Proceeding for Conference(Bahir Dar University).pdf

Mr. Tesema Belai (Asst prof.)Mr. Michael Mahari

EDITORS

ASSOCIATE EDITORS

EXECUTIVE COMMITTEE EDI-

Mr. Seleshi DemesseiMr. Ephrem Dagne

Mr. Mulugeta Azeze (Chairman)Dr. Pattem Anjaneyulu (Secretary)Prof. M. Jayachandran (Member)Dr. Nigus Gabbiye (Member)Mr. Yeneneh Tamirat (Member)Mr. Bereket Haile (Member)

contents

ELECTRICAL TECHNOLOGY AND ICT

FOOD PROCESSING

Charge Carriers Relaxation Dynamics on Chromium Doped Photorefractive Bismuth Sillenite Crystals

10

Stand Alone Hybrid Power System Design for Rural Electrification in Ethiopia

25

Wireless Backhaul and Standby Microwave Link between Bahir Dar and Woretta

35

Improved Design of Modulated Filter Banks and Performance Comparison of Filter banks with OFDM in

Multicarrier Communications

20

Wireless Transceiver Testbed for Education and Research Using SDR at BDU

42

Automatic Potometer for Creating Osmotic Stress in Plants Comparable to Field Moisture Stress

46

Bioconversion Of Parthenium hysterophorus into Valuable Organic Manure

54

Getasew Admasu Wubetu, E.Goovaerts

Zelalem Girma, Martin Braun

Gashaw Mihretu, M.Jayachandran, Agizew

Pattem Anjaneyulu

Solomon H Gebreyohannes

Sanjay Singh

Hiranmai Yadav, R, Eyasu Mekonnen, Vijayakumari, B

Economic Efficiency of Export–Oriented Cattle-Fattening Farms in East Shewa Zone: The Case of Adama City and Its

Surroundings, Ethiopia

59

Ethiopian Indigenous Knowledge of Traditional Foods and Beverages Processing: Needs for Modern Food Processing

Technology and Transformations

66

A Review on Recent Developments In Extended Finite Element Method (X-FEM)

78

Design of Molding Die for Shoe Sole 85

Dynamic formulation of 3 PRS PKM based on screw theory and Newton-Euler’s approach

92

Ergonomics Aspect of a Vehicle: The Case of Bajaj 102

Feasibility Study of Pumped Storage System for Application in Amhara Region, Ethiopia 107

Oumer Berisso

Tariku Hunduma

Migbar Assefa

Mesfin Seid Ibrahim

Hassen Nigatu

Mengist Hailemariam, Awol Mohammed

Mastewal Alemu Tilahun

MECHANICAL MANUFACTURING AND PROCESSING

Improvement in Productivity:A Case Study of Cement Production Enterprise

113

Abera Endesha Bekele, Ajit Pal Singh

contents

Modeling and Simulation of Biodiesel Production from Oil and Leather Industrial Wastes Using Aspen Plus 2006

150

Material Minimization of Straight-tooth Bevel Gears Using Particle Swarm Optimization (PSO)

119

Simulation and optimization of pump case casting 123

Work-Process-Oriented Approach of Improving Labor Productivity in Ethiopian Leather Product Manufacturing

129

Assessing Suitability of Recycled Aggregate for Use in Concrete

134

Biogas Generation from Food wastes from Students’ Cafeteria of Bahir Dar University

138

Biogas Production using Anaerobic Co-digestion of Sanitary Wastewater and Kitchen Organic Solid Waste of

Condominium House

143

Eshetu Getahuna, Menilk gebeyehub, Fitfiteb, Dr. Nigus Gabbiyeb

Yonas Mitiku, R. Srinivasa Moorthy

Abubeker Ahmed, Mekonen Gebreslassie

Sisay Geremew Gebeyehu

Woubishet Zewdu Taffese

Libsu, Chavan, R., Wonde

Martha Minale, Eng Teshome Worku

TRANSPORT AND MATERIAL

SUSTAINABLE ENERGY

contents

Calibration and Validation of SWAT Model and Estimation of Water Balance Components of Shaya River Watershed,

Genale-Dawa Basin, South-Eastern Ethiopia

207

Study of Solar Cooling Alternatives for Residential Houses in Bahir Dar City

157

Analyzing the Combined Impact of Climate and Land Use Changes on Sediment Yield and Stream Flow in the Upper

Gilgel Abbay Catchment, Ethiopia

178

Application of SWAT Model for Assessment of Best Management Practices (Bmps) On Soil Erosion/

Sedimentation in Gilgel Gibe Basin Watershed-Ethiopia

191

Assessing Reliabilities of Multipurpose Reservoir Using Stochastic Modeling and Scenario Based Simulations of Net

Inflows, the Case of Lake Tana, Ethiopia

201

A Simple Temperature Method for the Estimation of Evapotranspiration

167

Alemayehu Abate, Tena Alamirew, Megersa Olumana

Meron Mulatu Mengistu

Kirubel Mekonnen, Gebreyesus Brhane Tesfahunegn, Kassa Tadele

T.A.Demissie F., Saathoff, A.Gebissa, Y.Sileshi

Mulugeta Azeze, Hartmunt Eckstädt, Yilma Seleshi

Temesgen E.Nigussie, Assefa M., Melesse

WATER RESOURCES AND ENVIRONMENT

contents

Changes in Sediment Transport and Channel Morphology in a Micro-Scale Experimental Braided River

215

Intensity-Duration-Frequency (IDF) Analysis of Point Rainfall for Selected Meteorological Stations in Illu Aba

Borra Zone, Oromia Region

227

Microbiological Quality of Drinking Water and Hygiene sanitation Knowledge of the People in Adama Town of

Oromia Region, Ethiopia

237

Monitoring stations for runoff and sediment transport along Gumara River (Tana Basin): procedures and preliminary

results

241

Numerical Groundwater Flow Modeling Of the Meki River Catchment, Central Ethiopia

251

Simulation of Lake Tana Reservoir under Climate Change and Development Scenario

258

Trend Analysis of Runoff and Sediment fluxes using Statistical and Physically based models: Upper Blue Nile

Basin

267

Michael M.M

Elias Jemal, Desalegn Chemeda

Dr. Sukanta Bandyopadhyay

Mekete Dessie, Teshager Admasu, , Valentijn Pauwels, Niko Verhoest, Enyew Adgo, Jean Poesen, Ruben Maes, Jan Nyssen

Dereje Birhanu, Tenalem Ayenew

Getachew Tegegne, Dereje Hailu

T.Gebretsadkan, Y.A. Mohamed, G.D. Betrie, P. Van der Zaag, E.Teferi

contents


10

NCSTI-2012


Getasew Admasu Wubetu1,2 and E.Goovaerts2

1Bahir Dar University, College of science, Physics Department, P.O.Box: 79 Bahir Dar, Ethiopia 2 Experimental Condensed matter Physics, Department of Physics, University of Antwerp,

Univesiteitsplein 1, B-2610 Antwerpen, Belgium e-mail: [email protected]

Abstract: Fast and slow tunneling between the localized state of charge carriers in chromium doped bismuth sillenite B12SiO20 (BSO:Cr) was studied by photo-induced absorption (PIA) technique using frequency doubled nanosecond pulse excitation of ND:YAG laser. Using fitting procedure, the Cr doped BSO PIA decay is very well fitted by stretched exponential decay with highly correlated time constant τ and stretching coefficient α. To get the best fit standard statistical mean α were determined from the pump intensity measurements, leading to α = 0.22. From the linear fit of the Arrhenius plot, the thermal activation energy (Ea) is Ea = 0.016 ± 0.03 eV. This activation energy has lower limit of Ea ≈ 0 and upper limit of Ea ≈ 0.046 eV. It corresponds to a very weak temperature dependence of the PIA decay. This result is consistent with carriers tunneling between localized states for carriers relaxation. From the optical absorption spectra measurements, the Cr doped BSO has shifted the intrinsic edge to the longer wavelength and gave rise to new absorption bands in the visible and near infrared region (NIR) ranges.

Keywords : Neodymium-doped yttrium aluminum garnet, Cromium doped Bismuth Sillenite crystal, Photorefractive effect, optical density

I. INTRODUCTION

The photorefractive effect was first found in 1966 on optically-induced inhomogeneous crystal of LiNbO3, LiTaO3 at the Bell laboratories in New Jersey [1]. The effect was observed and other ferroelectric materials by measuring their respective refractive index. The study also was further extended by different groups in the crystals family of the bismuth sillenites, Bi12MO20 (BMO with M=Si, Ge, Ti) which are promising in data storage handling, real time holography, optoelectronics, and optical communications [2, 3, 4]. The non-doped and doped bismuth sillenite Bi12SiO20 (BSO) crystals studied here were grown using the Czochralski technique from stoichiometric melts [5]. Now a days, doped BSO has considerable scientific and technological interests including doping in a number of the transition series ions such as Cr, Mn, Fe,Co, Ru. The introduction of these extrinsic defects offers the possibility of modification of the photorefractive property of these crystals. The performance of such photorefractive properties in these crystals is strongly dependent on the impurities. The impurities act as donor or acceptor for mobile charge centers [6]. To get such microscopic information, a number of techniques could be useful like optical studies, electric

paramagnetic resonance (EPR), etc... Here, the study, however, focuses on the photo-induced absorption (PIA) which uses a pump-probe technique to monitor the relaxation dynamics for the excited charge carriers. These charge carriers are excited from localized states (photoactive centers) in the forbidden band gap to extended states in conduction (valance) bands [3]. In the conduction (valence) charge carriers move, are re-trapped and excited again.

When trapping occur at different center than the original one, a photo-induced absorption can be observed due to the presence of the shallow center other than the deep trap levels. The technological importance of the Cr doped BSO was investigated by different researchers. The doping of the sillenite crystals with Cr reduces the photoconductivity and increases the photoinduced absorption in the visible and near IR region [7]. This can be useful for an interesting application in improving the diffraction efficiency and increasing the speed of the materials at the wavelength of 633 nm [8]. Egorysheva et al and Mookrushina et al [9, 10], have shown that small amounts of Cr doped in BSO significantly change the light induced changes photoconductivity and the photorefractive properties in the crystals. This is coming from the modified ratio between excited donors and deep traps in the material which modifies the photorefractive properties. The article


11

NCSTI-2012

on reference [11], also described that the Cr dopants in sillenite at concentration levels of the order of 10-3 wt% markedly modify the properties of the photo-induced carriers and the photoconductivity. The aim of this work is to experimental investigate the relaxation of PIA after a nanosecond. Laser pulse on non-doped BSO and Cr doped BSO crystals. Characterization of carriers dynamics is done by pulsed laser excitation using a frequency doubled Nd:YAG laser. The laser beam hits the sample and excites the charge carriers. In order to monitor this process either a CW halogen lamp (long time range) or pulsed Xenon (Xe) lamp (short time range) is used to measure the change in optical density. Optical absorption spectra and time-resolved optical absorption studies are done on different time scales and probe wavelength for these samples. The study is also extended to include the dependence on laser pulse intensity and temperature. These yield information on the charge transfer processes which are at the origin of the photorefractive effect.

II. EXPERIMENTAL

2.1 SAMPLE PREPARATION The samples used in this work were non doped BSO and Cr doped BSO. These crystals were prepared in the crystal growth laboratory of the Bulgarian Academic of Science in Sofia. The bismuth sillenite single crystals were prepared by the chemical reaction between bismuth oxide (Bi2O3) and silicon oxide (SiO3) in the molar ratio of 6:1. They were grown by using Czochralski method in the [001] crystal direction with a low temperature gradient (5 – 70 C/cm) over the solution. The platinum crucibles that contain the appropriate melt solution were allowed to rotate at 20 rpm. These parameters allow achieving a growth rate of 0.7 mm/h. The same growing conditions were employed for all the samples to achieve an ingots size of height 60 mm and diameter 40 mm with a high degree of homogeneity. As shown in the table 3.1 the extrinsic carrier concentration of the sample were determined whereas the thickness of the polished crystal plates was measured using a digital micrometer. Table 1: Properties of polished crystal plates of non-doped and Cr doped BSO: Thickness using digital micrometer and extrinsic carrier concentrations from atomic absorption spectrometry. Type of samples Thickness

[mm] extrinsic carrier concentration [cm¡3]

BSO 0.91 no extrinsic carrier concentration

Cr:BSO 0.59 5.1x1018

2.2 EXPERIMENTAL SETUP The steps carried out in the experimental setup were the followings. First the frequency doubled ND:YAG laser was adjusted at a frequency of 10 Hz. Then, it was applied to the pump the flash lamp and Q-switch which were activated to obtain the optimal pulsed laser operation. Later, the laser intensity was adjusted using the polarizer of pulse energies in the range EP = 8 mJ/cm2 to 79.6 mJ/cm2 corresponding to peak intensity per pulse of IP = 1.33 MW/cm2 to 13.3 MW/cm2. This procedure allowed to acquiring a better signal and limits the damage on the samples. The frequency doubled Q-switched Nd:YAG laser at 532 nm was allowed into sample area that should overlap the Halogen/Xe probing light at an angle around 50 shifted in a quasi-collinear geometry. Activating the LP900 software for induced absorption to check overlap and alignment of the Laser and probe light in the sample with quasi-collinear mode. Then optimal output signal level for low band width (LBW) is around 1 V, it is 0.5 V for High band width (HBW) mode. This is adjusted using the irises. We were used fast measurements from (ns to 10 ms range) the Photomultiplier Tube (PMT) bias voltage was set with HBW preamplifier at setting 1 using the Xe pulsed probe light. In the case of slow time measurements (100 ¹s to 100 s) the halogen probe lamp was used with PMT bias voltage set at 9 or 10 V with LBW mainly at position 3 or 2 or 4 CW mode. Finally, the time interval between successive laser pulses was set to 60 s and the number of average per measurement to improve the signal to noise ratio was 10. Initially, the oscilloscope was triggered by a signal from photodiode. The pulse power was controlled with an adjustable beam attenuator consisting of rotatable half wave plate and two polarizers suitable for high laser. Suitable high wavelength pass filters and interference filters for the respective wavelengths of the probe light, together with the black paper screen were used to reduce the probe intensity light coming from Xe/halogen lamp in order to avoid limiting sample degradation and stopping the laser light. For the temperature dependency measurements, the samples were fixed in an optical window cryostat. A Pt thermocouple was mounted above the sample with estimated accuracy for the measured temperature of ± 1 K. For the BSO sample the measurements were in the temperatures range of 296 - 370 K where as for BSO:Cr in the range of 296 - 354 K. The resistance is correlated to temperature using a calibrated table.


12

NCSTI-2012

Fig. 1. The schematic scheme of the photo-induced absorption technique. The optical absorption spectra of the BSO and Cr doped BSO crystals were measured using a commercial Cary 5 in the visible and NIR range at room temperature. The instrument has two detectors. These are PMT detector for the UV-Vis in the wavelength range (175 - 800 nm) and the PbS detector in the NIR (800 - 3300 nm).

III. THEORY

3.1 Photorefractive effect on sillenite crystals Light from an ordinary source does not significantly alter the optical properties of the medium in which it travels. In such cases, the medium is said to be optically linear and its index of refraction and absorption do not depend on the light intensity. On the other hand, after the development of lasers in 1960s, which is the beginning of the field of nonlinear optics (NLO), it became possible to obtain highly coherent, monochromatic and intense light beam [12]. The photorefractive effect is a nonlinear optical effect seen in certain crystals and other materials that respond to light by altering their refractive index. Microscopically Photorefractive effect is resulting from the optical redistribution of the charge carriers (electrons and holes). The bismuth sillenite Bi12SiO20 (BSO) crystals have remarkable properties such as piezoelectricity, electro-optical active, and photoconductivity [13]. The contributions to the formation of the electric field in the stated application are coming from photoionization, carrier diffusion/drift and carrier trapping. The main applications of photorefractive materials like BSO, which is an attributed to its crystalline structure, are for storage of temporary and erasable holograms, coherent light amplification, optical information processing, optical interconnection and communication, and etc [14]. Most literatures reported that BSO has been grown by Czochralski techniques from stoichiometric melt of bismuth oxide and silicon oxide, with molar ratio Bi2O3 : SiO2 = 6:1 [2]. It has

a body centered crystalline structure in space group I23. The primitive cell consist of 24 atom of Bi, 2 atom of Si and 40 atom of O. Figure 2 shows the SiO4 tetrahedron in which one Si is surrounded by the four O atoms, as well as one type of the oxygen sites. The Bi atom are situated in a lower symmetry sites and omitted in figure 2.1 for simplicity. At room temperature the band-gap of non-doped BSO crystal is about 3.2 eV. It has a yellow color that comes from the broad absorption shoulder below the actual band edge attributed to Bi antisite defect [13]. The performance of BSO photorefractive materials could be improved by adding transition metal series ions. These might help to optimize the material characteristics like photorefractive sensitivity, speed of the charge transport, life time of the carrier trapping, etc.

Fig. 2. Sillenite Bi12MO20 (M=Si) with the space group of I23. The low symmetry Bi-O complex are omitted and only SiO4 tetrahedral positions in the elementary cell [15].

3.2 Band transport models in sillenite crystals Band transport models in non-doped BSO crystal have been reported by different researchers. Since photorefractive properties are related to carrier transport properties, it is necessary to understand the band and defect structure of non-doped BSO. In most literatures the proposed widely used band diagram of BSO has shown in figure 2.2 at room temperature. Figure 3 shows the electronic band structure of the non-doped BSO crystals with a band gap of 3.25 eV at room temperature [4]. Buse [1] described that this crystals have 1019 - 1020 cm-3 of Si or Bi vacancies and thus one intrinsically heavily defective which plays significant role in the photorefractive effect. The deep trap levels at about 2.5 eV from the conduction band corresponds to the absorption shoulder that leads to the pale yellow color of these crystals which is coming from the BiSi antisite defect. Ramz et al [3] showed that Bi+3 on the Si site can be coupled with the h+ in the surrounding oxygen tetrahedral. Many electron traps are found in the range 0.1 to 1 eV. Benjiloum et al [19], characterized the so called shallow trap levels by photoinduced transient current spectroscopy. On the other hand the deep trap levels are in the range


13

NCSTI-2012

of 1.2 to 2.5 eV from the conduction band for the electrons. The energy level of 2.8 eV below conduction band is generally believed to be inactive compensative acceptor for donor and does not participate in carrier generation and trapping for this crystal [8]. Introducing dopants of the transition ion series in the BSO crystals will create electron donors and traps in the band gap to change the mobility of the charge carriers, modifying the electrical and optical properties dramatically.

