σ i
N
L
L
O
L
∑σ
i =1
N
(8)
After computing, numerically or analytically, the gradient and Hessian matrices for the current
set of parameters, one can immediately move to the minimum by shifting the parameters
p k → p k + δp k , where the displacement vector δp k
derived in Eqn. (5), i.e.,
99
is determined from the linear system
⎡ α 11 α 12
⎢α
⎢ 21 α 22
⎢ M
M
⎢
⎣α M 1 α M 2
L α 1M
L α 2M
O
M
⎤
⎥
⎥
⎥
⎥
L α MM ⎦
⎡ δ p1
⎢ δp
2
⎢
⎢ M
⎢
⎣δ p M
⎡ β1 ⎤
⎡ δp1 ⎤ ⎡ α11 α12
⎤
⎢
⎥
⎢
⎥
⎥ ⎢
β
⎥ = ⎢ 2 ⎥ ⇔ ⎢ δp 2 ⎥ = ⎢ α21 α22
⎢ M ⎥
⎢ M ⎥ ⎢ M
⎥
M
⎢ ⎥
⎢
⎥
⎥ ⎢
⎣β M ⎦
⎣δp M ⎦ ⎣α M 1 α M 2
⎦
L α1M ⎤
L α 2 M ⎥⎥
O M ⎥
⎥
L α MM ⎦
−1
⎡ β1 ⎤
⎢β ⎥
⎢ 2 ⎥.
⎢ M ⎥
⎢ ⎥
⎣β M ⎦
(9)
One of the problems associated with Newton’s method (Levenberg, 1944; Kelley, 1999; Madsen,
et al., 2004; and Lawson & R.J. Hanson, 1974) is its divergence after successive iterations. At
the instant when χ 2 ( Pc
+ δP )
diverges we would like to retreat to its previous value
χ
2
( Pc )
and
then decrease the steps, δP and try again.
2
χ (p 1 ,p 2 )
200
160
'Steepest D escent'
is efficient far
from the m inim um
120
80
8
40
-8
-4
p2
p1
4
0
'N ew ton's M ethod'
is efficient near
the m inim um
-4
0
4
-8
0
8
Figure 1. Graph of the chi function: The chi-square (χ2) function versus two arbitrary
experimental parameters P1 and P2.
1.4. The Levenberg-Marquardt Algorithm
In order for the chi-square function to converge to a minimum rapidly, one needs a large step in
the direction along with the low curvature (near the minimum) and a small step in the direction
with the high curvature (i.e. a steep incline). The gradient descent and Gauss-Newton iterations
provide additional advantages. The LM algorithm is based on the self-adjustable balance
between the two minimizing strategies: the Vanilla Gradient Descent and the Inverse Hessian
methods.
100
Coming back to the steepest descent technique χ 2 is dimensionless but β
dimension as
1
p
k
has the same
, as indicated in Eqn. (3). The constant of proportionality between β
k
and
k
δ p k must therefore have the dimension of
in kg , then β
k
p k2 . For instance, if the parameter
p k is measured
kg −1 so the constant must have a dimension of kg −2 .
has obviously the units of
Therefore the unit cannot be the same for all parameters since they are generally measured in
different units ( p1 in Seconds, p2 in Meter…
p
in Ampere). Marquadt surmised that the
M
components of the Hessian matrix must hold at least some information about the order-of–
magnitude scale and dimension. Among the components of α -matrix the reciprocal of the
diagonal elements
α kk −1
have these dimensions. Hence he suggested that this must set the scale
of the constant. To avoid the scale becoming too large, it is divided by a dimensionless positive
damping term,
λ
(being positive ensures that
δpk
is a descent direction). Eqn. (3) is then
replaced by
⎡ δ p1
⎢ δp
2
⎢
⎢ M
⎢
⎣δ p M
⎡ α 11− 1
⎤
⎢
⎥
1
⎢ 0
⎥ =
⎥
λ ⎢ M
⎢
⎥
⎢⎣ 0
⎦
0
α
In more compact form, δ p k =
0 ⎤
⎥
0 ⎥
M ⎥
−1 ⎥
α MM
⎥⎦
L
−1
M
L
O
0
L
22
1
λα
⎡ β1
⎢ β
⎢ 2
⎢ M
⎢
⎣β M
⎤
⎥
⎥
⎥ .