Fig. 3. Schematic energy-band diagram of the BSO crystal at room temperature. There are different simplified models for photoinduced concentration of charges carriers and their application described in articles [3, 8, 16]. L1, L2 and L3 are photoinduced absorption center shown in figure 4 for different energy trap levels in BSO crystal. Applying a uniform light these centers leads to a movement that carriers from localized states (photoactive centers) in the forbidden band gap to extended states (conduction or valence bands) where they move, re-trapped and excited again. The positions of the forbidden band include the shallow traps and deep traps levels which have different physical reasons. L1 is the position of the deep trap levels which is around 2.5 eV from the conduction band consists of deep-level donor with plenty of electrons. This position of the electron excitation from the Bi+3/+4 Si to the conduction band. These are associated with the absorption shoulders which depend on electrons. Here the electrons ionized from the D0/+1 level are trapped at BiSi

3+/4+ level. L2 consists of shallow-level traps with the respect of the conduction edge which is around 0.3 to 1 eV. L3 consist of traps in a much deeper than L3. The excitation of the holes at this position from the acceptor position to the valence band have also similar effect on the absorption of electrons excitation from the valence band (VB) to the deep trap level BiSi

+4/+5 or alternatively the hole excitation BiSi

+4/+5 to the VB.

Fig. 4. The defects in BSO crystal [15].

3.3 Kinetics of Photoinduced Absorption The PIA spectroscopy can be studied by using nanosecond laser flash photolysis. The usual study in this techniques were done by propagating the ND:YAG laser pump beam at a small angle (50) away from the probe beam which result the change in optical density. The probe lamp can be either using the halogen lamp (slow time in the range millisecond to second) or (fast time in the range nanosecond to microsecond). The schematic diagram of the pump-probe technique is shown in figure 5. The change in optical density by nanosecond pump laser is determined by comparing the incident probe beam I0 (λprobe) and transmitted intensity through the sample, I(t, λprobe). Mathematically, the change in the optical density is related to the change on the intensity of the incident probe to the transmitted light beam expressed by the Lambert-Beer transmittance equation is given by

] [1

Fig. 5. Schematic diagram of the pump-probe technique. The kinetics of time-resolved optical decay is caused by a back reaction of the photolytically generated species of the initial transient. This back-reaction is thermal induced thus, its rate depend on the temperature. This dependence of the chemical reaction on temperature can be calculate by the Arrhenius equation:

),(),0(

log),(log),(probe

probeprobeprobe tI

ItTtOD


14

NCSTI-2012

[2]

Cr doped BSO significantly shift the band edge to the longer wavelength and gives also two absorption

bands in the visible and NIR ranges. The first absorption band for this crysta l is due to the Cr4+ defect as studied by Electric Paramagnetic Resonance (EPR) measurements in our group described in article [11]. There is also a pronounced photochromic effect observed due to Cr defects in BSO shifted to higher wavelength than in non-doped BSO.

τ is the inverse decay time constant, Ea is the activation energy, A is the frequency factor, kB the Boltzmann constant and T is the absolute temperature. Modifying equation 2 yields

[3]

IV. EXPERIMENTAL RESULTS AND DISCUSSION

4.1 Optical Absorption Spectra of Non-Doped And Doped BSO The spectral absorption measurements were done by Carry 5 Uv-Vis spectrometer on BSO and BSO:Cr at room tempreture. The spectral absorption shoulder of the non-doped BSO falls in the range of 494nm (2.5 ev) to 518 nm (2.4 eV) (green range). According to Oberschmid [23] the absorption shoulder is due to the contribution of an intrinsic antisite defect (Bi3++h) formed by occupation of the tetrahedrally coordinated Si4+ site by Bi3+, coupled with hole, mainly at the oxygen neighbor. The result indicates that many intrinsic defects are involved in the absorption. Fei-Fei et al [16] described that this broad absorption shoulder in the range of 2.3 - 3.2 eV is responsible for the yellow color to the crystal. The result we found is in this energy range. For the sample Cr doped BSO, the optical absorption spectra shows a band edge shift relative to non-doped BSO to 638 nm. This shift is coming from the Cr related defects. There is a low absorption band in the wavelength range of 612 - 656 nm. On the other hand, there are two high absorption bands formed in the wavelength range of 664 - 877 nm (visible) and 946 – 1044 nm (NIR). These absorption bands occur due to defects with different Cr valence state (Cr2+, Cr3+, Cr4+). Nechitailov et al [18] reported the Cr doped BSO formed two spectral absorption in BSO:Cr crystal which shift in absorption edge for to higher wavelength and the appearance of the new absorption bands in the range of 600 - 1060 nm. As it can be seen in figure 4.1, this report is consistent with our experimental findings. The two absorption bands are coming from dynamic charge transfer number of Cr ions (from Cr3+ to Cr2+ to Cr4+ )(or both) [19]. This shows the

300 400 500 600 700 800 900 1000 1100 1200 13000.0

0.5

1.0

1.5

2.0

2.5

3.0

3.5

Abs

orba

nce

wavelength [nm]

non-doped BSO Doped BSO:Cr

BSO:CrBSO

band edge shifted

shoulder

big band due to Cr4+

another band

TkE

B

a

Ae

1

[2]

ATk

E

B

a ln11ln

Fig. 6. Optical absorption spectra versus wavelength of non-doped BSO,Cr doped BSO and P doped BSO. 4.2 Time-resolved PIA decay in BSO and BSO:Cr Time resolved PIA measurements were performed on BSO and BSO:Cr crystals at room temperature. The laser pulse(¸λpump = 532 nm) was exciting at pump intensity of Ipump = 5 MW/cm2 for the crystals. The light induced absorption by the sample rises quasi instantaneously after the illumination by the short laser pulse, and returns to the original state when the illumination is over. The time dependence of the PIA allow to monitor and analyze the response for BSO:Cr crystal ( time range greater than 100 s) in the reversible photorefractive and photochromic effects. As a comparison, figure 7 the normalized optical density spectra decay in the illumination time of 400 ms for the BSO and BSO:Cr were plotted after exposure of nanosecond pulses of 5 MW/cm2 peak on the same condition . In this plot the PIA decay in non-doped BSO is completely finished in a few order of hundred ms whereas in the BSO:Cr this takes much longer time. This shows that a small amount of the Cr defects introduced into BSO significantly changes the photoinduced absorption and photorefractive effects coming from the different photoactive centers and traps levels.


15

NCSTI-2012

Fig. 7. The normalized time-resolved photoinduced absorption decay of non-doped BSO and Cr doped BSO. The pump wavelength 532 nm with peak intensity at 5 MW/cm2and probe wavelength of 600nm were used in 400 ms time range. 4.3 Curve Fits to the PIA decay in BSO and BSO:Cr From our measurements of PIA decay of carriers on BSO, the curves are well fitted using double exponential decay with two time constant the fast time constant and slow time constant. These time constant indicate the place where the carriers in the forbidden band which can be in the shallow trap level or deep traps level. The equation is given by [4] [4] where ∆OD(t = to) is the optical density before the laser excitation, which sets zero for ΔOD, initial amplitude at t= 0 which sets as 0, the ∆ODf (t = t0) and ∆ODs(t = t0) denote amplitudes of the fast and slow decay, τf and τs are the time constant respectively. The non-exponential optical densities decay spectra in non-doped BSO crystal could be well described by equation 4. Figure 8(a) shows the curve fit on the ∆OD spectral decay signal on BSO crystal with measuring time range of the 400 ms at room temperature. The same physical parameters were used as the previous section. This fit for non-doped BSO yields τf = 50 ms and τs = 1800 ms. Marinova et al [8] described using the same system, the position of the two shallow centers in 18 ms (for fast time) and 135 ms (for the slow time) constants. This difference could come from the growth condition of the BSO, or from the difference in the pump intensity in reference [14] was 12 MW/cm2 where as in our experiment 5 MW/cm2. A basic problem of the flash photolysis setup was low frequency instability of the baseline of ∆OD, mainly disturbing the long time range measurement. An improvement of probe light and

/or detector stability would be required to improve this activity.

0 50 100 150 200 250 300 350 4000.0

0.2

0.4

0.6

0.8

1.0

O

D

time [ms ]

BSO-Cr probe= 600 nm BSO probe= 600 nm

BSO-Cr

BSO 0 5 0 1 0 0 1 5 0 2 0 0 2 5 0 3 0 0 3 5 0 4 0 00 . 0 0 0

0 . 0 0 2

0 . 0 0 4

0 . 0 0 6

0 . 0 0 8

0 . 0 1 0

0 . 0 1 2

0 . 0 1 4

0 . 0 1 6

0 . 0 1 8

0 . 0 2 0

t h e d a t a D o u b le e x p o n e t ia l c u r v e f i t

t im e [ m s ]

OD

f = 5 0 m s

s = 1 8 0 0 m s

( a )

0 1 0 2 0 3 0 4 0 5 0 6 0 7 0 8 0 9 0 1 0 00 .0 0 0

0 .0 0 2

0 .0 0 4

0 .0 0 6

0 .0 0 8

0 .0 1 0

0 .0 1 2

0 .0 1 4

d a ta p o in ts a t 1 0 0 sd a ta p o in ts a t 1 s

s t r e c h e d e x p o n e t ia l c u r v e f i t fo r 1 s s t r e c h e d e x p o n e n t ia l c u r v e f i t fo r 1 0 0 s

O

D

t im e [s ]

s

( b )

Fig. 8. The double exponential curve fit for the photoinduced absorption decay non-doped BSO room temperature. The pump light ¸ = 532 with peak intensity of 5 MW/cm2 and probe wavelength 600nm was used in (a) 400 ms; (b) 1000 ms time ranges. The stretched exponential function is a generalization of the exponential function with one additional parameter, the stretching exponent. It is

given by [5] , where α is the stretching coefficient between 0 and 1, τ is the stretched exponential life time, OD(t = 0) is the initial amplitude at t= 0 which sets as 0 and the OD1(t = t0) is the amplitude at a time constant τ . α and τ are highly correlated. The non-exponential time-resolved PIA decay in BSO:Cr is well fitted using a stretched exponential. We used several methods to get the best fits, which is described in detail in section 4.4, by using standard statistics on each level of the intensities, we found the average α = 0.22 is describing very well the dark decay in BSO:Cr for the entire time range that was detected. Figure 8 shows the optical density decay spectrum which is well fitted by stretched exponential curve fit in equation 5 in BSO:Cr crystals in the 1 s figure 9 (a) and 100 s in figure 4.8 (b) time ranges. The multiple fits for BSO:Cr yields τ = 0.5 s for 1 s and 100 s time range. The carriers dynamics PIA decay in Cr doped BSO SO can be explained using the concept of behind

sf ettet ))(() 00

OD(t) =OD(t=0) + OD(t=to)e

tttt

ODttODtODtOD00

()( 0

1 0

0

( ) ( 0) ( )

t t

t OD t OD t t eOD


16

NCSTI-2012

stretched exponential decay. The science is related withon the carrier tunneling between the localized state in the forbidden band. The fast component is coming from the short tunneling and slow component from the long tunneling of the localized states. Fig. 9. The normalized time-dependent optical density of a BSO:Cr thin crystal plates at room temperature. The pump wavelength 532 nm with peak intensity with 5 MW/cm2 and probe wavelength of 600 nm at time range of (a) 1 s; (b) 100 s time ranges.

4.4 Intensity Dependency of PIA Decay In BSO:Cr The intensity dependent measurements were performed for BSO:Cr sample at room temperature. The pump peak intensity of the pulses was run from 1.33 - 13.33 MW/cm2 with a step by step increment of 1.33 MW/cm2. The PIA decay curves for each pulse intensity were fitted by a stretched curve fit. From these curve fits α and τ parameter were determined as in equation 5. The mean value of α for the increase in pump intensities were determined and found that α = 0.22. Fixing this α, the curve fits were repeated to get the time. It is well fitted for our measured non exponential photoinduced absorption decay versus time range of 100 s. The pump intensity values used for this selective optical density graphs were 1.33, 3.99, 7.98, and 9.31 MW/cm2 respectively. The respective stretched time constant were τ = 2.12, 0.21, 0.42 and 0.19 s respectively. Figure 10 (a) shows the maximum amplitude versus pump intensity per pulse in Cr doped BSO crystal. It has depicted an increased behavior of higher intensity values. This graph also depicts saturation value of the maximum amplitude

for some higher intensity value. The observed behavior of this graph is an exponential increasing of the maximum amplitude and salutation at higher intensity values. Figure 10 (b) reveals the investigation of stretched time constant found using the strongly correlated average α = 0.22 versus pump intensity per pulse from 1.3 - 13.3 MW/cm2 per pulse in Cr doped BSO. It has been observed that the stretched time constant for high intensity values with constant behavior and high time constant for the lower intensity values for this crystal. These are related with lower frequency instability in the baseline at lower pump intensities. The high time constant for the lower intensity value comes from the low signal to noise ratio. For non-doped BSO, the temporal evolution depends on the intensity of the pump light, because the amplitude of the decay signal depends on the concentration of charges in the shallow traps [20]. At low light intensity the shallowest traps are highly ionized at room temperature and do not contribute much to the absorption..

0 . 0 0 .2 0 .4 0 .6 0 .8 1 .00 .0 0 0

0 .0 0 2

0 .0 0 4

0 .0 0 6

0 .0 0 8

0 .0 1 0

O

D

t im e [ s ]

d a t a p o in t s s t r e c h e d e x p o n e t ia l c u r v e f i t

s

( a )

0 1 0 2 0 3 0 4 0 5 0 6 0 7 0 8 0 9 0 1 0 00 . 0 0 0

0 . 0 0 2

0 . 0 0 4

0 . 0 0 6

0 . 0 0 8

0 . 0 1 0

0 . 0 1 2

0 . 0 1 4

O

D

t im e [ s ]

d a ta p o in t s a t 1 0 0 sd a ta p o in ts a t 1 s

s t r e c h e d e x p o n e t ia l c u r v e f i t fo r 1 s s t r e c h e d e x p o n e n t ia l c u r v e f i t fo r 1 0 0 s

s

( b )

0 2 4 6 8 10 12 140.000

0 .005

0 .010

0 .015

0 .020

0 .025

Experim ental result L inear fit

O

Dm

ax

Intensity [M W /cm 2 ]

(a )

0 4 8 1 2 1 6 2 0 2 4 2 8 3 2 3 6 4 00

1

2

3

4

5

6

7

F i t t e d t im e c o n s t a n t

[s

]

In t e n s it y [ M W /c m 2 ]

( b )

Fig. 10. (a) Maximum OD amplitude versus pulse intensity; (b) stretched time constant versus pump intensity in Cr doped BSO crystal at time range of 100 s at room temperature. The pump light wavelength 532 nm with intensity of 1.33 - 13.33 MW/cm2 and probe wavelength of 600 nm.


17

NCSTI-2012

4.5 TEMPERATURE DEPENDENCY RELAXATION OF OD IN BSO AND

BSO:CR The temperature dependence PIA decay measurements were performed on BSO and BSO: Cr crystals. It was carried out at peak intensity of 5 MW/cm2 and monitored by CW halogen lamp with probe wavelength 600 nm. The position of the trap levels and carriers tunneling were analyzed by a double exponential decay fit in non-doped BSO and by stretched exponential decay fits in Cr doped BSO crystals in order to determine an activation energy using Arrhenius law. This could help to explain the photorefractive effect of these crystals, the photochromic effect and the induced absorption which could help real application in holography, optical data storage handling, etc...

4.5.1 NON-DOPED BSO The time-resolved PIA measurement in the range of 296 K to 370 K in BSO crystals were analyzed using a double exponential decay fits. These yields the slow time and fast time constants for this crystal. Based on equation 2 and 3, Arrhenius plot were performed from the determined time constants, to determine the activation energy of the two classes of shallow traps corresponding to slow and fast component figure 11 (a) and 11(b)). The activation energy is determined by using slope found in equation 3 Ea = 0.17± 0.02 eV (frequency factor of A = 410 1/s or 405.96 1/s) for the slow and Ea = 0.43± 0.02 eV (frequency factor of A = 5.04x108 1/s ± 1.95 1/s) for the fast components. Marinova et al [14], reported the 0.46 eV for the slow time which is different from our result and 0.41 eV which is very close to our result. The previously reported values for the depth of the shallow trap in non-doped BSO are ranging from 0.32 eV to 0.50 eV. Khrimov at al [21] obtained the value of the activation energy in BSO which is 0.42 eV very close to our result. This was determined from the dark decay of the photorefractive grating formed by pulsed excitation experiments.

2 .6 2 .7 2 .8 2 .9 3 .0 3 .1 3 .2 3 .3 3 .4 3 .5

- 0 .8

- 0 .6

- 0 .4

- 0 .2

0 .0

0 .2

0 .4

0 .6

0 .8

1 .0

1 .2

1 .4

d a ta p o in t fo r s lo w t im e lin e a r f it

ln(1

/ s [

s-1]

1 0 0 0 /T [K -1 ]

(a )

2 .6 2 .7 2 .8 2 .9 3 .0 3 .1 3 .2 3 .3 3 .4 3 .52

3

4

5

6

7

d a ta p o in t fo r f a s t t im e lin e a r f it

1 0 0 0 /T [K -1 ]

ln(1

/ f [s

-1]

( b )

Fig. 11. The Arrhenius plot of the reciprocal of the decay time ln(1/τ ) versus 1000/T made for the measurements in the temperatures range 296 -370 K the dotted data for slow and fast component using double exponential decay for non doped BSO which results the activation energy (a) Ea = 0.17± 0.02 eV for the slow; (b)Ea = 0.43 ± 0.02 for the fast time constant. The probe wavelength is 600 nm and pump wavelength of 532 nm with intensity of 5 MW/cm2.