⎥
⎦
(10)
βk.
kk
In order to combine Eqns. (9) and (10), Marquardt defined a diagonally-enhanced new
matrix:
δ
kl
α kl′ = α kl (1 + δ kl λ ), where the value of the Kronicker delta function is given by
⎧ 0
= ⎨
⎩ 1
⎡ α 11′
⎢α ′
⎢ 21
⎢ M
⎢
⎣α M′ 1
α′ -
α 12′
′
α 22
M
α M′ 2
for
k ≠ l
for
k = l
L
L
O
L
such that
α 1′M ⎤ ⎡α 11 (1 + λ )
α 12
L
⎢
⎥
α 2′ M ⎥ ⎢ α 21
α 22 (1 + λ ) L
M ⎥ =⎢
⎥ ⎢
′ ⎦ ⎣
α MM
M
M
αM1
αM 2
(11)
101
O
α 1M
α 2M
L α MM
⎤
⎥
⎥
⎥
M
(1 + λ )⎥⎦
where λ is a dimensionless constant, and α kl is replaced with α kl′ in Eqn. (5) which yields
M
∑ α ′ δp
l =1
kl
⎡ δ p1
⎢ δp
⎢ 2
⎢ M
⎢
⎣δ p M
1
= β k or
⎤ ⎡α11(1 + λ )
α12
⎥ ⎢
α 22 (1 + λ )
⎥ = ⎢ α 21
⎥ ⎢
M
M
⎥ ⎢
αM 2
⎦ ⎣ αM1
α1M ⎤
L
α 2 M ⎥⎥
L
⎥
O
M
⎥
L α MM (1 + λ )⎦
For very small value of λ , the displacement vector
−1
⎡ β1
⎢β
⎢ 2
⎢ M
⎢
⎣β M
⎤
⎥
⎥.
⎥
⎥
⎦
(12)
δp k , obtained from Eqn. (12) is close to the
one, obtained by the pure Inverse Hessian technique, Eqn. (9), which is a good step in the final
stages of the iteration, near the minima. If χ 2 = 0 (or very small), then we can get (almost)
quadratic final convergence. However, if λ is very large, then the matrix
α kl′
is forced in to
being diagonally dominant, so Eqn. (12) goes over to be identical to Eqn. (10), this is good if the
current iterate is far from the solution. It means that, by increasing the parameter λ we approach
the ‘steepest descent’ limit (i.e. a short step in the steepest descent direction). Thus, the damping
term λ influences both the direction and the size of the step, and this leads us to make a method
without a specific line search. To reduce the computational errors (especially near the minimum
point), it is recommended to find the derivatives of the model function χ 2 ( X , P ) analytically.
Let’s first prepare the LM algorithm, with flow chart. The minimization process is iterative. One
(
)
2
2
starts with a reasonably small value of λ . At every successful iteration: χ mew < χ cur , it is
reduced by a factor of 10, moving towards the ‘inverse Hessian’ regime. Otherwise it retreats to
the ‘steepest descent’ regime by being increased by a factor of 10. The stop criteria are necessary
to avoid an endless iteration cycle. When one or more combination of the following stopping
criteria are satisfied, then the fitting process stops:
i. When the total number of iterations entered by the user attains.
ii. When the minimum value of χ 2 ( Pc ) to exit iteration attains.
iii. When the absolute shift of the chi square, χ 2 ( Pc + δ P ) − χ 2 ( Pc )
below some a certain
threshold or decreases by negligible amount. The program can also be set to ‘PAUSE’ when
χ 2 ( Pc + δ P ) a start to diverge then continues after press enter key.