4.5.2 CR DOPED BSO The time-resolved PIA measurement in the range 296 K to 354 K in BSO crystals were analyzed using a stretched exponential decay fit. The measurements were done at probe wavelength of 600 nm and 700 nm. We were used an optical cryostat for our temperatures dependent measurements in range of 296 K to 370 K which increase 7 K in each step. The result we found for PIA carriers were determined using stretched exponential. The carriers decay well fitted by strongly correlated stretched exponential time constant τ by setting average α = 0.22. We found that for corresponding temperatures values (314K, 321 K, 336 K, and 349 K) with the respective time constant equivalence of (0.52s, 0.66 s, 0.49s, and 0.66 s). Consequently, using the complete time constant for each temperatures were plotted using the Arrhenius law in equation 2 and 3. Figure 12 presents the plot as the ln(1/τ ) versus 1000/T. The thermal activation for the plot were determined as Ea = 0.016 ± 0.03 eV (frequency factor A = 6.3 1/s or 0.5 1/s) from the linear fit. It has lower limit Ea ≈ 0 and upper limit Ea ≈ 0.046 eV . This shows the temperatures independency behavior in the relaxation of OD in this crystal. A very different mechanism has to be introduced to explain the stretched exponential behavior and the very weak temperature dependency. The micro-mechanism includes the carrier trapping, where the the recombination occur at the shallow levels and deep centers via an exponential distribution of the localized states in the band tail. Another reason is the carriers tunneling in a localized state of the Cr defects which do not involve significant activation energy. This behavior may occur from the system where the localized states are energetically far below a mobility edge [22] or the density of the


18

NCSTI-2012

localized state in large, such that the tunneling probability is significantly large than trapping probability. Since no any additional evidence (no published paper on the market the temperatures dependency of Cr doped BSO), we can only give our own discussion about the physical meaning of the temperatures independency behavior in light-induced absorption relaxation. The possible explanation could be a large tunneling probability between deeper Cr related levels in BSO. This will dominate the transport processes in carrier relaxation dynamics which also could lead to a exponential.

2.75 3.00 3.25 3.500.00

0.25

0.50

0.75

1.00

1.25

1.50

1.75

2.00

data point for fast time linear fit

ln(1

/ [

s-1]

1000/T(1/K)

Fig. 12. The Arrhenius plots ln(1/τ ) vrs. 1000/T made for the measurements in the temper atures range 296 -354 K the stretched exponential decay time constant for Cr doped BSO. The thermal activation energy from the linear fit is Ea = 0.016 eV with lower limit of Ea ≈ 0 and upper limit of ≈ 0.046 eV. The probe wavelength is 600 nm and pump wavelength of 532 nm with intensity of 5 MW/cm2.

V. CONCLUSION From the optical absorption spectra measurements, the Cr doped BSO has shifted the intrinsic edge to the longer wavelength and gave rise to new absorption bands in the visible and NIR ranges. The carriers relaxation dynamics in non-doped BSO and Cr doped BSO was studied using the PIA change by nanosecond pulse excitation using a frequency doubled Nd:YAG laser. The carriers relaxation time of BSO completed in few order of milli second whereas the Cr doped BSO completed in hundred seconds. The PIA decay of BSO is well fitted using non-exponential decay which is coming from the carriers trapping by the shallow and deep trap levels. On the other hand, Cr doped BSO is well fitted using stretched exponentials decay with a high correlated physical parameters α and τ which are related with microoptical mechanism of the carrier

tunneling between localized state of each Cr ions than carrier trapping. The pump intensity dependence behavior was investigated from 1.3 - 13.3 MW/cm2 per pulse for BSO:Cr. It has been observed that the stretched time constant is near to constant for high intensity values while higher time constants are formed for the lower intensities for this crystal. These high time constants seem unreliable and may result from low signal to noise ratio at low excitation intensities. Maximum amplitude versus pump intensity curve has shown a near linear behavior, with the sign of saturation at the higher intensity values. For non-doped BSO crystals, the time-resolved PIA decay relaxes first gradually then faster while increasing the temperature from 296 K to 370 K. From the Arrhenius plots, the thermal activation energy is calculated using the slope value. The slow decay component has an activation energy of Ea = 0.17± 0.02 eV and for the fast one Ea = 0.43 ± 0.02 eV. The thermal activation energy of the BSO:Cr crystals has been calculated from the Arrhenius plot. From the linear fit Ea = 0.016 ± 0.03 eV, with large uncertainty which have lower limit of Ea ≈ 0 and upper limit of Ea ≈ 0.046 eV. This shows nearly temperature independent behavior in this crystal. The possible explanation could be a large tunneling probability between deeper Cr related levels in BSO. This will dominate the transport processes in carrier relaxation dynamics which also could lead to a stretched exponential behavior.

IX. REFERENCES [1] G. Porter:Flash Photolysis and some of its applications Nobel Lecture (1967). [2] V. Marinova, I.Ahmad, E.Goovaerts, Draft paper: Relaxation of the light-induced absorption changes generated by short pulse excitation in Ru-doped Bi12SiO20. submitted to Appl. Phys. B (2009). [3] F. T.S Yu and S. Yin: Material, properties and application Photorefractive Optics, 26-40 (1996). [4] Yu Zheng: Melt growth of the BSO (Bi12SiO20): Critical issue for the growth in micro-g environment. Massachuset Institute of Technology, PhD thesis (1999). [5] Anita Fuchsbauer: Nanosecond Transient Spectroscopy on Conjugated Polymers, Diploma thesis. University of Linz, Austria (July, 2005). [6] V. Marinova, S. H Lin, V. Sainov, M. Gospodinov, and K. YHsu: Light-induced properties of Ru-doped Bi12TiO20 crystals. J. Opt. A: Pure Appl. Opt. 5 (2003).


19

NCSTI-2012

[7] R. Oberschmid Absorption center of Bi12GeO20 and Bi12SiO20 crystals. phys. stat. sol. (a) 89, 263 (1985). [8] P. Guter: Photorefractive Materials and Their Applications 2 ISBN-10: 0-387-33924-8, 9-45. (June, 2005). [9] A. A. Nechitailov, M. V. Krasin'kova, E. V. Mokrushina, A. A. Petrov N. F. Kartenko*, and V. V. Prokof'ev Impurity Defects in Cr-Doped Bi12SiO20 and Bi12TiO20 Crystals. Inorganic Materials, Vol. 36, No. 8, pp. 820,825. (2000). [10] A. V. Egorysheva, V. V. Volkov, C. Coya, and C. Zaldol: Tetrahedral Cr4+ and Cr5+ in Bi12TiO20 Single Crystals. phys. stat. sol. (b) 207, 283 (1998). [11] E.V. Mokrushina, A.A. Nechitailov, V.V. Prokoiev: Effect of a low chromium impurity on properties of photoinduced charge carriers in Bi12TiO20 and Bi12SiO20 single crystals. Optics Communications 123 592-596 (1996). [12] F. Ramaz, L. Rakitina , M. Gospodinov, Bernard Briat: Photorefractive and photochromic properties of ruthenium-doped Bi12SiO20. OpticalMaterial s 27, 15471559 (2005). [13] R. Oberschmid: Absorption center of Bi12GeO20 and Bi12SiO20 crystals. phys.stat.sol. 89, 263 (1985). [14] Chenting Lin: Single crystal growth and characterization of BSO (Bi12SiO20). Massachuset Institute of Technology, PhD thesis (1994). [15] I. Ahmad: Spectroscopic study of doping-compensated doped gallium arsenide and of Paramagnetic defects in photorefractive bismuth sillenites. Antwerp University, PhD thesis September (2009). [16] Li Fei-Fei, Jing-Jun, Kong Yong-Fa, Huang Hui, Zhang Guangi, Yang Chun-Hui, Xu Yu-Heng: Light-Induced Absorption in Nominally Pure Bismuth Silicon Oxide. Vol.18, No.12, 1595 (2001). [17] A.A Nechitailov,M. V. Krasinkova, E. V Mokrushina, A. A Petrov, N. F Kartenko, V.V Prokoiev: Correlation Between the Impurity Content, the Average Charge State of Chromium Cations and Optical absorption in Sillenite Crystals Doped by Chromium in Wide Range of Concentrations. Crystal.Res.Technol 36, 29(147156) (2001). [18] A. A. Nechitailov, M. V. Krasin'kova, E. V. Mokrushina, A. A. Petrov, N. F. Kartenko, and V. V. Prokof'ev: Impurity Defects in Cr-Doped Bi12SiO20 and Bi12TiO20 Crystals. Vol. 36, No. 8, pp. 983-989. (2000). [19] I. Ahmad, V. Marinova, and E. Goovaerts: High-frequency electron paramagnetic resonance of the hole-trapped antisite bismuth center in photorefractive bismuth sillenite crystals. Phys. Rev. B 79, 033107 (2009).

[20] k. Buse: Photorefractive crystals for holographic storage. Appl. Phys. B64, 233-291 and Appl. Phys. B 64 391 (1997). [21] A. L. Khrimov, A. A. Kamshilin, M. P. Petrov: Opt. Comm. 77, 139 (1990). [22] M. H Cohen, E.N. Economo, and C. M Soukolis, Bi12SiO20 monocrystals doped with transition metals. Non-Cryst. Solids 59, 15 (1983).

Improved Design of Modulated Filter Banks and Performance Comparison of Filter banks with OFDM in Multicarrier...

20

NCSTI-2012

Improved Design of Modulated Filter Banks and Performance Comparison of Filter banks with OFDM in Multicarrier

Communications Pattem Anjaneyulu

School of Computing and Electrical Engineering, Bahir Dar University, Bahir Dar, Ethiopia email: [email protected]

Abstract — Filter banks have been found wider applications in digital and wireless communications.

High speed digital data transmission has been used in modern wired and/or wireless communications systems such as WiMax, Wireless LAN, DAB, T-DVB, 3G/4G (Long Term Evolution-LTE) wireless mobile communications etc. Orthogonal Frequency Division Multiplexing (OFDM) has been used as the basic technique in all these systems. One of the drawbacks of OFDM is reduced transmission bit rate because of the cyclic prefix (CP) used in this system. An emerging technique proposed to be as an alternative to OFDM is applicability of digital filter banks. Filter banks designed for reduced stop band attenuation, and less distortion in frequency response are required in these applications. In the present work an improved design of modulated filter banks has been proposed. Filter banks designed in the proposed method is compared with the filters using already available methods in terms of stop band attenuation, magnitude and aliasing distortions. Then the performance of these filter banks in Matlab simulations of transmultiplexer and conventional OFDM systems is compared for transmission of data and an image data through AWGN channel for multicarrier communication applications. It is found that the filter banks designed in the proposed method are with higher stop band attenuation, and reduced distortions. Also the performance of the transmultiplexer communication is better than that of OFDM at lower SNR scenarios. Keywords- QMF, cosine modulated filter banks, transmultiplexers, OFDM, multicarrier communications

I. INTRODUCTION

Current modern high data rate digital wired and/or wireless communication systems such as WiMax, WiFi, Digital audio, video broadcasting and even 3G/4G systems (LTE, E-UTRA, HSPA) have been based on OFDM technique in one or the other way. One of the drawbacks of OFDM is less throughput because of some redundant bits are to be used as CP in these systems, albeit it is robust against interference effects. This reduced bit rate is to be accounted for in high speed future generation wireless applications. One of the promising alternative to OFDM, and emerging technique is filter bank multicarrier communications. The applicability of filter banks in speech signal compression, image coding, and in digital communications have been established and in future wireless communications it has been recognized [1]-[3]. Efficiency of filter bank based systems in cognitive radios have already been realized and an European task force has been formed ( FP-7) to study future radio systems based on filter bank multicarrier technology under PHYDYAS project.

Filter banks with perfect reconstruction and minimum distortion of the transmitted signal in the noisy environment is desired in high data rate applications. Initially a type of filter banks designed

for two channels signal transmission also called as quadrature mirror filters (QMF) later the theory has been extended to multi-channels as cosine modulated filter banks (CMFB) [4]-[7] for high fedility of the signal. Also it has been shown that the near perfect reconstruction performance can be achieved in these filter banks [6]-[12] eventhough ideal conditions are not possible. Hence, design of filter banks for perfect reconstruction with minimum distortion in frequency response characteristics and aliasing is of considerable interest in the recent past.

Filter banks implemented as transmultiplexers have been proposed to be as an alternative to OFDM in multicarrier communications. Applicability of filter bank based transmultiplexers in communications has been already established [13]-[15]. And in the future wireless is under progressing. The impetus for the present work is to study feasibility of filter bank systems in competent to OFDM. This problem is accounted in this work by designing a prototype filter and a comparison of performance is made on transmultiplexer and OFDM. Results have been incorporated in this paper. This paper is organized as follows: In the following section-II, design of prototype filter, which is a basic building block for cosine modulated filter banks, filter bank based transmultiplexers and OFDM techniques are


21

NCSTI-2012described. In section-III design examples of the filter banks and a comparison of performance of the filter banks with OFDM has been attempted. In the followed sections-IV and -V, discussion, and conclusions are given respectively.

II. DESIGN OF FILTER BANKS AND MULTICARRIER TECHNIQUES

From the theory of QMF filters it was shown that combination of analysis filters followed by decimators at the transmitter side and synthesis filters followed by interpolators at the receiver side can nearly perfectly reconstruct the input signal in the subband coding, provided the filters satisfied the necessary following conditions.

1()(2)/(2 Mjj ePeP for 0<ω<π/M-----(1)

0)( jeP for ω≥π/M --------------------------------(2)

1()(

2)/(2 Mjj ePeP -----------------(3)

where eq.1 and 2 are satisfied by the filters, it results minimum distortions in the magnitude response and aliasing. And it can be shown that the two equations are combined in the objective function given by equation 3 and near perfect reconstruction of the signal can be obtained at the minimum of the objective function given by equation 3.

When QMF filter theory has been extended to multichannels, it was shown that cosine modulated filter banks constructed from a basic low pass prototype filter is sufficient for near perfect reconstruction of the original signal. A good theory

Fig. 1. Transmultiplexer System

has been well documented [6]-[12]. In the present work a basic lowpass prototype filter is designed using linear optimization technique. In this technique initially the stopband, pass band, order of the filter, error tolerance, etc. are fixed and objective function is evaluated iteratively by varying band edge frequencies until the prescribed error tolerance is achieved [6]-[9], [12].

A transmultiplexer system basing on filter bank is shown in Fig.1. In this system signals of different sources are applied to interpolators before the synthesis filter bank. The outputs of the filter banks is combined at the transmitter and this serial stream is passed through AWGN channel. The channel output is then applied to decimators followed by analysis filter banks as shown in Fig. 1. The outputs of the analysis filter banks are taken as the reconstructed signal streams.

Fig. 2. A basic block diagram of Orthogonal Frequency Division Multiplexing (OFDM)

A block diagram of OFDM is shown in

Fig.2. It is consisting of upper chain of sections in the transmitter and lower chain of sections in the receiver. In the upper chain binary data stream is applied to signal mapper (not shown) followed by serial to parallel form converter. The parallel outputs are applied to IFFT section. Cyclic prefix is added to the outputs of IFFT and are then converted to serial form. The serial data stream is passed through AWGN channel (C) in between transmitter and receiver sections as shown in the Fig.2. At the receiver in the lower chain, the serial data stream is converted into parallel form and then cyclic prefix is removed. And the parallel stream is subsequently passed through, FFT, parallel to serial and signal demapper (not shown) sections. The serial output of the demapper can be recognized as the transmitted binary data stream.


22

NCSTI-2012 III-Design Examples and Performance comparison:

Magnitude response of 32 channel (M) and for a length of 512 filter bank using Park McCellan (P.M) method is shown in Fig.3. The upper panel shows the magnitude response using PM algorithm before the optimization. In this a stopband attenuation of 100dB is obtained. The filter designed for the same configuration in the proposed method is shown in the lower panel (Fig.3b). Interestingly an improved stop band attenuation of 130dB is obtained in the proposed method.

Fig.3. Magnitude response of 32 channel filter Upper panel: Filter using P.M method Lower panel: Filter designed using proposed method. . Method No.of

bands Length Stpbnd

Attn. (dB)

Mag. error

Aiasing error

Cruisere and Mitra

32 512 118 1.002 3x10-7

Kaiser window

32 466 100 1.002 4x10-7

Cosh window

16 97 45 1.002 2x10-3

Cosh window

08 45 35.8 1.002 3.79x10-3

Proposed Method

32/16 512/256 130 1.001 2x10-7

Table-I. Comarative analysis of low pass prototype filter

Fig. 4. Upper panel: Convergence of objective function Lower panel: (a) Magnitude error (b) Aliasing Error

Fig. 5. Magnitude response of 16 channel filter Upper panel: Filter using P.M method Lower panel: Filter designed using proposed method.

Convergence of the optimization is shown

in Fig.4 (upper panel). It shows that at a faster rate


23

NCSTI-2012and after 54 iterations minimization of objective function is achieved. The other features of the filter such as magnitude distortion and reconstruction errors are shown in Fig.4a and 4b (lower panel of Fig.4) respectively. These parameters are obtained as 1.002 and 2x10-7.

Plots of magnitude response of filters for 16

channel, length 256 are shown in Fig.5. Upper panel shows filters without optimization and the lower panel for filters in the proposed method. In this filter bank also a stopband attenuation of 130 dB is obtained. Also low magnitude and aliasing errors (not shown) are obtained. A comparative analysis of the filters designed in the proposed method and that of the other methods is shown in Table-I. This table is constructed from the data of filters designed using P.M. and window based methods [6]-[10]. Improved performance of the filters in the proposed method is observed when compared to other methods.

Fig. 6. Performance of transmission of data through tranmultiplexer (data 2) and OFDM (data 1) Upper panel: Transmission of data Lower panel: Transmission of image In order to study performance of these filters in multicarrier communications, Matlab simulations have been carried out for transmultiplexer and

OFDM systems. A plot of performance comparison of data transmission through a transmultiplexer in 32 channel configuration and OFDM system for the same number of channels is shown in Fig.6 (upper panel). In this figure, data 1 (blue line) is corresponding to OFDM and data 2 (red line) is for transmultiplexer. Mean square error (MSE) is evaluated in both the systems.

An impressive performance is observed in the transmultiplexer transmission at low SNRs between 4 to 5 dB. This is an interesting one, since most of the wireless transmissions require low error rate at low SNRs. At higher SNRs, performance of OFDM is found to be superior to transmultiplexer. Performance comparison of the same systems for image transmission is shown in Fig.6 (lower panel). In this plot, data 2 (red) is for OFDM and data 1 (blue line) is for tranmultiplexer. In this transmission also, a low error rate is obtained in transmultiplexer at low SNRs and at SNRs i.e. above 4dB, performance of OFDM is found to be significant.

IV. DISCUSSION

As shown in the above figures, there is an improvement of nearly 30 dB less stopband attenuation in the proposed method. Also amplitude and reconstruction errors are much less when compared to other methods. The improved preformance characteristics of the filters in the proposed method can be recognized not only from the filters based on Parks McCellan algorithms but also from window based designs as shown in the Table-I. It was demonstrated in [10] that the filters based on Cosh window are with superior performance to Kaiser window based design. In this work a better performance than the Cosh window based designs can be noticed. Eventhough the filters using Cosh window are with a little lower order, this feature can be superseded by the improved characteristics of higher stopband attenuation, minimum error distortions in the proposed method. An interesting result observed in this work is an improved performance of filter bank based transmultiplexer when compared to data transmission through OFDM at low SNRs. Similar result is observed in the image data transmission also. Thus a low error rate has been noticed in the simulations of the transmission of data and image through transmultiplexer when compared to OFDM at low SNRs (i.e.below 4 dB) and above this range OFDM is superior to filter banks.


24

NCSTI-2012

V. CONCLUSIONS

Improved characteristics of high stopband attenuation, and low error distortions are observed in the proposed filter design when compared to other methods. Also the performance of transmultiplexer based on these filters is found to be superior to that of conventional OFDM at low SNRs. This is a promising feature for achievement of higher spectral efficiency in filter bank based multicarrier communication systems for high data rate communication systems. This is to be investigated further in different multipath channel conditions in addition to AWGN channel. Also this is to be verified under different delay Doppler spread conditions for broadband mobile communications.

Acknowledgements: The author is highly

thankful to the Director of the School of computing and Electrical Engineering and higher authorities of the Institute of Technology and Bahir Dar University for their constant encouragement in carrying out this work.

REFERENCES

[1] M.Renfors, Pierre Siohan, B.Farhang-Boroujeny, and Faouzi Bader, “Filter banks for Next generation Multicarrier Wireless Communications” EURASIP Journal on Advances in Signal Processing, pages 2, Volume 2010 (journal editorial comments).

[2] A.N. Akansu, Xueming Lin, and Marc de Courville,”Orthogonal Transmultiplexers in Communication: A review” IEEE Transactions on Signal Processing, vol. 46, pp. 979-995, April 1998.

[3] Vidar Ringset et al., "Performance of a filter bank multicarrier (FBMC) physical layer in the WiMax context, “Future Network & Mobile Summit 2010 conference proceedings, ISBN: 978-1-905824-18-2.

[4] J. D. Johnston, “A filter family designed for use in quadrature mirror filter banks,” Proc. IEEE Int. Conf. Acoustics, Speech, Signal Processing, pp. 291–294, 1980.

[5] J.H. Rothweiler, “Polyphase Quadrature Filters- A

new subband coding technique” in Proceedings of the Int. Conf. Acoustics, Speech, and Signal Processing, MA, Apr. 1983, pp. 1280-1283.

[6] C.D. Creusere and S.K.Mitra, "A simple method for designing high quality prototype filters for M-band pseudo QMF banks”, ," IEEE Transactions on Signal Processing, vol. 43, pp. 1005-1007, April 1994.

[7] P.P. Vaidyanathan, "Multirate Systems And Filter Banks. Englewood Cliffs, NJ: Prentice-Hall, 1993.

[8] N.J. Fleize, “Multirate Digital Signal Processing”, John Wily Sons.

[9] S.K. Mitra “ Digital signal Processing – A computer based approach”,

[10] Yuan-Pei Lin, and P.P. Vaidyanathan, "A Kaiser

Window Approach for the Design of Prototype Filters of Cosine Modulated Filter Banks," IEEE Signal Processing Letters, vol. 5, no. 6, pp. 132-134, June 1998.

[11] Martin, Fernando et al., “ Prototype Filter Design Techniques for cosine modulated transmultiplexers”, ECCTD, August, 28-31, 2001, Finland.

[12] J.Ogale, and Alok Jain,”Design of an M-Channel Cosine Modulated Filter Bank by New Cosh Window Based FIR Filters”, World Academy of Science, Engineering and Technology, vol 72, pp. 239–244, 2010.

[13] M.Vetterly, "Perfect transmultiplexers," IEEE Int. Conf. on Acoustics, Speech, and Signal Processing , vol. 4, pp. 2567-2570, 1986.

[14] Waldhusier et al., “Multicarrier Systems & Filer Banks”, Adv. Rad. Sci., vol 4, pp-165-169, 2006.

[15] R.P. Ramachandran and P.Kabal, “Tranmultiplexers: Perfect reconcstruction and compensation of channel distortion” Singal Processing vol 21, 261–274, 1990.


25

NCSTI-2012


Zelalem Girma1, Martin Braun2

1 Arba Minch University and Fraunhofer IWES, P.O. Box 21, Arba Minch, Ethiopia

email:[email protected] 2 University of Stuttgart and Fraunhofer IWES, Pfaffenwaldring 47, D-70569 Stuttgart, Germany

Abstract — This paper presents the development of an effective approach of design, simulation and analysis of stand-alone hybrid renewable energy resources for typical rural village in remote area situated in SNNPR region of Ethiopia. It emphasizes the renewable hybrid power system to obtain a reliable autonomous system with the optimization of components size and Levelized Cost of Energy. The main power of the hybrid system comes from the photovoltaic panel batteries / inverter system, while the diesel generator is used as backup units. The optimization software used for this paper is HOMER. HOMER is design software that determines the optimal architecture and control strategy of the hybrid system. Critical decision variables like the size of the PV array, size and number of battery string, inverter, size of diesel Generator, dispatch strategy, are given weight in the optimization process. Wind speed and solar radiation data have been taken from NASA's meteorological department. A remote village with energy consumption of 279 kWh/day and 64 kW peak power demand was considered. The simulation results indicate that the proposed hybrid system would be a feasible solution for distributed generation of electric power for stand-alone applications at remote village with 200 households with average of five family members per household. An innovative approach of determining rural electric load for remote village which does not have electric access has been proposed.

Key words- Hybrid system, HOMER, Photovoltaic, Diesel Generator

I. Introduction Reliable access to electricity is a basic precondition for improving people’s lives in rural areas, for enhanced healthcare, education, and for growth within local economies as well as to meet millennium development goal in 2015. At present, more than 80% people in Ethiopia do not have access to electricity in their homes. Almost all of these people live in rural areas; most have scant prospects of gaining access to electricity in the near future. The Ethiopian Government tried to connect this rural location by using national grid extension for the last two decades. However, still the current electricity access is below 50% and the real connection is less than 14% [1]. In this scenario the rural people who have very low load demand with dispersed settlement will not get electricity in the near future

Energy is a key component of any poverty eradication and sustainable development strategy and is critical to the achievement of the millennium development goals. Better access to sustainable energy service for rural people in Ethiopia is prerequisite for the sufficient supply of lighting, communication systems, and the development of income generating activities as well as the improvement of the public health situation. One of the main problems of standalone system such as solar as well as wind energy is the fluctuation of energy supply, resulting in intermittent delivery of power and causing problems if the supply continuity is required. This can be

avoided by the use of standalone hybrid systems. A hybrid power system can be defined as a combination of different, but complementary energy generation system based on renewable energy or mixed (RES- with a backup of Liquefied Petroleum Gas (LPG)/diesel/gasoline gen_set). Hybrid systems capture the best features of each energy resource and can provide “grid-quality” electricity, with a power range of one kilowatt to several hundred kilowatts. They can be developed as new integrated designs within small electricity distribution systems (mini-grids) and can also be retrofitted in diesel based power systems. Hybrid systems can provide a steady community-level electricity service, such as village electrification, offering also the possibility to be upgraded through grid connection in the future.

Proposed Hybrid power system in this papertypically relies on renewable energy to generate 95% of the total supply. The large share of renewable makes this system almost independent and lowers the energy prices over the long- term, and the diesel generator set is used as a backup to assist in periods of high loads or low renewable power availability. The battery backup size is lower due to back up system and suffers less stress than in a 100% renewable power system, prolonging battery lifetime significantly and reducing replacement costs.

II. Methodology The simulated hybrid renewable energy system

comprises of wind turbine, Photovoltaic (PV) array with power converter, battery and Diesel generator.


26

NCSTI-2012The battery is added into the system as a backup unit and act as a storage system. This system is designed specifically for an off grid system at remote area to supply power 24/7 bases. The system is designed by considering remote village called Dembile which, is located around 80km from Arbaminch town, in Bonke woreda in SNNPR region of Ethiopia. The solar and wind resource data of the remote site was taken from online data of NASA Methodological department [11]. The field surveys has been conducted to get daily load profile and energy usage pattern of the village. Since the village does not have Electricity access, the daily load profile of electrified village with the same socio economic condition with the selected village has been taken for simulation. The HOMER software is used to determine the optimal sizing and operational strategy for a hybrid renewable energy system based on three principal tasks which are simulations, optimization and sensitivity analysis. The following subsection discusses on the three principal tasks of the HOMER software. A. HOMER: Simulation

HOMER simulates the operation of the system based on the components chosen by the designer. In this process, HOMER will perform the energy balance calculation based on the system configuration consisting several numbers and sizes of component. It then determines the best feasible system configuration which can adequately serve the electric demand. HOMER simulates the system based on the estimation of installing cost, replacement cost, operation and maintenance cost, fuel and interest rate.

B. HOMER: Optimization Optimization process is done after simulating the

entire possible solutions of hybrid renewable energy system configuration. HOMER display a list of configurations sorted based on the Total Net Present Cost (TNPC). It can be used to compare different types of system configuration from the lowest to the highest TNPC. However, the system configuration based TNPC is varied depending to the sensitivity variables that have been chosen by the designer.C. HOMER: Sensitivity Analysis

The HOMER software will repeat the optimization process for every selection of sensitivity variables for the hybrid renewable energy system. The sensitivity variables are such as the global solar, wind speed and the price of diesel fuel. Then, the list of various configurations of hybrid renewable energy will be tabulated from the lowest to the highest TNPC. The optimal solution of hybrid renewable energy system is referring to the lowest TNPC.

III. System Configuration

The typical wind-solar hybrid power generation systems include PV system, WT system, battery units, diesel generator, related electric devices and loads. Wind-solar hybrid power generation systems can be divided into three classes according to bus bar forms, including pure AC bus bar system, pure DC bus bar system and hybrid AC-DC bus bar system. The three classes systems have different features. The detail explanation of each configuration found in [19]-[27]. In this paper, AC-DC configuration is used due to its advantage compared to other configurations.

Fig.1 AC-DC Hybrid configuration

AC -DC configuration has superior performance over the other type of hybrid system. In this scheme the renewable energy source and the diesel generator supply a portion of load demand directly, resulting in higher overall system efficiency. The diesel generator and the inverter can operate in stand alone or parallel mode. This offers some combination of the source for meeting the load. When the load is low, either the diesel generator or the battery can supply the load. However, during peak load both sources are operated in parallel mode. Due to this parallel operation the initial capacity of diesel generator and inverter can be reduced. In this scheme a controller is needed to supervise the operation of the system, selecting the most appropriate mode of operation to supply a certain load without power interruption.

IV. System Description and Simulation From the design point of view, the optimization

of the size of hybrid plants is very important, and leads to a good ratio between cost and performances. Before the system sizing, load profile and available renewable resource of the site should be evaluated. The load profile for hybrid system was created from result of survey of electrified village with the same socio economic status of the selected case study village. The daily load and hourly load was calculated by using spreadsheet program


27

NCSTI-2012

EXCEL. The load profile of the village is tabulated in Table 1.

A. Electrical load information As seen from the survey of some rural villages in

Ethiopia the electrical load demand is very low dominated by lighting load. In this study, 200 rural household with average family size of five, public and commercial centers are considered.

Around three water pumps are assumed to deliver the water need. One pump used for school, health clinic and a milling house; and the remaining two for house use. The selected type of water pump has a capacity of 150W power rating, with a pumping capacity 20 liter/minute. The required amount of water needed per family is ~100 liter/day, for cattle ~25 liter /day, for school, health center, shopping center and milling house~2000 liter/day. The above assumption is based on country average consumption of water per person and per cattle, i.e. the average consumption is 20 liter/person/day and 25liter/cattle/day. Since in the village we assumed 200 households with average of five members per family and three cattle per households, the total consumption of water per day is around 36m3/day. Three water pumps with capacity of 20 l/min can provide more than 36m3/day if it runs for 10 hour per day. A water tank of capacity 43 meter cube is considered and at full load the pumps draw 0.75kW of electrical power and pumps 3.6m3 per hour. The peak deferrable load is 0.75kW, which is the rated power of the pump. It would take 12 hour for the pumps at full power to fill the tank. So the storage capacity is 12 hour times 0.75kW, which is 9kWh. It would take for the pumps 10 hour at full power to meet the daily requirement of water for the village. So the average deferrable load is 10 hour per day times 0.75kW, which is 7.5kwh/day.

By referring the load profile given in Fig.2 and Fig.10, 279kWh/day is the average estimation of daily energy consumption of primary load with 64 kW peak and 4.2 kWh/day deferrable loads with 750 watt peak. Three water pumps which have capacity of 20 l/min have been selected as deferrable load in the simulation. In order to analyze uncertainty in the future, load sensitivity analysis has been done by 10% and 20% increment of the load. The monthly load demand of this village is shown in Fig. 3. It is observed that the annual peak load of 64 kW has occurred in March, Jun, July and December. The daily peak load occurred from 6:00 up to 13:00.

Fig.2 Daily load curve of the village

Fig.3 Monthly load variation of the village

B. Photovoltaic (PV) economic information and solar resource

The size of a PV array system in the optimum system is 78 kWp. While the total capital cost is $218,400 and the replacement cost is $0 since the project life time is the same as PV array life time, which is 25 year. The design accounts for the decrease in PV efficiency panels with the ambient temperature. The solar radiation data is taken from NASA meteorological department database. The array slope angle is set to 15 degree and the array azimuth is 0 degree which is referring to the South direction. The lifetime for this PV array system is 25 years with a de-rating factor of 90% and ground reflectance of 20%. Fig.4 shows the average monthly solar radiation data of the village where maximum radiation occur in the month February and the minimum radiation available in the month of July, which is the raining season of the region. Fig.5 shows the average daily solar radiation where the maximum occur around noon 12:00. It is seen from the data that the site has tremendous solar resource potential with average radiation of 6kWh/m2/ day. This is the reason that 95% of electrical energy come from Photovoltaic array while the rest 5% diesel generator in optimum system.


28

NCSTI-2012

Fig.4 Monthly average solar radiation of the site Fig.5 Hourly solar radiation curve of the site

Table1. Load profile of the Village

Residential load Type No of house

holdAppliance type Ratting (W) No. of

ApplianceRun time h/day kWh

CFL lamp 15 4 5 0.3 Tape Recorder 50 1 10 0.5 Television 250 1 8 2

High class House Hold

10

Total kWh/day /high class house hold 2.8kwh/day CFL lamp 15 3 4 0.18 Tape recorder 50 1 10 0.5

Middle classHouse Hold

100

Total kWh/day /middle class house hold 0.68kWh/day

CFL lamp 15 1 4 0.6kwh Low class House Hold

90Total kWh/day /low class house hold 0.6kWh/day

Public and commercial load

CFL lamp 15 15 5 1.125 Tape recorder 75 1 8 0.6 Television 250 1 6 1.5 Computer 700 3 7 14.7 Refrigerator 200 1 8 1.6

School 1

Others 250 1 6 1.5 Total kWh/day 21.025 kWh/day

CFL lamp 15 8 8 0.96 Tape recorder 50 1 6 0.3 Television 250 1 14 3.5 Lab. equipment 1000 1 12 12Refrigerator 200 1 8 1.6

Health Center 1

others 250 1 6 1.5 Total kWh/day 19.86 kWh/day

CFL lamp 15 15 8 1.8 Tape recorder 75 5 8 3Television 250 4 6 6Razor 20 8 12 0.48 Grinding mill

machine12000 1 14 168

Commercial Load

others 250 6 1.5 Total kWh/day 180.78kWh/day


29

NCSTI-2012.

C. Wind turbine parameters & wind resource A monthly average wind data of the village

were taken from online data of NASA meteorological department. This is an average of the last five year and shown in Fig.6. Wind energy is clean, free and inexhaustible. It can be harnessed for producing electricity by using a wind turbine. Wind turbine will use the force from air current flowing across the Earth’s surface which is called surface wind. The wind power is defining from the wind velocity and area of the wind flow:

P=0.5AV3…………….(1)

WhereP: the power in the wind (watts) : Air density (1.225 kg/m3 )

A= π/4) D2, the cross-sectional area through which the wind passes (m2), D: is the diameter of the turbine blade

V: wind speed normal to A (m/s)

From Equation (1) the power in the wind increases as the cube of wind speed. This means, for example that doubling the wind speed increases the power by eightfold and small variation in wind speed produces wide variation in wind turbine power output. The cost of a turbine increases roughly in proportion to blade diameter, but power is proportional to diameter squared, so bigger machines have proven to be more cost effective. The wind speed of selected village is very low and it ranges from 3m/s to 4.5 m/s with monthly average of 3.91m/s.

Fig.6 Average monthly wind speed over the year

The wind speed over a year is presented in a weibull distribution form in the Fig.8. The autocorrelation factor of 0.85 is measured based on the hour to hour randomness of the wind speed. The diurnal pattern strength of 0.25 represents as the strength of a wind speed

Fig.7 wind speed variation with height

Fig.8 Weibull probability distribution function of wind speed

D. Diesel Generator The diesel power plant of 5 kW is used in

optimal configuration. The diesel price with four discrete values of 1$/L, 1.2$/L and 1.5$/L are used for the sensitivity variables. At present, the diesel price is about 1$/L in Ethiopia. The lower heating value is 43.62MJ/kg, density of the fuel is 820kg/m3 and carbon content is 88% and sulphur content is 0.33%.The diesel generator forced to operate from 6:00 to 14:00 and from 18;00 to 20:00 for optimum system to supply nighttime load and to avoid frequent startup of generator which reduce its life.E. Battery

The type of battery that used for the system is Surrette 6CS25p model with the rating of 6V, 1156Ah, 6.94kWh with lifetime throughput 9645kWh. The cost for one battery is $900 with the replacement cost of $800. The battery stack is containing several numbers of batteries and the battery string contains 8 batteries in series with bus voltage of 48v. Total of 96 batteries are used in


30

NCSTI-2012

consid 4 and

assumptions of component F

converter is 10 years with the efficiency of 9

suppliers [3]-[21]-[26] and summarized in Table 2.