102
Figure 2. The LM Algorithm with a flow chart.
The update rule is used as follows. If the error goes down following an update, it implies that our
quadratic assumption on χ 2 is working and reduce λ (usually by a factor of 10) to reduce the
influence of gradient descent. On the other hand, if the error goes up, we would like to follow the
gradient more and so λ is increased by the same factor. If the initial guess is good but χ
2
does
not fall down to the required minimum value, we have to change the initial value of λ slightly.
103
2. IMPLEMENTATION OF THE LM ALGORITHM
In this paper Gauss’s elimination and Gauss’s Jordan matrix inversion methods are used to
determine the shift parameters. Among the several tests made on real and complex non linear
functions, only three examples are illustrated to see how much this method is effective and faster
than the other methods.
2.1. Test on real three dimensional wave function
(xi , yi )
The first test is applied to two dimensional data coordinate
fi
and data value
where i = 1 − 210 , e.g., at i = 7 ( x 7 = −7, y 7 = −1, f 7 = 6.452 ) .
Table 1. Experimental data for irregularly shaped surface.
xi
-7
-6
-5
-4
-3
-2
-1
1
yi
2
3
4
-
-
-
5
6
7
-
-7
-1.029
6.743
13.3
15.99
14.91
12.72
19.56
33.59
14.18
3.027
2.91
2.546
2.389
-8.428
-6
-7.211
0.544
6.464
7.324
2.644
-6.841
-20.33
4.615
5.203
10.67
13.85
11.29
4.334
-3.384
-5
-9.661
-5.837
-3.782
-5.371
-10.97
-22.67
-49.52
40.49
21.73
18.62
17.54
13.99
8.64
3.572
-4
-7.136
-8.94
-11.96
-15.41
-18.64
-26.28
-52.52
54.98
26.54
16.56
12.09
9.181
8.176
8.796
3.058
9.637
5.997
-3
-2
-1
0
1
-1.112
4.853
6.452
1.905
-0.942
-6.983
-0.665
7.76
14.72
11.16
-13.68
-7.916
2.596
12.53
14.99
-17.6
-11.2
-0.96
6.545
13.81
-16.22
-4.794
10.17
21.22
19.46
-15.65
-27.89
3.837
10.87
22.71
41.83
31.66
47.46
21.57
30.86
40.53
16.97
5.508
0.567
0.605
-
-
-
-
5.224
2.158
8.837
10.48
10.26
-4.33
-
-
-
-
-
-
30.86
21.57
19.46
13.81
14.99
11.16
-
-
-
-
-
-
47.46
31.66
21.22
6.545
12.53
14.72
-1.905
-
-
-
41.83
22.71
10.17
0.96
2.596
-7.76
-6.452
-
-
0.942
-
2
-5.997
4.33
10.26
10.48
8.837
2.158
-5.224
10.87
3.837
4.794
11.2
7.916
0.665
-4.853
3
-9.637
-3.058
0.605
-0.567
-5.508
-16.97
-40.53
27.89
15.65
16.22
17.6
13.68
6.983
1.112
4
-8.796
-8.176
-9.181
-12.09
-16.56
-26.54
-54.98
52.52
26.28
18.64
15.41
11.96
8.94
7.136
5
-3.572
-8.64
-13.99
-17.54
-18.62
-21.73
-40.49
49.52
22.67
10.97
5.371
3.782
5.837
9.661
-
-
-
-
6
3.384
-4.334
-11.29
-13.85
-10.67
-5.203
-4.615
20.33
6.841
2.644
7.324
6.464
0.544
-
-
-
-
7
8.428
2.389
-2.546
-2.91
3.027
14.18
33.59
19.56
12.72
14.91
15.99
-13.3
6.743
104
7.211
1.029
Figure 3 (a). Graphs of experimental values f i (blue) and numerical or computed values
f ( X , Pc ) (yellow) before iteration.
Figure 3 (b). Graphs of experimental values f i (blue) and numerical or computed values
f ( X , Pc ) (yellow) after iteration
105
From the above results (Table.1), one can easily see that the data (surface) follows the wave
function having the form
f ( X , P) = f ( x, y, p1 , p2 , p3 ) = p1
sin( p3 x)
cos( y )
− p2
− p3 sin( y ) cos( x) .
2
y +1
x
(
)
We have then made a fitting, using the LM approach, in order to find the values of the
parameters ( p1 , p2, , p3 ) that best fit f ( X i , P) with f i (see Fig. 3 (a) and (b)).
In this case the dimension is q = 2 and the numbers of parameters are M = 3 . After initializing
( p1, p2, , p3 ) the values found from the iteration are χ 2 = 0.0 ,
The function now have the from f ( x, y ) = 7
p1 = 7.0, p2 = 11.0 and p3 = 54.0 .
sin(11x)
cos( y )
− 54
− 11sin( y ) cos( x) .
2
y +1
x
(
)
As one can see from the above results, the LM model is highly useful when it is implemented to
complicated-shaped surfaces. What is also important here is here that selecting an appropriate
type of function (such as sine, power, decay, etc functions) and lambda. The shift parameters are
not that much changed by normalized random errors only minimum of chi-function increases.
Hence, based on the above two figures (Figs. 3 (a) and (b)), one can conclude that new
equations/relations and modifications to the already existing formulas can be obtained from
experimental data having disturbed/complicated surfaces.
2.2. Test made on complex two dimensional function
In ellipsometery the complex ratio ρ =
rp
= tan
r
Ψ e
is measured, commonly
j∆
s
expressed in terms of the two real parameters Ψ and ∆ i.e. ρ = tan Ψe j∆ . The inversion of this
formula to get suitable value of real and imaginary part of the refractive index is some what
difficult to do analytically, and even numerically inversion of complex functions using LM
algorithm is not yet well developed.
Let us consider an oblique reflection and transmission of optical plane wave at the planner
interface between two semi-infinite homogeneous optically isotropic media air and glass with
complex index of refraction n = nr + jk . The ratio of the complex reflection coefficient, ρ , is
related to the angle of incident by
ρ = tan Ψ e
j∆
⎡ sin
= ⎢⎣
2
θ − cos θ 0
sin
2
(n r
+ jk
θ − (n r + jk
106
)
2
)2
− sin
cos θ
2
θ ⎤⎥
⎦
2
.
The algorithm has been tested on an actual data taken in a PSA-ellipsometry on acrylic glass
sample for a wave length of light 450nm . After successive iterations the following results has
been recorded.
Table 2. Experimental data and computed values of ρ and n .
i th
measuremen t
Data
Experimental
Values found from the successive iterations
coordinate
data values
Computed values
Actual error
(θi / deg )
f i = ρi
f ( X i , Pc ) = ρ (θ i , n)
ρi − ρ (θi , n)
1
-0.11726-
-1.1721756E-01-
-4.2445958E-05-
j0.00134
j1.3177083E-03
j2.2291672E-05
-0.06301-
-6.2998131E-02-
-1.1868775E-
j0.00135
j1.3587392E-03
05+j8.7391818E-06
-0.03577-
-3.5782781E-02-
1.2781471E-
j0.00135
j1.3766416E-03
05+j2.6641530E-05
-0.00847-
-8.5111084E-03-
4.1108578E-05-
j0.00143
j1.3926749E-03
j3.7325081E-05
52+j0
2
54+j0
3
55+j0
4
56+j0
N=4
q = 1, m = 1,
Initialization
CHI = 1.1195837E - 09 - j7.0279005E - 10
p1c = n = 3 + j0.3
ABS(CHI) = 1.32188E - 9
n = 1.50009620 + j2.90234271E - 3
The real and imaginary part of the refractive index of the glass found from the iteration is
nr = 1.5000962 and k = 0.00290234 271 respectively. The fitted values of the reflection
coefficient have up to 5 decimal precision (one can also get high precision by selecting
appropriate lambda till the errors arise only form the experiment imperfection and machine error.