V. Operation Principle of Standalone Hybrid

he ene

ith surplus power going to charge the battery bank.

optimal system. The quantities of batteries string 16strings.ered in simulation are 8, 10, 11,12,13,1

Table 2.Technical data and study of . Power converter A power electronic converter is used to maintain

the flow of energy between ac and dc components. The optimal size of power converter used in this system is 32 kW. The capital cost, replacement costs and maintenance for this equipment for 1kw is $700, $700 and 50 $/ year respectively. Seven different sizes of converter which are 15kW, 20.kW, 30kW, 32kW, 35kW, 40.kW and 45 kW are considered in the simulation. The lifetime for one unit of

5%.Technical data and cost of components used in

the design of hybrid system is taken from the web site of corresponding manufacturers and

SystemThe operation principle of proposed hybrid

system is based on dispatch strategy and energy management unit. The PV array which is the basic load supplier will charge the battery bank when there is an excess power remaining after meeting the load demand. Since output power of PV module is in DC mode, they must be converted to the AC power by using an inverter so that it can be utilized to meet the AC load demand. If PV array cannot meet the demand, the battery bank will not be charged, but will be discharged to supply the demand. The allocation of resource based on load condition is called dispatch strategy. The dispatch strategy for a hybrid energy system is a control algorithm for the interaction among various system components. The control strategy determines t

rgy flows from the various sources of energy. HOMER can model two dispatch strategies,

cycle charging and load following. Which is optimal depends on many factors, including the sizes of the generators and battery bank, the price of fuel, the O&M cost of the generators, the amount of renewable power in the system, and the character of the renewable resources. Under the load followingstrategy, whenever a generator is needed it produces only enough power to meet the demand. Under the cycle charging strategy, whenever a generator hasto operate, it operates at full capacity w

VI. Result and Discussion The proposed hybrid renewable energy system for

the village is shown in Fig.10. It consists of primary load of 279kWh/day, with peak load of 64kW and deferrable load of 7.5kWh/day with peak load of 750W. Deferrable load is electrical load that must

PV Array DC Capital cost $4000 Replacement cost 00$30O& M cost $0Efficiency 15%Lifetime 25Tracking system No traking Wi rnd Generato DC

Technology –Excel BWC R/48

power 7.5 kW DC,48V Hub Height 50m Capital cost $30000 Replacement cost 00$250O& M cost $50 Lifetime 10 yearDiesel Generator ACCapital cost 500$/kW Replacement cost 400$/kWO& M cost r0.015$/houLifetime 25000hr Battery DCTechnology S25 Surrette 6CCapacity 6.94kWhNominal Capacity Ah 1156Voltage 6VMin. state of charege 60%Capital cost $900 Replacement cost $800O& M cost ear10$/yEfficiency 80%Lifetime 10year Converter AC/DC/AC Capacity 30kWCapital cost $700 Replacement cost $700O& M cost 50$/yearEfficiency 90%Lifetime 10 yearSystem Data Project life time 25 year Operating strategy following LoadSpinning reserve 10Set point SOC 60% Maximum annual capacity shortage 20% Daily noise of load 15% Hourly noise of load 30%


31

NCSTI-2012be met within some time period, but the exact timing is not important. Loads are normally classified as deferrable because they have some storage associated with them. Therefore, in this simulation three water pumps are considered as deferrable load. The software considered this deferrable load as critical load when the water tank is empty. Therefore, the Combination of equipment is depending up on the optimization procedure and sensitivity variables.

system

orted based on the T

a rational result f hybrid renewable energy system

ergy System Considering

ti

L,w

converter with minimum COE of $0.401/kWh.

Fig.9 Hybrid power system configuration

A. Optimization of hybrid renewable energy without considering sensitivity variables

HOMER performs the optimization process in order to determine the best solution in terms of component size and Total Net present cost of hybrid renewable energy system based on several combinations of equipment. Hence, multiple possible combinations of equipment could be obtained for the hybrid renewable energy system due to different size of the equipment. In the optimization process it simulates every combination in the search space. The feasible one will be displayed at optimization result s

otal Net Present Cost (TNPC).The combination of system components is

arranged from most effective cost to the least effective cost. The optimization results of hybrid renewable energy system are obtained for every selection of the base case i.e. fuel price 1$/L, primary load 279 kWh/Day, real interest of 10% and PV cost and replacement multiplier of 1 .Table 3 shows a list of optimization results for the hybrid renewable energy system without considering the

sensitivity variables. The total net present cost for optimum combination with out considering sensitivity variable is $412,720 and cost of energy (COE) is 0.538$/kWh with total renewable fraction of 89%. However, sensitivity variables should be taken into account in order to obtain o

B. Hybrid Renewable EnSensitivity Variables

Sensitive variables are very essential to consider uncertainty of input variables such as wind speed, solar radiation, load variation, etc. in the future. In this simula on the following sensitivity variables

y

.7

are used. Primary load 279,300 and 350 kWh/da Real interest rate 6 ,8 and 10% Photovoltaic capital cost multiplier 1 and 0 Diesel fuel price 1$/L, 1.2$/L and 1.5$/L

The above sensitivity variables are used in the optimization process to obtain the best configuration of hybrid renewable energy system consisting of diesel generator, PV array system, battery storage and/or power converter with total net present cost of $443,627 and cost of energy $0.401/kWh as shown in Table 4. In this case study, the system consisting of diesel generator, PV array, battery storage and power converter yields to the most economical cost with the minimum TNPC of Energy. The second cost effective system from categorized simulation result is wind-PV system. As seen from wind speed profile of the village, the monthly wind speed is very low and the output from wind turbine is insignificant. The energy obtained from different components of hybrid renewable energy system is shown in Fig.10. The PV array produced 146,066kWh/yr that is 95% of the total energy served. The remaining 5% of total energy is served by the diesel generator, which is 6,896kWh/yr. This system produced 13.8% of excessive energy. Table 4 shows that the TNPC of PV-Generator battery hybrid system become economically feasible when the primary load is varied from 279kWh/day to 350kWh /day and annual real interest rate 10%. The system is also economical even if the diesel fuel price increase from the current 1$/L to 1.5$/

hich is actual expecting due to soaring oil price As it is observed from simulation PV/gen/battery

system is still optimum for wide variation of load and diesel fuel price. Therefore, the optimum system for the case study village is PV/gen/battery system with


32

NCSTI-2012Table3. Optimization result without considering sensitivity variable

Table 4. Optimization result with sensitivity variables

Fig10. Monthly average Electrical energy production of hybrid energy system

Fig.11 cost summery by component

Fig.12 Cost summery by cost type


33

NCSTI-2012

Fig.13 Cash flow summery

VII. CONCLUSIONS

Reaching the non electrified rural population is currently not possible through the extension of the grid in Ethiopia, since the connection cost is not economically feasible due to dispersed rural settlement and low load factor. Further, the increases in oil prices and the unbearable impact of this energy source on the users and on the environment, are slowly removing conventional energy solution such as fuel gen_set based systems, from the rural development agenda. Grid extension and off grid hybrid solutions are complement each other rather than compete. Off-grid hybrid stand alone renewable energy are economically viable in lieu of grid extension in areas where there is low load factor, low population densities and difficult geographic terrain to be crossed

From simulation result, the combination of PV array, diesel generator, battery storage and converter brings to the optimal configuration of hybrid renewable energy system applicable to be used as an off-grid system for selected village of 200 house hold in southern region of Ethiopia with cost of energy $0.401/kWh. Since the solar resource potential of the site is high 95% of the energy is produced from solar array and 5% from diesel generator. The energy storage system and inverter should be replaced two times during project period. However, the last time replacement of the battery occurred in 24th year from 25 year of total project life time salvage value of around $90,000 left at the end of project period. As seen from the simulation result the designed system can provide 24 hour electricity for the village without interruption.

Despite their significant benefits to the environment and great long-term potential for sustainable energy development, hybrid power systems are currently in an economic disadvantage position because of their high installation costs compared with traditional electric generation as seen from Fig.12. In the majority of cases, the incentives from federal and state governments and local utilities are necessary to make a hybrid system economically viable, which, in turn, makes the incentive policies so critical to the widespread deployment of such systems.

The cost of energy (COE) of the optimal system is higher than the current grid price of electricity, which is highly subsidized by the government. In this simulation we did not consider any subsidy from the government. However, if we consider the Economic and social values that stand alone hybrid system brings to community this system is still feasible in comparison with grid extension to remote site, which is far away from access to grid line and its is more reliable with high quality of Electricity when compared to stand alone PV system.

REFERENCES

[1]Ethiopian Electric Power Corporation (online) .Available: http://www.eepco.gov.et/. Accessed on 10, Apr, 2011

[2]Alliance for Rural Electrification, “Hybrid power systems based on renewable Energies; suitable and cost competitive solution for rural Electrification”


34

NCSTI-2012[3]North Arizona wind and sun (online. Avalable:

http://www.solar-electric.com/hiposopa.html.Accessed on 10, Dec, 2011

[4] ESMAP Technical Paper 121/07, December 2007 [5] Sharew Anteneh Mekonnen, “solar energy assessment

in Ethiopia modeling and measurement”, Msc thesis, Addis Ababa University, Ethiopia

[6] M.A. Elhadidy and S.M. Shaahid, “Parametric study of hybrid (wind + solar + diesel) power generating systems”, Journal of Renewable Energy, Vol. 21, pp. 129-139

[7] Traca A, et al. “Source reliability in a combined wind–solar–hydro system”.,IEEETrans Power Apparatus System1983;PAS-102(6):1515–20.

[8]Elhadidy MA, Shaahid SM. “Feasibility of hybrid (wind+solar) power systems for Dhahran, Saudi Arabia”. Journal of Renewable Energy, Vol. 16, pp. 970-976

[9]Tomas Markvart; “Sizing of hybrid PV-Wind Energy System”, solar energy vol 57 no.4 pp 227-281

[10]H. A. Aulich, F. Raptis, “Rural Electric Hybrid Plants State of the Art and Future Trends”

[11]NASA,(online)Available: http://eosweb.larc.nasa.gov. Accessed on 20 September, 2011

[12]Givler,T. and Lilienthal, P. (2005). “Using HOMER Software, NREL’s Micro-power Optimization Model, to explore the Role of Gen-sets in Small Solar Power Systems; Case Study: Sri Lanka Technical Report” National Renewable Energy Laboratory,USA.

[13]Goldemberg, José, “Rural energy in developing countries” In World Energy Assessment: Energy and the Challenge of Sustainability, United Nations Development Program (UNDP), Brazil, 2000

[14]Bergey Wind power (online). Available: http://www.bergey.com. Accessed on 15 Sep, 2011.

[15]Tomas Markvart, 1996. “Sizing of hybrid PV–wind Energy systems”, Solar Energy 59, pp. 277-281.

[16]HOMER V2.68 beta, National Renewable Energy Laboratory (NREL), 617 Cole Boulevard,(online) Available : http://www.nrel.gov/homer.

[17]A. K. Akella, R. P. Saini, M. P. Sharma, “Sizing andCost Analysis for Integrated Renewable Energy System in aStudy Area”

[18]Akanksha Chaurey, Tara Chandra Kandpal: “Assessment and evaluation of PV based decentralized rural electrification: An overview”Renewable and Sustainable Energy Reviews 14 (2010) 2266–2278

[19]M.H.Nehrir, C.Strunz “A Review of hybrid renewable /alternative Energy for Electric Power Generation”, IEEE Transaction on Sustainable Energy Vol, 2 No.4, October 2011

[20]Stutenbaumer U, Negash T, Abdi A. “Performance of small-scale photovoltaic systems and their potential for rural electrification in Ethiopia”. Renewable Energy 1999; 18(1–2):35–48.

[21]Solar market Research and analysis (online) Available: http://www.solarbuzz.com/node/3184,accessed on 15 Sep, 2011

[22]World Bank (online) Available: http://www.worldbank.org/energy. Accessed on 14 May, 2011

[23]Adriaan N. Zomers, “Rural Electrification”, Ph.D. Thesis, University of Twente, 2001

[24]Wuyuan Peng, Jiahua Pan, “Rural Electrification in China:” History and Institution China & World Economy / 71 – 84, Vol. 14, No. 1, 2006

[25]E.M. Nfaha, J.M. Ngundam “Simulation of off-grid generation options for remote villages in Cameroon”, Renewable Energy 33 (2008) 1064 -1072

[26]Hani S. Alganahi, Kamaruzzaman S, “Experimental Study of Using Renewable Energy in Yemen”Australian Journal of Basic and Applied Sciences, 3(4): 4170-4174, 2009 ISSN 1991-8178

[27]R.Ramkumar: “Renewable Energy Resources and Developing Countries”, IEEE transactions on power apparatus and systems, Vol. PAS-102, No. 2, Feb.-1983, pp-502-510

[28]SMA Solar Technology AG (online). Available www.sma.de , accessed on 20 Feb,2012.

[29] Masters and Gilbert, “Renewable and Efficient Electric power systems”, John Wiley& sons inc. publication, ISBN 0-471-28060-7-2004

^¨


35

NCSTI-2012


Gashaw Mihretu1, M.Jayachandran1, Agizew2

1 School of Computing & Electrical Engineering, Bahir Dar University Ethiopia 3 Ethio Telecomm, Bahir Dar, Ethiopia

Abstract — A standby microwave link is a terrestrial communication system, used as an

alternative link for fiber optics or the area which is difficult to practically implement fiber optics. It uses a beam of radio waves in the microwave frequency range to transmit video, audio or data between two locations, which can be from just a few meters to several kilometers apart. Presently the communication to woretta is serviced only on optical fiber, and any disruption of optical fiber results in isolation of the region. This paper is intended to provide a concept and insight of the need of planning and designing a standby microwave link between Bahir Dar and Woretta. The traffic from and to Woretta is likely to increase in future as it is one of important centre for growth of economy and investment in Amahara region due to its richness in land & water resources,. Hence it is pertinent to provide reliable and efficient alternate communication link. The site survey, fade margins, frequency planning, path budget calculation, and performance evaluation activities are discussed in this paper and the result is simulated by using relevant software’s. A direct link is not possible due to terrain consideration and a two hop link is suggested.

Keywords- : Line Of Sight, First Fresnel Zone, Microwave, Path Profile, Propagation Loss, Fade

Margin, Link Budget and International Telecommunication Union Regulations.

I. INTRODUCTION Microwave link is a terrestrial communications

system, and is very important in Ethiopia because most of its geographical areas are not suitable for wired communications systems. It is an alternative link for fiber optics in the area which is difficult to practically implement fiber optics, in both short and long-haul telecommunications. It uses a beam of radio waves in the microwave frequency range to transmit video, audio, or data between two locations.

Microwave frequencies are useful for terrestrial and satellite communication systems, both for fixed and mobile. In the case of point-to-point microwave link, antennas are placed on a tower or other tall structure at sufficient height to provide a direct, unobstructed line-of-sight (LOS) path between the transmitter and receiver sites. In the case of mobile radio systems, a single tower provides point-to-multipoint coverage, which may include both LOS and non-LOS paths.

Any frequency within the electromagnetic spectrum associated with radio wave propagation is referred as Radio Frequency (RF). When an RF current is supplied to an antenna, an electromagnetic field is created that then is able to propagate through space.

Microwave link design is a methodical, systematic and sometimes lengthy process that includes: Site survey, Loss/attenuation Calculation, Fading and fade margins calculations, Frequency planning and interference calculations, Quality and

availability calculation activities. The whole process is iterative and may go through many redesign phases before the required quality and availability are achieved [4]. This process is show by the following block diagram in figure 1.

Fig 1 Microwave link design process flow chart

Frequency planning

Interference analysis

Link budget

Propagation losses

Branching losses Other losses

Fading predictions

Rain attenuation

Site survey

Multi path propagation

Diffraction - refraction losses

Quality & availability calculation

ITU-R P.530-xx propagation models


36

NCSTI-2012

II .MICROWAVE LINK SITE SELECTION AND PATH PROFILE PREPARATION

For all microwave links, it is imperative to actually perform physical path survey and not to rely just on the maps and/or aerial photographs. A path profile is a graphical representation of the path traveled by the radio waves between the two ends of a link .The path profile determines the location and height of the antenna at each end of the link, and it insures that the link is free of obstructions and propagation losses from radio phenomena, such as multipath reflections [5]. A) The Proposed Link -Bahir Dar To Woretta

The first step of microwave link design is identification of suitable terminal radio sites from figure 2 which is the area depicting Bahir Dar, Zege, Woretta and Lake Tana. After this the path profiles are determined (i.e. Coordinates of the of the radio terminals, maximum hop length, visibility line of sight of the terminals).

The proposed link is from Bahir Dar to Worotta and its path profile is show in the table 1 below

Radio terminal sites

Path profile Bahir dar Worotta

Latitude 11035’726”N 37O55’505”N

Longitude 37O23’ 196”E 37O41’665”E

Elevation 1798m 1850m

Maximum hop distance

49.4km 49.4km

Table-1 Coordinates and hop distance from Bahir Dar to Worotta

The coordinates of the sites are collected by using GPS (global positioning system) from the two stations.

Fig 2. Map of area of interest

Based on the above, path profile of the two sites, the maximum hop distance and the maximum height of obstacle are shown in the figure 3 below. This is done by using Global mapper 12 software.

Distance in km

Woretta BDR

Fig 3. Path profile from Bahir dar to Worotta

From the path profile shown in the figure 3, we see Bahir dar and Worotta are at a height (elevation) of 1795m and 1825m respectively. Between the two terminals the maximum obstacle is located at a distance of 29.421km from Bahir dar with a height (elevation) of 1949m. The elevation of the obstacle is very high; there is no clear line of sight (LOS) between the two radio terminals. To have clear line of sight, it requires around 2000m tower height, which is not practical and not recommended. The best solution for this problem is to find other alternative path and it is discussed below.