The interesting thing doing with complex function is, we only solve the derivative of ρ ( θ , n )
with respect to
n
i.e.
d ρ (θ , n )
dn
to find nr and k (not ρ ( θ
dn
,n)
and
r
ρ (θ , n )
) . During
dk
interpolation and extrapolation, unlike the Aitkens and Lagrange interpolations, graphs
interpolated using LM model follow the right path (with little regression).
107
Extrapolated graph for the complex function ρ(θ,n)
* indicates experimental value
0.4
-0.0010
0.3
0.2
Imaginary of ρ (ρI)
Real of ρ (ρr)
(where n=1.5000962+j0.00290234271)
0.1
0.0
-0.1
-0.0011
-0.0012
-0.0013
-0.2
-0.3
-0.0014
-0.4
-0.0015
-0.5
40
45
50
55
60
Angle of incident θ/deg
65
70
Figure 4. Extrapolated graph for the complex function ρ (θ , n) with * and ▪ representing
experimental and numerical values respectively.
2.3. Test on complex two dimensional power function
The third test was made on complex three dimensional power functions (their derivatives are
logarithmic functions). Consider the following experimental data:
Table 3. Experimental data on 2D power functions.
i
xi
yi
1
6+j2
- 1- j 6
151.1271
j 41.47818
2
5+j8
29+j 0
-318.893
j 710.7169
3
-3+-j0.5
-7+j 1
34.97808
j 96.72046
4
-4+j 2
0+j 5
61.8854
- j 24.1816
5
-5+j 5
-9.9- j 3
260.2891
j 413.5324
6
-6- j 1
-4+j 1
14.13067
j 120.9102
N =6
108
fi
The data is fitted with the function f ( x , y , p1 , p 2 , p 3 , p 4 ) = x p + y p + p 3 xy + p 4 . For this
1
case the value of
p
1
= 2 − 0 .5 j
,
q = 2
2
and M = 4 . During Initialization of the parameters with
p 2 = 2 + 0 .5 j ,
p 3 = 2 + 0 j 0 .5
and
p 4 = 2 + j 0 .5
Pc = 2, 2, 2, − 0 .5,0 .5,0 .5,0 .5 ), the appropriate value of λ
(equivalent
to
used near 0.001 is 0.0012 .
Figure
5 (a). Graphs of the experimental and numerical data at different number of iterations.
Figure 5 (b). Graphs of the experimental and numerical data at different number of iterations.
109
Figure 5 (c). Graphs of the experimental and numerical data at different number of iterations.
Figure 5 (d). Graphs of the experimental and numerical data at different number of iterations.
The function becomes f (x, y, p1 , p2 , p3 , p4 ) = x (1− j 0.6 ) + y 2 + (− 0.05 + j5)xy − 9 + j3 . From the Figs. 5
(a)-(d), we can see that the LM is not affected by the order of the data (ascending or descending).
110
Based on the above results we can conclude that the LM algorithm is popular method and has the
following advantages
(i)
The parameters converge rapidly around the minimum in multi dimensional surfaces with
complicated landscapes.
(ii)
Even though the initial guess is poor, LM fits partly/most of the parameters to make
fresh start.
(iii)
The convergence speed needed to reach the minimum, is not significantly influenced by
the number of parameters.
(iv)
The shift parameters are not that much changed by normalized random errors. Only the
minimum of the chi-function increases.
(v)
Normalized random errors do not bring much change on the convergence speed, etc. Like
any other non-linear optimization techniques, the LM algorithm method in finding global
minimum is not guaranteed (this however can be secured by initializing parameters with
a better guess).
3. SUMMARY
We extended the framework of the LM algorithm to real and complex multi-dimensional
functions. The results show that LM is very efficient when Gradient Descent and Newton’s
methods separately failed to converge. In this paper we developed two programs (one for real
and the other for complex or imaginary values) that work for any number of parameters, any
number of dimensions and coordinate systems: Cartesian, Curvilinear etc. We believe that the
algorithm also provides a concert support when someone wants to make a check at the instant of
a fitting or when solving complex functions. Last but not least the LM method develops user’s
trust on the algorithm during fitting complicated surfaces and/or graphs.
4. ACKNOWLEDGEMENTS
We would like to acknowledge the moral support of all our staff members at the Department of
Physics and the material support of the same department, College of Natural and Computational
Sciences, Mekelle University. We are also grateful for the referees (internal and external) and the
editors of the Momona Ethiopian Journal of Science for their critical and constructive comments
and to the opportunity the Journal has given us to publish our paper in the first volume.
111
5. REFERENCES
Arumugam, M. 2003. EMPRR: A High-dimensional-Based Piecewise Regression Algorithm. pp.
4-16.
Avriel, M. 2003. Nonlinear Programming: Analysis and Methods. Dover Publishing. ISBN 0486-43227-0.
Bates D. M. & Watts, D. G. 1988. Nonlinear Regression and Its Applications. Wiley, New York.
Box, M. J., Davies D. & Swann, W.H. 1969. Non-Linear optimisation Techniques. Oliver &
Boyd.
Coope, I. D. 1993. Circle fitting by linear and nonlinear least squares. Journal of Optimization
Theory and Applications, Plenum Press, New York, 76 (2).
Gill, P. R., Murray W. & Wright, M. H. 1981. The Levenberg-Marquardt Method §4.7.3 in
Practical Optimization. Academic Press, London, pp. 136-137.
Kelley, C. T. 1999. Iterative Methods for Optimization, SIAM Frontiers in Applied Mathematics,
18, ISBN 0-89871-433-8.
Lampton. M. 1997. Damping-Undamping, Strategies for the Levenberg-Marquardt Nonlinear
Least-Squares Method. Computers in Physics Journal, 11(1): 110 – 115.
Lawson C.L. & Hanson, R.J. 1974. Solving Least Squares Problems. Prentice-Hall.
Levenberg, K. 1944. A Method for the Solution of Certain Non-Linear Problems in Least
Squares. The Quarterly of Applied Mathematics, 2: 164–168.
Lourakis, I. A. 2005. A Brief Description of the Levenberg-Marquardt Algorithm Implemented
by levmar, Institute of Computer Science Foundation for Research and Technology Hellas (FORTH) Vassilika Vouton.
Madsen, K., Nielsen H.B. & Tingleff, O. 2004. Methods for Non-linear Least Squares Problems,
Informatics and Mathematical Modeling. 2nd Ed., Technical University of Denmark.
Marquardt, D. 1963. An Algorithm for Least-Squares Estimation of Nonlinear Parameters.
SIAM Journal on Applied Mathematics, 11: 431–441.
112
Momona Ethiopian Journal of Science (MEJS)
Guidelines to contributors
General
i. Momona Ethiopian Journal of Sciences (Mom. Eth. J. Sci.) publishes high quality solicited and
unsolicited papers in various science disciplines. The manuscripts are subject to peer review.
Unpublished papers of scientific value and which fulfill the standards of the editorial aims of
Momona Ethiopian Journal of Science will be readily accepted. Acceptance for publication
depends on recommendations by at least two of the editors. The editor-in-chief reserves the
right to reject any manuscript submitted, whether on invitation or otherwise, and to make
suggestions and modifications before peer review.
ii) It also publishes review articles, short communications and book reviews, comments: very
short papers (maximum 3 pages) containing opinions, contributions to current scientific
debates or responses to the published articles.