A) Recommended Two hop link—Bahir Dar to Zege and Zege to Woretta

From figure 3 it is clearly seen that due to LOS obstruction, there is no chance of establishing the direct microwave link between the stations Bahir dar and Woretta. Hence a repeater is suggested. Use of a repeater enables transmission system to regenerate or replicate the signal distorted by transmission loss. It acts as a relay between the two stations when it is impossible for radio signal to reach up to another station with enough power level due to LOS (Line Of Sight) problem and far distance.

We can establish the repeater station at Zege and relay microwave radio signal between these two stations without the significant attenuation of signal.

.

Fig 4. Repeater between Bahir Dar and Woretta.


37

NCSTI-2012

The alternative path viewed for analysis has two hops, one from Bahir- Dar to Zege and the other from Zege to Worotta with repeater at Zege. Further by having ADM equipment at Zege, we can add / drop data, voice and video sevice and thus we will be having alternate link to Zege too. The path profiles for this link are shown in the table 2 below. Path from Bahir Dar To Zege

The maximum hop distance between the two terminals is 12.90km. Between these terminals there are no obstacle because of Lake Tana, and both terminals has approximately the same elevation. So there is a clear line of site between the two sites. The path profile of the link between the two sites analyzed with the help of Global mapper 12 software is shown Figure 5 below.

Distance in km Fig 5. Dist from Bahir Dar to Zege and their elevation

Path from Zege to Woretta The maximum hop distance between the two

terminals is 46.7km and the obstacle exists at 19.557km distance from zege with maximum elevation (height) of 1876m. As shown in the figure below, the Zege terminal has grater elevation than the obstacle height, so there is clear line of site.

Woretta

Distance in km Fig 6. Dist between Zege to Woretta and their elevation

IV. FIRST FRESNEL ZONE (FFZ) CALCULATION

A Fresnel zone is defined as a family of ellipsoids that can be constructed between a transmitter and a receiver by joining all the points for which the excess path delay is an integer multiple of half wavelengths or it is the area that the signal spreads out. The radius of first Fresnel zone can be expressed by the following equation.

F1= √λ*d1*d2/d1+d2 (1)

Where F1 = FFZ radius in meters d1 = dist from the one site to the obstacle in meter d2 = dist from the obstacle to another site in meters λ = wavelength of the transmitted signal in meters For LOS propagation between the two sites, the maximum FFZ radius is calculated by the following formula

F1=8,657√D/f (2) Where D= the maximum hop distance in meter f = operating frequency of the link

A) FFZ - Bahir Dar To Zege

The maximum FFZ radius of this path is calculated by equation (2) using D = distance from Bahir Dar to Zege in meter =12.90km = 12900m F = operating freq of the link (2.4GHZ) F1=8,657√D/f = 8,657√12900 /2.9 = 18.258 m

B) FFZ - Zege To Woretta The radius of the FFZ for the above link is

calculated by equation (1) using f = operating frequency of the link (2.4GHz)

λ = 3*10^8m/s/2.4*10^9Hz = 0.125m d1 = dist from Zege to the obstacle = 19.557km = 19557m d2 = dist from the obstacle to woretta =27.150km = 27150m

zege

BDR

ZegeTable-2. Path profile of hop between Bahir dar to zege and Zege to Worotta

Radio terminal sites Path profile Bahir dar Worotta Zege Latitude 110335’60

”N 11O55’60”N 11O42’6.

882”N Longitude 37O23’60

”E 37O41’24960”E

37O20’31.9605”E

Elevation 1798m 1850m 1950m Maximum hop distance

12.9km from Zege

46.7 km from Zege

46.7 km to Worotta &12.9km from Bahir dar


38

NCSTI-2012 F1=√(0.125*19557*27150)/(19557m+27150m) =√66,3715.568.75/46707m=√1,421.019=37.696m

From the above calculation values between the

two sites there is a clear line of sight and 60% of FFZ is free from any obstructions, the FFZ radius from Zege to Woretta is shown in the figure 7 below

Distance in km

Fig 7. Radius of FFZ from Zege to Bahir Dar

V. MICROWAVE LINK PATH ANALYSIS The path analysis (or link budget) is carried out

to dimension the link. What is meant here is to establish operating parameters such as transmitter power output, parabolic antenna aperture (diameter), and receiver noise figure [7]. It is a calculation involving the gain and loss factors associated with the antennas, transmitters, transmission lines and propagation environment, to determine the maximum distance at which a transmitter and receiver can successfully operate.

A) Propagation Losses Free-Space Loss Electromagnetic waves are

attenuated while propagating between two geometrically separated points. The free-space path loss model is used to predict received signal strength when the transmitter and receiver have a clear, unobstructed line-of-sight path between them. The attenuation is directly proportional to the square of distance and frequency and gives the free-space loss that represents most of the total attenuation caused by wave propagation effects. The frequency and distance dependence of the loss between two isotropic antennas is expressed in absolute numbers by the following equation.

LFSL = (4πd/λ) 2 = (4πdf / c) 2 [dB] (3)

Where d = dist between transmit and receive antennas (m) λ = operating wavelength (m) c = speed of light in vacuum (m/s) f = frequency (Hz)

If the frequency and the distance are expressed in terms of kilo meter and gaga hertz, the FSL are given by the following equation

LFSL = 92.45 + 20log (f) + 20log (d) [dB] (4)

Where f = frequency (GHz) d = LOS range between antennas (km)

The FSL is directly proportional to the length of the path and the operating frequency of the link, the relation is shown in the graph below.

Fig 8. FSL vs. distance and frequency

B) Receive Signal Level (RSL) The receive signal level (RSL) is the power level

entering the first active stage of the receiver. In most cases since the same duplex radio setup is applied to both stations the calculation of the received signal level is independent of direction. Receiver sensitivity threshold is the signal level at which the radio works satisfactory with the errors at a specified bit rate.

RSL can be calculated by the following formula

RSL = Pt –Lctx + Gatx – Lcrx + Garx – FSL –Rt [dBm] (5) Where RSL Rx (receiver sensitivity threshold) Pt = output power of the transmitter (dBm) Lctx, Lcrx = Loss (cable, connectors, branching unit) b/n Txr/Rxr and antenna (dB) Gatx = gain of transmitter antenna (dBi) Garx = gain of receiver antenna (dBi) FSL = free space loss (dB) Rt = rain attenuation (dB)

C) Rain attenuation An Electromagnetic wave, traveling in a given

direction through a rain cell, loses part of its power in that direction, as a result of absorption and scattering effects. In the impact with a raindrop, the total power lost depends on the "drop cross section", which is given by the sum of a scattering cross section and an absorption cross section. The drop cross section is a

F1

LOS

Woretta

zege


39

NCSTI-2012function of the drop radius and of the signal wavelength.

By integrating the power lost in the impact with a single raindrop to all the raindrops in a given volume (rain cell), the total loss produced within that rain cell can be estimated. As a result, the specific rain attenuation γ (dB/km) can be expressed, as a function of the rain rate R (in mm/h), by the following exponential formula

ᵞ =a. Rb (6)

where the parameters α and b are functions of the frequency and polarization. Values for α and b are available in the tabular form in a number of publications and it is shown in the table 3 below

Once the specific rain attenuation is found, the total rain attenuation is given by

A = ᵞ.L dB (7)

Where L is effective path length

Freq,GHz ah av bh bv

1 0.0000387 0.0000358 0.912 0.88 2 0.000154 0.000138 0.963 0.923 4 0.00065 0.000591 1.121 1.075 6 0.00175 0.00155 1.308 1.265 7 0.00301 0.00265 1.332 1.312 8 0.00454 0.00395 1.327 1.31 10 0.0101 0.00887 1.276 1.264 12 0.0188 0.0168 1.217 1.2 15 0.0367 0.0335 1.154 1.128 20 000751 0.0691 1.099 1.065 25 00124 0.113 1.061 1.03 30 0.187 0.167 1.021 1 Table-3 Specific attenuation coefficients

Lake Tana has rain intensity varies from 50 to 240 mm/hr, we can take the average, values 145 mm/hr.

For 2.9 GHz and 2.4 GHz frequencies a and b values are not specified, and hence interpolated from the table-3. For vertical polarization the values of

a = 0.000402 and b = 0.999

ᵞ = a. Rb = 0.000402*(145^0.999) = 0.058dB/km

The rain attenuation of the two paths is calculated as follows. For the first path Rain attenuation = 0.058 dB/km*12.90km = 0.748 dB For the second path Rain attenuation = 0.058 dB/km*46.7km = 2.7 dB

The relationship between frequency and rain intensity are shown in the graph below

Fig 9 Attenuation vs. rain intensity, for different signal frequencies, vertical (lower line) and horizontal (upper line)

polarizations

D) Fading and Fade Margin Fading is defined as the variation of the strength

of a received radio carrier signal due to atmospheric changes and/or ground and water reflections in the propagation path. Incorporating sufficient fade margin is the most important step in microwave link design. If the margin is too small, the link will be unstable, and as a result, sufficient availability of the link or quality of the provided services cannot be guaranteed. On the other hand, unnecessarily large margin makes the link more expensive (higher performance, larger and more expensive antennas) and increases the cost of creating the microwave link.

Fm= RSL - RS (8) Where Fm = fade margin RSL =received signal level RS = receiver sensitivity

VI .LINK BUDGET CALCULATION OF THE RECOMMENDED LINK

The recommended link has two hops. The link budget is calculated to know the reliability of the link to be designed by selecting different equipments with the appropriate rating. The specification of the equipments used at the transmitter and receiver site is shown in the table 3 below. General Specifications of equipments

Freq. Bands (GHz) 2.4

Tx/Rx Spacing (MHz) 55/116 Transmission Capacity 8E1 Modulation QPSK Freq. Synthesizers Step Size (KHz) 250

Freq. Stability (ppm) ± 10


40

NCSTI-2012

System Configuration (1+0), (1+1) Co-channel standby system, (1+1) Channel hot standby system

Power Supply (V) DC -48V ± 20%

Power Consumption 35

Transmit and Receive Power

Transmitting IF (MHz) 310

RF Output Power (dBm) 24

Receiving IF (MHz) 70 RF Input Level Range (dBm) -20 ~ -90

Receiver Threshold BER=1E-6 (dBm) 8 x E1 - 84

Antenna gain(dBi) 30 Coaxial cable

Type of cable

Operating frequency Loss in dB per 100 feet

LMR-900 2.4Ghz 2.9

Table-4 Specifications of equipments used at transmitter and receiver site

A) Link budget calculation from Bahir dar to Zege In order to calculate the link budget, we use the

specifications of the equipment given in table-4.

Free space loss (FSL) It is the first and most important step in link

budget analysis, and it depends on the operating frequency and the hop distance. Free space loss is calculated using equation (4) using, F = frequency (GHz) = 2.9 d = line-of-sight (LOS) range between Bahir Dar to

Zege (km) =12.90

LFSL = 92.45 + 20log (2.9) + 20log (12.90) [dB] = 92.45+9.24+22.2=123.95 dB

Received signal level (RSL) It is the amount of power reached at the receiver

unit, and it can be calculated by equation (5) using Pt = output power of the transmitter (dBm) = 24 Lctx, = Loss (cable, connectors, branching unit at Tx)

= 6.467 dB +0.025dB=6.49dB,becase the cable length at the transmitter side is 68m,it has 6.467dBloss and 0.025is connector loss.

Lcrx = Loss (cable, connectors, branching unit at Rx) 5.7 dB +0.025dB=5.725dB, because the cable length at the transmitter side is 60 m, it has 5.7dBloss and 0.025is connector loss between transmitter/receiver and antenna (dB)

Gatx = gain of transmitter antenna (dBi) =30 Garx = gain of receiver antenna (dBi) =30 FSL= free space loss (dB) =123.178

Rt = rain attenuation (dB) = 0.748 dB RSL=24dBm +30dBi -6.49Db +30dBi -5.725dB - 123.95 dB - 0.748 dB = -52.913 dBm

Fade margin It is the difference between the received signal

levels to the receiver sensitivity (receiver threshold). It is guaranty to the reliability of the link and calculated as follows by equation (7) using RSL=received signal level = -52.81 dBm RS= receiver sensitivity = -84dBm

Fm= RSL - RS = -52.913 dBm - (-84dBm) = 31.087 dB

B) Link budget calculation from zege to woretta In order to calculate the link budget, we use the

specifications of the equipment given in table-4.

Free space loss (FSL) FSL is calculated by equation (4) using

f = frequency (GHz) = 2.4 d = line-of-sight (LOS) range between Bahir dar

to zege (km) =46.7 LFSL=92.45 + 20log (2.4) + 20log (46.7) [dB] =92.45+9.24+21.48= 133.436 dB

Received signal level (RSL) It is the amount of power reached at the receiver unit, and it can be calculated by equation (5) using Pt = output power of the transmitter (dBm) =24 Lctx,=Loss (cable, connectors, branching unit at Tx) = 5.7 dB +0.025dB=5.725dB, because the cable

length at the transmitter side is 60 m, it has 5.7dBloss and 0.025is connector loss b/n transmitter/receiver and antenna (dB)

Lcrx = Loss (cable, connectors, branching unit at Rx) 5.7 dB +0.025dB=5.725dB, because the cable length at the transmitter side is 60 m, it has 5.7dBloss and 0.025is connector loss between transmitter/receiver and antenna (dB)

Gatx = gain of transmitter antenna (dBi) =30 Garx = gain of receiver antenna (dBi) =30 FSL = free space loss (dB) = 123.178 Rt = rain attenuation (dB) = 2.7dB RSL = 24dBm + 30dBi - 5.725dB + 30dBi - 5.725dB -

133.436 dB - 2.7dB = -63.586 dBm

Fade margin It assures the reliability of the link and calculated as follows, by equation (7) using RSL= received signal level = -51.393 dB RS = receiver sensitivity = -84dBm Fm = RSL – RS = -63.586 dBm - (-84dBm) = 20.414 dB


41

NCSTI-2012As we see from the link budget calculation of the

two links, the first hop has fade margin of 31.087 dB and the second hop has 20.414 dB. This indicates the link has the ability to provide guaranteed quality of service.

VII. CONCLUSION A direct standby microwave link between Bahir Dar

and Woretta as was initially proposed because of the importance of Woretta as economic growth centre. This direct link could not be materialized due Non-LOS propagation. An alternative proposal consisting of two hops from Bahir Dar to Zege and from Zege to Woretta has been considered and Path budget calculated. The result is found to be consistent with practice. Further by having repeater at Zege we have the facility of direct communication to Zege which does not exist presently.

The suggested microwave link is to be implemented and its performance and evaluation should be carried out in future if an alternate/standby link is required for important city of Woretta.

REFERENCES

[1] Ray Horak, Telecommunications and Data Communications Handbook, 2007 [2] Luigi Moreno, Point-To-Point Radio Link Engineering, 2001-2010 [3] MD. Rakib Al Mahmud and Zaigham Shabbir Khan

,Analysis And Planning Microwave Link To Established Efficient Wireless Communications, Blekinge Institute of Technology, September ,2009

[4] Harvey Lehpamer, Microwave Transmission Networks, Planning, Design and Deployment, Pg 106 McGraw – Hill Professional Engineering, 2004

[5] Michael F. Young, Planning a Microwave Radio Link, 2002.

[6] Roger L. Freeman, Telecommunication System Engineering, Fourth Edition, 2004.

[7] J. Frank Jimenez, Fundamentals of Radio Link Engineering, March 1999.

[8] Harvey Lehpamer, Microwave Point-to-Point Systems in 4G Wireless Networks and Beyond

[9] Constantine A. Balanis, Antenna Theory, Analysis and Design, Third Edition, 2005, John Wiley & Sons, Inc.


42

NCSTI-2012

Wireless Transceiver Testbed for Education and ResearchUsing SDR at BDU

Solomon H Gebreyohannes School of Computing and Electrical Engineering, Bahir Dar University, Bahir Dar, Ethiopia

e-mail: [email protected]

Abstract – Wireless transceiver testbed provides a complete working platform either to prove concepts, implement projects or innovate some wireless communication applications. This can be done through software defined radio (USRP/ GNU Radio). The main objective is to develop a uniquely defined but general wireless transceiver testbed which will be vital to help students to make their knowledge real in laboratories and researchers to undergo existing and new researches in research centers.

Key words – SDR, USRP, GNU Radio, python.

I. INTRODUCTION

Wireless transceiver testbed provides a complete research platform to actually implement wireless applications. It can collect many prototypes to a homogenous wireless communication platform. It mainly consists of transmitter and receiver. It can potentially used for wireless and wired systems. There are different types of wireless testbeds. Many of them are uniquely implemented in different universities and research centers. University of Surrey, UK, [1] developed campus wide wireless network testbed concentrating on practical research in mobile and fixed-wired data-communication systems based on mixed transmission technologies. 802.11 indoor wireless testbed is implemented in wireless network research center, University of California [2]. In the University of Utah [3], emulab, which refers facility and software system, is used to develop, debug, and evaluate wireless systems. Hydra, a wireless mutihop testbed [4] and WARP, a unified wireless network testbed for education and research [5] are also examples of well known wireless testbeds. Software Defined Radio, USRP peripheral and GNU Radio software, for example, can be used for such testbed purposes.

A software-defined radio system (SDR) is a radio communication system where components that have been typically implemented in hardware

(example: mixers, filters, amplifiers, modulators/ demodulators, detectors, etc) are instead implemented by means of software on a personal computer or embedded computing devices [6]. Universal Software Radio Peripheral, or USRP, is an RF device with a simple design that allows for a wide range of SDR related uses. Since the USRP is not a stand-alone SDR, its architecture requires signal processing functions which have to be done on other host devices via USB 2.0. it allows one to create a software radio using any computer with a USB 2.0 or Gigabit Ethernet port. There are two types, type 1 and type 2, of the USRP family. Various plug- on daughterboards allow the USRP and USRP 2 to be used on different radio frequency bands. Daughterboards are available from DC to 5.9 GHZ at this time. The entire design of the USRP family is open source. USRP work with GNU Radio, a free-software (open source) framework for the creation of software defined radios. GNU Radio works on Linux, Windows, Mac OSX, FreeBSD and NetBSD [7],[8]. GNU Radio is a free software development toolkit that provides the signal processing runtime and processing blocks to implement software radios using readily-available, low-cost external RF hardware and commodity processors. It is widely used in hobbyist, academic and commercial environments to support wireless communication research as well as to implement real-world radio


43

NCSTI-2012systems. GNU Radio applications are primarily written using the Pyhon programming language, while the supplied, performance-critical signal processing path is implemented in C++ using processor floating point extensions. Thus, a developer is able to implement real-time, high-throughput radio systems in a simple-to-use, rapid-application-development environment. While not primarily a simulation tool, GNU Radio does support development of signal processing algorithms using pre-recorded or generated data, avoiding the need for actual RF hardware [9]. USRP being closely coupled with the GNU Radio software radio framework forms a flexible and powerful platform.