Copyright policy
The Momona Ethiopian Journal of Science accepts manuscripts for consideration with the
understanding that the manuscript has not been published and is not under consideration for
publication elsewhere. By submitting their work authors agree to assign copyright ownership to
the Editorial Office of MEJS in the event that the work is published in the Momona Ethiopian
Journal of Science.
Original articles will be considered for publication if they have been published previously as
short abstracts, but not if they have been published previously as extended abstract (>1000
words). This applies to both electronic and paper publications. Submitted manuscripts must be
accompanied by a copyright release statement signed by all authors. A copy of release form can
be obtained from the Editorial Office of MEJS. Authors may use multiple copies of the form,
each with an individual author’s signature are also acceptable.
Authorship policy
The journal has a policy on authorship for manuscripts submitted to Momona Ethiopian Journal
of Science. The policy is that authorship credit should be based only on
113
ƒ
Substantial contributions to conception and design, or acquisition of data, or analysis and
interpretation of data
ƒ
Drafting the article or revising it critically for important intellectual content and
ƒ
final approval of the version to published
The above conditions must, all be met. Acquisition of funding, the collection of data, or general
supervision of the research group, by themselves, do not justify authorship. Each author of all
multiple –authored manuscripts published in the journal of the Momina Ethiopian Journal of
Science will be required to confirm in writing, prior to publication of the manuscript, that he or
she satisfies the journal criteria for authorship. Authors have to submit statement of authorship to
the Editorial Office of MEJS.
114
Instructions to Authors
I. Nature of the manuscript
1. The manuscript should be written in English.
2. Should be as concise as possible, less than 21 printed pages including the abstract. Longer
manuscripts can only be accepted with exceptional cases. Shorter submissions take priority when
space is limited.
3.
Statement
of
authorship
should
be
submitted
with
the
manuscript.
Momona Ethiopian Journal of Science requires a signed confirmation by all Authors and should
state that:
i)
the work as submitted has not been published or accepted for publication, nor is being
considered for publication elsewhere, either in whole or substantial part;
ii)
the work conforms to the legal requirements of the country in which it was carried
out, including those relating to conservation and welfare;
iii)
all authors and relevant institutions have read the submitted version of the manuscript
and approve its submission; and
iv)
all persons entitled to authorship have been so included.
II. Submission of Manuscript
All manuscripts should be submitted in both hard copies (three) with signature of authorship
confirmation to Momona Ethiopian Journal of Sciences, Mekelle University, P.O.Box 231,
Mekelle, Ethiopia and Electronic submission: all manuscripts should be submitted electronically
to the email address [email protected]. The electronic submissions should be in Rich Text Format
(RTF). The manuscript must be submitted ready for the printer, i.e. typewritten, double-spaced
on one side of the paper with margins of 1 inch in all sides, on format A4. Lines should be
numbered.
Text marking: Italics, should be used for all scientific names from the level of the genus
downward.
115
III. Standard papers
Standard papers should be divided into abstract, introduction, methods, results, discussion,
acknowledgments and reference sections.
Momona Ethiopian Journal of Science publishes original contributions containing original
scientific findings in any of the science fields. A single file should be prepared containing the
title page, abstract, main body (text), acknowledgements (if any), references, tables, figures and
appendices (if any) in Microsoft Word or LaTex. Portable Document Format PDF) or Postscript
(PC) or Encapsulated Postscript (EPS) files can be submitted along with the LaTex files but are
not eligible for submission only by their own.
1. Title page
This should include the full title, the name(s) of author(s), addresses: institute or laboratory of
origin, name, postal address and email address of the author(s) and not more than six keywords,
which should be relevant for literature searching.
Name, address as well as fax and phone number, e-mail-address of the author responsible for
correspondence.
2. Abstract
All papers should include a summary, in short numbered paragraphs, limited to about 3% of the
length of the text, and in any case to not more than 300 words. This should provide a concise
statement of the scope of the work and its principal findings and be fully intelligible without
reference to the main text (it must contain background, objective, Methods, Results and
conclusion).