II. SYSTEM ARCHITECTURE

A basic SDR system may consist of a personal computer equipped with a sound card, or other analog-to-digital converter, preceded by some form of RF front end. Significant amounts of signal processing are handed over to the general-purpose processor, rather than being done in special purpose hardware. Such a design produces a radio which can receive and transmit widely different radio protocols (signal formats) based solely on the software used. Here, the main idea is to use USRP/ GNU Radio combination to make SDR system for wireless transceiver. The general setup, perhaps, looks like Figure 1.

Figure 1 Wireless transceiver setup

Relative to other systems, SDR doesn’t usually need hardware configuration nor dedicated hardware. The necessary materials are the peripheral (like USRP), PC, antenna and connectors. They should be connected and arranged in the order shown in Figure 1.

The second, of course the main task, is the software consideration and installation matter and make it ready for use. Theoretically, any high level programming language can be used for the system though the difficulty among them varies significantly.

III. BENEFITS OF THE SYSTEM

There are two main advantages in implementing and using this system, especially for SCEE, BDU. First, any kind of wireless application can, theoretically, be implemented in this platform. Some applications have been tested in Bahir Dar University and explained in section IV. Second, it can provide areas of study and research. In this paper, four main categories are selected; such as, antennas, peripheral, computer programming and performance evaluation of the system. Antennas act as a bridge to connect the system (strictly, daughterboards) and RF world (channel). The study may range from knowing each antenna types, which are ready for sale for each frequency ranges, to design a specific one. In any way, simple or complex, there are lots of things to be considered and is purely an engineering task. As an example, there are different antenna types in [8] which are ready for use. And also, there is a prototype antenna for beacon satellite receiver in [10]. Peripheral is the main part of the system. It seems it is difficult to study and perhaps, to improve, but possible. Still advanced, the USRP motherboard can be studied and/or improved or may be changed. Ettus [8] developed another board (other USRP series) which can support MIMO capabilities. In [5], there is a new successful independent development of the board. The main advantage of SDR in general is to use software instead of dedicated hardware here and there. This task can be categorized in a couple of groups. One is to make the computer ready for use. This includes investigating specifications of the PC, installing the pre-requisites, required packages, specifying the platform and any other needed. In short, it is to make the computer ready to run specific programs which are in turn useful for different applications. Second, it is to write the codes of different wireless applications and to test them. GNU Radio library [9] itself has many applications like FM transmitter/ receiver, AM


44

NCSTI-2012transmitter/ receiver, spectrum analyzer and other communication toolboxes. Other people are also interested in developing applications. Lee [11], for example, designed software-defined radar prototype using GNU Radio. Yessika [12] discussed Global Positioning System (GPS) signal acquisition performance and procedures. Scaperoth [13] described an interface between a Cognitive Engine and SDR platform which modifies (i.e., configures) the radio’s operation. Recently, Bruhtesfa [16] used software defined radar to characterize human activity in order to achieve wide acceptance in protecting one’s privacy instead of using cameras. To choose the USRP/ GNU Radio combination as a good alternative, the performance should be studied and compared with the existing other systems. There may be variations of considering the evaluation and comparison, mostly prior to the objective of the specific project or research. Eavluating a GNU Radio testbed [19] is an example of checking how well GNU Radio is used for wireless testbed.

IV. EXISTING AND FUTURE DEVELOPMENTS AT BDU

In Ethiopia, USRP/ GNU Radio technology is first introduced at Bahir Dar University. Satellite receiver is implemented and working at Physics laboratory, BDU. Performance of USRP is tested and an application is developed by the same author – “GNU Radio Physical Layer Implementation for the Study of Performance and Development of Applications”, published on proceedings of 5th Scientific Conference on Electrical Engineering by ESEEA [20]. FM and TV signals reception and (live) FM transmission were successfully done in SCEE, BDU. There is also a plan to build private GSM network and satellite receivers to check as to where the system works on one hand and to utilize it on the other hand.

V. CONCLUSION

USRP and GNU Radio are better ways either to develop an application and/ or undergo experiments as proof of concepts, if not; it helps to innovate new developments.

REFERENCES

[1] Center for Communication Systems Research, University of Surrey, UK. http://www.ee.surrey.ac.uk/CCSR/home[2] University of California, http://www.ucr.edu/[3] Emulab, http://www.emulab.net/[4] Hydra – “A Wireless Multihop Testbed”, http://www.ece.utexasedu/~rheath/research/prototyping/mimoadhoc/[5] WARP, IEEE Computer Society, http://www.computer.org/portal/web/csdl/doi/10.1109/MSE.2007.91[6] M. Dillinger, K. Madani, N. Alonistioti, “Software defined radio: architectures, systems and functions”, 2003 [7] D. A. Scaperoth, “Configurable SDR Operation for Cognitive Radio and Applications”, Virginia Polytechnic Institute and State University, 2007 [8] Matt Ettus. Ettus Research LLC. URL:http://www.ettus.com, 2011. [9] Eric Blossom. “GNU Radio’, URL: http://gnuradio.org/redmine/wiki/gnuradio.[10] Mamoru Y. “Research for Sustainable Humanosphere”, Kyoto University, 2009. [11] L.K. Patton, “A GNU Radio Based Software-Defined Radar”, Wright State University, 2007. [12] Yessika Yaneth Gutierrez, “GNU GPS Radio Development with Advanced Acquisition”,U of Texas, San Antonio, 2009. [13] D.A. Scaperoth. “Configurable SDR Operation for Cognitive Radio, Applications using USRP”, Virginia P.I. University, 2007. [14] N. Manicka, “GNU Radio testbed”,University of Delaware, 2007. [15] E.B. Mikkelsen “ The Design of a Low Cost Beacon Receiver System using Software Defined Radio”, Norwegian U, 2009. [16] B.E. Godana, “Human Movement Characterization in Indoor Environment”,Delft University , Netherlands, 2009. [17] Emulab GNU Radiohttp://www.emulab.net/tutorial/docwrapper.php3?docname=gnuradio.html, 2007.


45

NCSTI-2012[18] K. Madke, S. Choi. “A Flexible MAC/PHY Multihop Testbed”, IEEE Vehicular Technology Conference, 2007 [19] S. Valetin, H. Malm, and H. Karl, “Evaluating the GNU Software Radio Platform”, U of Paderborn, , 2006. [20] Solomon H., “GNU Radio Physical Layer Implementation for the Study of Performance and Development of Applications”, 5th Scientific Conference on EE by ESEE Association, 2011.


46

NCSTI-2012


Sanjay Singh School of Plant Sciences College of Agriculture and Environmental Sciences

Haramaya University, Haramaya Ethiopia E-mail : [email protected]

Abstract - The simulation in the laboratory of in situ mode of field moisture stress which originates from top of the soil and travels down the root-zone, has posed a formidable challenge to a range of scientists including plant biologists, agronomists, plant nutritionists, plant biotechnologists, horticulturists, ecologists, environmentalists, meteorologists, hydrologists, irrigation engineers etc alike. The invention of a technique for the uptake of water, nutrient, water stress and salt stress studies which could overcome the deficiencies of and the difficulties in field, lysimeter, pot or aerosol investigations for precise studies of root function in relation to water absorption and plant water-use, nutrient uptake and its use-efficiency, salt uptake and its impact on crop production, would be a great leap forward in experimental plant biology, agriculture, horticulture, forestry, hydrology and environmental science. We have invented an automatic potometer technique which is simple, easy to operate, inexpensive, non-cumbersome and very closely resembles in situ field moisture stress development under laboratory conditions. This technique is novel as it is superior to all those of the past and presently in the use for water stress studies. This can be used to measure the whole-plant response including growth parameters, plant water status and water-use, CO2 exchange, stomatal conductance, photosynthesis, nutrient and salt uptake, metabolic events, osmo-regulation, tissue-specific gene expression and hormone signalling etc accurately with precise degree and duration of plant stress in the laboratory. Information so-gathered might be useful in modelling crop-water stress effects aimed at developing dynamic cropping system for efficient water-use by plants in various dryland environments, scheduling irrigation, monitoring plant sustenance and drought tolerance, determining plant salt tolerance and nutrient-use-efficiency, addressing environmental issues and unravelling physiological and metabolic mechanisms to give way to creating transgenic or other new variants of food, fiber, fuel, forage, fruit and forest plants having potential to fight against drought, salt accretion and nutrient deficiencies across the globe.

Key words- Cell viability, computer-guided operation, environmental factors, field moisture stress, leaf rolling, mercury, Oryza sativa L., osmotic solution, potometer, proline, transpiration, water extraction, water potential

I.INTRODUCTION

Water is elixir of life, be it plant or animal, prokaryote or eukaryote. Although, it is an essential constituent of life but its abundance or scarcity, both are harmful for terrestrial organisms [1]-[4]. The present era is, however, witnessing serious water scarcity in all the tracts of the world, i.e. arid, semi-arid, and humid regions, as evidenced by continually receding ground water table with the rivers, lakes and water reservoirs losing their serving capacity, hence turning dry [5]-[7]. This water shortage is more than likely to swell up in future to take the shape of alarming water crisis which might eventually prove devasting under threats accruing from global warming and climate change [8], [9]. That is why, it has become a serious cause of concern to agriculturists, plant growers and planners all over the world [10]-[12]. Looking at the grave agricultural water situation in the offing, it is imperative to call for effective research propositions aimed at targeted plant responses to diminished water supply with a view to easing its harmful impact on world food production and food security [4], [13]-[15]. Although, a large body of information on crop water

deficit exist in the literature, but they largely, if not wholly, suffer from one kind of lacunae or the other [16], [17]. The field and lysimeter studies among them, for example, though provide useful results but are often time-consuming, mostly uncontrollable and difficult to monitor variables encountered therein [18]-[20]. On the other hand, investigations of pot culture and the like-ones based on water stress created by withholding the supply of water share the truth of inheriting restrictions on longitudinal root growth and its proliferation imposed by pot size and elevated root temperatures particularly when grown either in the open air or glasshouse [21]. A step next to it is subjecting the root system to the osmoticum [22] or root aerosols [23] which run the risk of being wholly and abruptly exposed to only a single magnitude of root-zone water stress at a time. All of these conditions are far from field situations wherein moisture stress originates from the top of the soil and travels down the root-zone slowly whose initiation, intensification and progression mainly depend upon weather conditions, crop coverage, moisture holding and retaining capacity of soil and shoot-root characteristics of plants (Fig.1, 2) [24].


47

NCSTI-2012

The simulation in the laboratory of in situ mode of field moisture stress which originates from the top of the soil and travels down the root-zone has posed a formidable challenge to a range of scientists including plant biologists, agronomists, plant nutritionists, plant biotechnologists, horticulturists, ecologists, environmentalists, meteorologists, hydrologists, irrigation engineers etc alike. However, the research team headed by Professor T.N. Singh at Faizabad in India has succeeded, though partially, in creating plant water stress under laboratory conditions but the osmotic stress development in the rhizosphere being stationary in nature was not quite similar to that of field moisture stress but very close to the situation that exists in the field at a particular time during soil moisture depletion (Fig. 3) [25]. Therefore, the lack of a suitable laboratory technique comparable to that of slow and progressive in situfield moisture stress development originating from the top of the soil and travelling down the root-zone, is inevitable hurdle for the precision studies on root function in relation to water absorption and its use, uptake of nutrients and salts and their impact on germinating seeds, plant seedlings and saplings, plant sustenance and survival, and adaptation of crops to limited water availability [13], [15], [26]-[30]. The accurate information on plant responses to in situsoil-water scarcity in the root-zone as when and how the perception of water stress translates into a plant condition that imparts detrimental impact, hardly needs to be emphasized for the fuller understanding of both the intrinsic and extrinsic control mechanisms of plant growth, development and production [3], [4], [13], [15]. The invention of a technique which could overcome the deficiencies of and the difficulties in field and lysimeter investigations and pot culture and

Fig. 1. Diagrammatic representation of a typical radial and downward slow and steady expansion of root dry zone as a consequence of progressive soil moisture depletion by maize plant (adopted from Singh and Singh, 2000).

aerosol studies at whole-plant level would be a greatleap forward in the field of experimental plant biology, agriculture, hydrology and environmental science. We have invented an automatic potometer technique which is simple, easy to operate, inexpensive and non-cumbersome and very closely resembles in situ field moisture stress in the laboratory. This technique is novel as it is superior to all those of the past and presently in the use for water stress studies which enables measurements of whole plant responses to continually increasing water stress in space and time by subjecting the intact plant roots to progressively graded osmotic solutions, a situation of water stress development that often exists under field conditions. In this communication, we have presented the results of changes in leaf rolling and leaf water potential as physiological indicators and proline accumulation and triphenyl tetrazolium chloride reduction test for cell viability as metabolic indices of water stress in intact rice plant (Oryzasativa L.) experimented under graded osmotic solutions in especi

Fig. 2. Pattern of radial and downward expansion of root dry zone and soil moisture depletion by maize plant as affected by variable dates of sowing in summer (adopted from Singh and Singh, 2000).

ally-designed automatic potometer in the laboratory.

II.MATERIALS AND METHODS The newly fabricated automatic potometer consists

of three portions joined together through the ground jointing (Fig. 4). The left side portion is made of U-shaped glass tubing with its one arm measuring 26 cm equipped with an air passage at the top below the mouth and the other arm is 24 cm long and 1 cm in diameter each. The short arm is connected to the middle portion of the apparatus which is 12 cm long arm with excessive length of a flexible PVC joint to enable the adjustment of any desired degree of


48

NCSTI-2012elevation angle between this arm and the short arm of U-tube. A thistle-like structure to be loaded with mercury therein is fitted on the mouth of the longest arm of the U-shaped tubing and the third (last) portion having a reservoir-like structure is joined to the end of the middle portion. The test plants were anchored in the solution reservoir through the rubber cork. The angle of elevation between the short arm of the U-shaped tubing and the middle portion of the apparatus consisting of sidearm was 1000 in order to avoid intermingling under gravity of graded series of osmotic solutions, yet not to cause any anti-geotropic effect on roots, as large proportions of it in cereals and other plants do grow radially and proliferate upwardly under field conditions [24], [27], [29]-[31]. The whole apparatus was clipped at five places to keep it in the place on a swinging wooden board when loose, to make the loading of osmotic solutions easy by raising it obliquely upward. Subsequently, the obliquely-raised side of the wooden board was pulled down and brought parallel to the plane of the table. The experiment was conducted with rice at the Crop Research Station, Masodha, Faizabad, India during rainy season in 2007. The first step was to fill the mercury in the U-shaped tubing from the farthest side, first up to the height of 10 cm only followed by loading from the second portion of 5 graded polyethylene glycol (MW 4000) solutions of 1, 0.75, 0.50, 0.25 and 0.12 M concentrations respectively. The next step was to load the osmotic solution of the highest concentration in the remaining length of the short arm of U-shaped tubing followed by gently loading of 2.5 ml each of the remaining graded osmotic solutions, one after another in decreasing order of concentration, by an automatic syringe pipette through a narrow PVC tubing of desired length attached to its end. The loading of solutions can also be accomplished by safety pipette filler or dropper. To the open end of the middle portion was then attached the third portion of the apparatus having a reservoir-like structure of 6x2.5x1.5 cm length, breadth and height respectively, which also had an extended sidearm of 16 cm length having a side feeding hole at the end with the constricted open distal tapering end as well. The whole structure including U-shaped tubing connected with the middle portion through a flexible PVC joint and the reservoir-like assembly accompanied by its extended sidearm were graduated to the required marks. Subsequently, one-month-old five test plants intact with six leaves and roots of medium duration hybrid rice cv. NDRH-2 maturing in about 135 days and bred for irrigated ecosystem grown in Hoagland nutrient solution [32], were taken out and their side-tillers which were beginning to emerge, were removed and anchored in the reservoir-like structure of the apparatus pushing the roots with a stream of gently forced water towards the horizontal extended arm having tapering end. Since we were concerned with the situation of scanty water supply to the plants by using osmoticum, i.e. the development of plant

water stress, the role of the amount of intact roots in the maintenance of plant water balance under such conditions was also examined by removing half the numbers of roots against those having all the roots intact. This portion was then connected to the open end of the middle portion of the apparatus and was filled with the nutrient solution either through the feeding tube made in the cork or feeding hole made in the extended sidearm. The further addition of mercury into the U-shaped tubing then commenced drop by drop at pre-determined rate under gravity from the thistle-like container fitted in the open mouth of the U-shaped tubing. The pre-regulated rate and size of mercury droppings kept on pushing forward @ 2.5 cm day-1 the graded osmotic solutions [22] towards roots beginning from root-shoot junction of the test plants. With the passage of experimental time, the entire root system was ultimately engulfed and subjected on the way to continually graded osmotic solutions increasing in osmotic potential (more negative in values) acropetally with time, i.e. at last, the most concentrated osmotic solution being towards root-shoot junction and the least concentrated one towards the tapering end of the potometer. This arrangement of osmotic solutions in the rhizosphere provided a situation that often exists under conditions of in situfield moisture stress originating from top of the soil advancing slowly downward [24], [25]. The existence of various bands of graded osmotic solutions accompanied by conspicuous absence of an abrupt change thereof forming inter-phase between any two neighbouring solutions had been earlier confirmed by using dyes of various colours for each band. The operative microclimate components constituted temperature which fluctuated between 28-32°C, the relative humidity varied between 77-83% and feeble air circulation produced by ceiling fan located at the height of 2 metres and a radial distance o

tions and the results were analysed statistically [37].