3. Introduction
This should contain a clear statement of the reason for doing the work, outlining essential
background information but should not include either the results or conclusions.
4. Materials and Methods
This should be concise but provide sufficient details to allow the work to be repeated. Product
and manufacturer names: Where specific named materials/products are mentioned or named
116
equipment used (including software packages), these should be identified by their manufacturer,
followed by the manufacturer's location (e.g. town, state, country), or a source reference should
be given if a standard or replicated procedure is being followed or uses.
5. Results
This should not include material appropriate to the Discussion. Pertinent results should be
reported using text, table and figures. Restrict tables and figures to those that are necessary. Do
not duplicate data in tables and figures and, where possible, avoid repeating data presented in
tables or figures in the text. All tables and figures should be referred to in the text. In general the
emphasis should be on the size of effects rather than their statistical significance.
6. Discussion
This should highlight the significance of the results and place them in the context of other work.
A combined results and discussion section are allowed.
New and important findings should be emphasized but, as a rule, data previously presented in the
Method and Results sections should not be repeated. Implications and limitations of the findings
and their practical application should be discussed. The length of the Discussion should be
commensurate with the number of important findings. Usually the discussion can be less than
750 words in length. It is generally not necessary to restate the conclusion at the end of the
Discussion.
7. Acknowledgments
8. References
All references in the text should be listed in the reference section. References should not appear
in the reference list if they are not quoted in the text. References should be listed in alphabetical
order and then further sorted chronologically if necessary. The following usage should be
confirmed.
a) Reference to a journal publication:
117
Mergeay, J., Declerck, S., Verschuren, D. & De Meester, L. 2006. Daphnia community analysis in
shallow Kenyan lakes and ponds using dormant eggs in surface sediments. Freshwater
Biology, 51: 399-411.
b) Reference to a book:
Moss, B. 1998. Ecology of fresh waters, man and medium. Blackwell Scientific Publications,
London, pp. 557.
c) Reference to a chapter in an edited book or articles of monographs (symposia, etc):
Tudorancea, C., Zinabu Gebremariam & Dadebo, E. 1999. Limnology in Ethiopia. In: Wetzel, R.
G. & Gopal, B. (eds.), Limnology in Developing Countries 2: 63-118.
In the text, references should be cited chronologically, author’s surname should be followed by
year of publication: (Finlay et al., 1988; Koschel, 1995; Koschel and Adams, 2003;).
If several contributions by the same author were published in one year, a distinction of these
should be made by adding a, b, or c, e.g. (Koschel, 1981a; Meyer and Huss, 1999a, b).
Ethiopian names should be in direct order, that is, the full first name followed by the father’s
name; Gebremariam Hailu et al. 2001, but not G.mariam, H. et al. 2001.
Titles of journals should not be abbreviated. Unpublished material, except for PhD theses, should
not be included among the references, but should be cited as (Kebede Tesema et al., 2005
unpubl. Data) in the text.
9. Tables
Each table should be typed on a separate page. Tables should be numbered consecutively in
Arabic numerals and attached to the end of the text after the references. They should be supplied
with headings and kept as simple as possible and should be referred to in the text as “Table 1”,
etc.
10. Figures
Figures should be numbered consecutively in Arabic numerals with proper legends. Legend for
figures (incl. photographs) should be listed consecutively on a separate page after tables.
118
Other columns
1. Review article
These are aimed at giving an overview of particular subject suitable for a wider audience that
includes recent advances in an area in which an author has been actively engaged.
2. Book reviews
A critical evaluation of recently published books in any discipline of sciences can be published.
3. Short communication
It must report completed work, not preliminary findings. They are an alternative formats for
describing smaller piece of work. The report should not be more than five pages and should not
contain more than two figures, tables or combinations.
© College of Natural and Computational Sciences, Mekelle University
119