III.RESULTS AND DISCUSSION

s the osmotic potential kept decreasing through its

f 3 metres from the test plants. The symptoms of plant water stress were gauged

daily for six consecutive days by assessing physiological indices such as the rolling behaviour [33] of the fully developed second leaves from the top and the water potential [34] of the same leaves.However, the metabolic indices such as prolineaccumulation [35] and cell viability by tetrazolium chloride reduction test [36] were determined in the pooled samples of the remainder of the leaves. Experiments were conducted in three replica

The plant traits studied using this laboratory technique clearly demonstrated that the pattern of stress-induced changes in leaf rolling, leaf water potential, proline accumulation and cell viability (Table 1) were quite similar to the results normally obtained from field moisture stress studies [24]. A


49

NCSTI-2012

TABLE 1 Physiological and metabolic symptoms of plant water stress resulting from continually graded osmotic solutions in automatic potometer resembling in situ field moisture stress [Mark the symbolic osmotic stress front advancing towards roots turning turgid rice plant (cv. NDRH-2) into flaccid condition with the rising levels of mercury

U-shaped tube]. in

Symptoms of t water stress Symbolic movement Day of Root fractio Physiological indices Metab lic indicessta

plann o

tus of osmotic experiment exposed to Leaf rolling Leaf water Proline Cell ili

osmotic stress (score) potential(-bar)(mgg-tysolutions into the

wt__________________________________________________________ ________ _ ______ ______________

Half roots 1 4.0 0.30

2nd All roots 1 5.1 0.33


3rd All roots 2 8.1 0.70

Half roots 2.3 9.0 0.70 1.

4th All roots 3.3 12.0 1.12


5th All roots 4.6 16.0 2.05 0.

Half roots 5 16.8 2.06 0.41

6th All roots 5 19.8 3.05 0.23

Half roots 5 19.8 3.10 0.22

)(ODat530nm) root-zone ___ __ _ _

1st All roots 1 4.3 0.30

.31

1.32

1

1.24

1.24

1.17

12

0.87

0.87

43

0.9

LSD0.05 0.2 0.22

0.09

1000

5

Osmotic stress yet toreach the r t-zoneoo

Graded osmotic solutions a vancing

towards roots under mercur pressuredial

2 15 4 3

Nut solutionrient

Test p tion)lant (turgid condi

Test plant

100

5

Osmoti ress half-wr h routec st ay

th oug its

Gancingpresraded osmotic solutions adv

towards roots under mercurial s reu

5 4

(intermediate condition)

Nutrient solution

23Roots

1000

5

Osmotic gtravel

stress havinled full le gth of its routen

Graded osmotic solutions adv g

towards roots under mercurial pressureancin

5

N rient solutionut

Test plant (flaccid condition)

Roots2 1

5 4 3


50

NCSTI-2012continued movement towards the root-shoot

junction and beyond it subjected the plants to increasing water stress. This was evident from the leaf rolling behaviour (score 5) and the decline of leaf water potential to about -20 bars at the end of sixth day. Proline accumulation also increased proportionately from 0.32 mg to 3.0 mg g-1 dry wt followed by comparable loss of cell viability due to plant water stress created by graded osmotic solutions [36]. Interestingly, variation in the root-shoot (leaf area) ratio introduced by removal of half the numbers of roots did not materially alter the plant responses to water stress. Thus, these results conclusively suggest that the automatic potometer technique of simulating plant water stress in the laboratory is quite similar and comparable to that in situ field moisture stress which guarantees precise, accurate and desired magnitude of stress development in space and time, overcoming the deficiencies of and difficulties in field, lysimeter, pot and aerosol studies [18]-[21], [23], [25]. Our laboratory technique of simulating water stress by employing graded series of osmotic solutions is indeed comparable to in situ field moisture stress as water extraction by growing plants and its evaporation from surface under field conditions (evapotranspiration) result in soil drying originating from its top layer creeping down the root-zone which causes plant water stress leading to retardation or arrest of crop growth and reduction in production (see Fig. 1, 2). Likewise, failure of irrigation water supply or diminished water availability or irregular rainfalls during wet season inviting dry spells which all accelerate the process of soil drying also result in development of soil matric potential at 5-10 cm depth falling at times to less than -15 bars, the permanent wilting point [24], [28]. Such values of matric potential in drying soils within or beyond the wilting point imply that the risk of mortality from desiccation for germinating seeds, young seedlings and saplings, and retardation or arrest of crop growth, increases with advancing stress until it is overcome by either irrigation water or resumption of rains [3], [5], [29], [38], [39].

In conclusion, this technique is novel as it is superior to all those of the past and presently in the use for water stress studies conducted at whole-plant level either in the field or lysimeters or pots under laboratory or growth cabinets. Further, most importantly, it has potential to revolutionise researches on uptake of water, nutrient and salt with their physiological and metabolic consequences of immense significance within shortest time possible. It can be used for investigating one or all growth parameters, either at vegetative or reproductive stage, with an array of physiological responses such as plant water status and water use, CO2 exchange, stomatal conductance, photosynthesis, uptake of nutrient and salt, metabolic events, osmo-regulation, tissue-specific gene expression and hormone signalling etc in plants under water stress of any desired degree and duration in space and time. In this technique both the

degree and duration of plant water stress can be controlled by regulating the rate and size of mercury droppings in the U-shaped tubing and by varying the concentrations of osmotic solutions used. It is simple, easy to operate, inexpensive, non-cumbersome and rapid in functioning which establishes its superiority over those in situ field moisture stress studies. Further, refined studies could now be possible to conduct in the laboratory precisely and accurately without sacrificing for the time factor waiting in the hope of desired soil moisture or experimental weather conditions to arrive for conducting the experiments free from troubles of the kind normally encountered in field investigations [20]. Thus, this technique very closely resembles in situ field moisture stress development in the potometer. The size and capacity of the apparatus can be either shortened or enlarged as per one’s needs. Further, the effect of re-watering or relieving of stress can also be studied in this system by either partially or wholly flushing out the osmotic solutions via the tapering end or feeding hole by injecting water through the feeding tube in the rubber cork as happens in situ field moisture stress removal following irrigation or rainfall [5]. To make this technique far more efficient, precise, accurate and time-saving we are attempting to control the functioning of this apparatus on the principle of computer-guided mode of operation tuned to continuous monitoring and amalgamation of the effects of a series of environmental factors such as heat load, radiation and irradiance, vapour pressure, relative humidity, wind velocity etc, all affecting concurrently the transpiration by plants subjected to osmotic stress.

This technique might be shared by a range of scientists to study plant processes and their behaviour aimed at developing dynamic cropping system for efficient water-use by crop species in various dry land environments [3], [14], [28], scheduling irrigation [5], [40], monitoring plant sustenance [13], determining salt uptake by plants and its impact [41], [42], measuring nutrient-use-efficiency [27], addressing environmental issues [8], [9] and unravelling physiological and metabolic mechanisms [4], [13], [15], [43] to give way to creating transgenic or other new variants of food, fiber, fuel, forage, fruit and forest plants having potential to fight against drought and salt accretion across the globe [44]-[46].


51

NCSTI-2012

Fig. 3. Diagrammatic representation of the apparatus used for simulating root-zone water stress by using stationary graded osmotic solutions of polyethylene glycol (MW 4000) (adopted from Singh et al. 2011).

Fig. 4. Diagram of an automatic potometer for simulating continually graded osmotic stress in the root-zone of test plants resembling in situ field moisture stress development.


52

NCSTI-2012

REFERENCES

[1] F. Yamamoto, T. Sakata, and K. Terazawa, “Physiological, morphological and anatomical responses of Fraxinus mandshurica seedlings to flooding,” Tree Physiology, vol. 15, pp. 713-719, 1995.

[2] T.T. Kozlowski Flooding and Plant Growth (ed.) Academic Press, Orlando, USA. 1985.

[3] T.C. Hsiao, J.C. O'Toole, and V.S. Tomar, “Water stress as a constraint to crop production in the tropics.” in Priorities for Alleviating Soil-Related Constraints to Food Production in the Tropics (ed. anonymous),International Rice Research Institute, Los Banos, Philippines. 1980, pp. 339-369.

[4] L.G. Paleg, and D. Aspinall The Physiology and Biochemistry of Drought Resistance of Plants (eds.) Academic Press, Sydney, Australia, 1986. 2011. [5] R.J. Lascano, and R.E. Sujka, Irrigation of Agricultural Crops (eds), 2nd edition. Agronomy Monograph 30, ASA-CSSA-SSSA, USA, 2007. and Plant Water, Vol. 1-3, Utah Agr. Exp. Sta., Utah [6] T.V. Cech, Principles of Water Resources: History, Development, Management, and Policy (ed.) John Wiley & Sons Ltd., New York, USA, 2008.

[7] M. Power, “Peak water: Aquifers and rivers are running dry, how three regions are coping” WiredMagazine, vol. 16, no. 5, online May 2008. Plants and Soils San Diego, California, Academic [8] J.T. Houghton, L.G. Filho, B.A. Callander, N. Harris, A. Kattenburg, and K. Maskell, “Climate Change 1995: The Science of Climate Change” (eds.) in Intergovernmental Panel on Climate Change,Cambridge University Press, UK, 1996.

[9] R.W. Spencer The Discovery of Global WarmingHarvard University Press, USA, 2003.

[10] M.S. Swaminathan Sustainable Agriculture: Towards Food Security Konark Publishers Pvt, New Delhi,

India, 1996. [11] P. Dasgupta, “The economics of food,” in Feeding a World Population of more than 8 Billion: A Challenge to Science (Waterlow, J. C., ed.). Oxford University Press, UK 1997, pp. 19-36. [12] L.T. Evans Feeding the Ten Billion: Plants and Animals Cambridge University Press, UK (HB871.e.92), 1998. [13] H.G. Jones, and T.J. Flowers Plants Under Stress: Biochemistry, Physiology and Ecology and Their Application to Plant Improvement Cambridge University Press, UK, 1989. [14] O. Ito, J.C. O’Toole, and B. Hardy, Genetic Improvement of Rice for Water-Limited Environments (eds.) International Rice Research Institute, Las Banos, Philippines, 1999. [15] R.A. Richards, “Physiological traits used in the breeding of new cultivars for water-scarce environments”, Agric. Water Mgmt., vol. 80, pp. 197- 211, 2006. [16] H.D. Barrs, “Determination of water deficits in plant tissues”, in Water Deficits and Plant Growth, Vol. 1 (T. T. Kozlowski, ed.). Academic Press, New York, USA, 1968, pp. 236-347.[17] P.J. Kramer, “Measurement of plant water status: historical perspectives and current concerns”, Irrig. Sci., vol. 9, pp. 275-287, 1988. [18] K.W. Finlay, and G.N. Wilkinson, “The analysis of adaptation in a plant breeding programme”, Aust. J. Agril. Res., vol. 14, pp. 742-754, 1963. [19] S.A. Eberhart, and W.A. Russell, “Stability parameters for comparing varieties”. Crop Sci., vol. 6, pp. 30-40, 1966. [20] R. Knight, “The measurement and interpretation of genotype-environment interactions,” Euphytica, vol. 19, pp. 225-235, 1970. [21] A.D. Schneider, T.A. Howell, and J.L. Steiner, “An

evapotranspiration research facility using monolithic lysimeters from three soils,” Appl. Engg. Agri., vol. 9, pp. 227-232, 1993. [22] B.E. Michel, and M.R. Kauffmann, “The osmotic potential of polyethylene glycol 6000,” Plant Physiol., vol. 51, pp. 914-916, 1973. [23] R. Greenwald, M.H. Bergin, J. Xu, D. Cohan, G. Hoogenboom, and W.L. Chameides, “The influence of aerosols on crop production: a study using the CERES crop model,” Agricultural Systems, vol. 89, pp. 390-413, 2006. [24] Sanjay Singh, and T.N. Singh, “Response of maize crop to early and late summer sowing conditions,” Indian J. Plant Physiol., vol. 5 (NS), pp. 307-313, 2000. [25] Sanjay Singh, T.N. Singh, and J.S. Chauhan, “A New Laboratory Technique for Creating Water Stress,” Journal of Crop Improvement, vol. 25, pp. 769–778,

[26] R.J. Hanks, and R.W. Brown, Proceedings of the International Conference on Measurement of Soil

State University, Logan, UT, USA, 1987.[27] G.J.D. Kirk Rice Roots: Nutrient and Water Use (ed.) International Rice Research Institute, Los Banos, Philippines, 1994.[28] P.J. Kramer, and J.S. Boyer Water Relations of

Press, New York, USA, 1995.[29] K. Kikuzawa, and H. Koyama, “Scaling of soil water absorptions by seeds: an experiment using seed analogues,” Seed Sci. Res., vol. 9, pp. 171-178, 1999.[30] G. Singh, T.N. Singh, and Sanjay Singh, “Trinodal rooting in rice: a new parameter on drought resistance,” Indian J. Plant Physiol., vol. 4 (NS), pp. 232-235, 1999. [31] Y. Waisel, A. Eshel, and U. Kafkafi Plant Roots: The Hidden Half (eds.) Marcel Decker and Co., New York, USA. 1996. [32] D.R. Hoagland, and D.I. Arnon, The water-culture method for growing plants without soil. Univ. Calif. Agric. Expt. Sta. Circ., USA, vol. 347, 1938. [33] Sanjay Singh, and T.N. Singh, “Morphological, chemical and environmental factors affecting leaf rolling in rice during water stress,” Indian J. Plant Physiol., vol. 5(NS), pp. 136-141, 2000. [34] E.B. Knipling, “Measurement of leaf water potential by the dye method,” Ecology, vol. 48, pp. 1038-1041, 1967. [35] L.S. Bates, R.P. Waldren, and I.D. Teare, “Rapid determination of free proline for water stress studies,” Plant and Soil, vol. 39, pp. 205-207, 1973. [36] T.N. Singh, L.G. Paleg, and D. Aspinall, “Stress metabolism. III. Variations in response to water deficit in barley plant,” Aust. J. Biol. Sci., vol. 26, pp. 57-76, 1973.[37] J.C.R. Li, Statistical Inference (Vol.I). Edwards Brothers Inc., Ann Arbor, Michigan, USA, 1967. [38] E.T. Nilsen, and D.M. Orcutt The Physiology of Plants under Stress: Abiotic Factors John Wiley & Sons Ltd, New York, USA, 1996. [39] A.B. Elizabeth, “Plant responses to water deficit,” Trends Plant Science, vol. 2, pp. 48-55, 1997. [40] C. Giulivo, and G. Zerbi, IV International Symposium on Water Supply and Irrigation in the Open and Under Protected Cultivation (eds., available in ActaHort CD-rom format only). ISHS Acta Horticulturae, Padova, Italy, vol. 228, 1988. [41] E. Blumwald, and A. Grover, “Salt tolerance,” in Plant Biotechnology, Current and Future Applications of Genetically Modified Crops (N. Halford, ed.). Agritech Publications, New York, USA, 2006, pp. 206-224. [42] D. Aspinall, “Metabolic effects of water and


53

NCSTI-2012 salinity stress in relation to expansion of the leaf surface,” Aust. J. Plant Physiol., vol. 13, pp. 59-73, 1986. [43] H.J. Bohnert, and R.G. Jensen, “Strategies for engineering water stress tolerance in plants,” Trends Biotech., vol. 14, pp. 89-99, 1996. [44] B.C.Y. Collard, C.M. Vera Cruz, K.L. McNally, P.S. Virk, and D.J. Mackill, “Rice molecular breeding

laboratories in genomics era: current status and future considerations,” Intl. J. Plant Genomics, pp. 1-25, 2008. [45] E. Pennisi, “The blue revolution, drop by drop, gene by gene,” Science, vol. 320, pp. 171-173, 2008. [46] D.A. Lightfoot, “Blue revolution brings risks and rewards,” Science, vol. 321, pp. 771-772, 2008.

.

Bioconversion of Parthenium Hysterophorus into Valuable Organic Manure

54

NCSTI-2012


Hiranmai Yadav, R1. , Eyasu Mekonnen2 and Vijayakumari, B3

1 School of Natural resource management and Environmental Sciences, College of Agriculture, P.O. Box #138,

Haramaya University, Dire Dawa, Ethiopia Email: [email protected]

2 Department of Botany, Avinashilingam University For Women, Coimbatore – 641 043,Tamil Nadu, India

Abstract:- Organic farming is defined as production system which avoids or largely excludes the use of synthetically compounded fertilizes, pesticides, growth regulators and livestock feed additives to the maximum extent feasible. The main aims of organic farming are to achieve food and nutritional security, to encourage long term maintenance of soil fertility, crop productivity and soil health, to encourage and enhance biological cycles within the farming system, involving microorganisms, soil flora, soil fauna, plants and animals. This also help in conserving, developing and utilizing the natural resources in the efficient way, to recycle/reuse biomass materials either on farm or elsewhere in order to minimize pollution that may result from agricultural practices. Biodegradable material of microbial, plant or animal origin shall form the basis of the fertilization programme in organic farming. Parthenium hysterophorus Linn, an exotic weed belongs to the family compositae spread alarmingly like a wild fire with devastating effect on environment, health and agriculture. The species Parthenium hysterophorus also known as congress weed or congress grass, has become common in India, Australia and parts of Africa. In some areas, outbreaks have been of almost epidemic proportions, impacting crop production, livestock and human health. Parthenium hysterophorus has, in a very short time period, emerged as one of the most troublesome weed species in eastern Ethiopia. The weeds were collected before flowering and decomposed with and without earthworms. The macro and micro nutrients, physic chemical characters were studied and found suitable to be used as suitable organic manure for different crops.

Keywords- Parthenium hysterophorus, compost , vermicompost, nutrient value

I.INTRODUCTION

Organic farming is defined as production system which avoids or largely excludes the use of synthetically compounded fertilizes, pesticides, growth regulators and livestock feed additives to the maximum extent feasible. The main aims of organic farming are to achieve food and nutritional security, to encourage long term maintenance of soil fertility, crop productivity and soil health, to encourage and enhance biological cycles within the farming system, involving microorganisms, soil flora, soil fauna, plants and animals. This also help in conserving, developing and utilizing the natural resources in the efficient way, to recycle/reuse biomass materials either on farm or elsewhere in order to minimize pollution that may result from agricultural practices. Biodegradable material of microbial, plant or animal origin shall form the

basis of the fertilization programme in organic farming.

Parthenium (Parthenium hysterophorus L., Asteraceae) is an aggressive invasive alien weed species[1], native to the Americas but now widely spread in Asia, Africa and Australia [2]. Partheniumweed was first introduced accidentally into Ethiopia in the 1970s. It was first reported from Ethiopia in 1988 at Dire- Dawa and Harerge, Eastern Ethiopia [3] and subsequently found near Desse, North-eastern Ethiopia. Both are major food-aid distribution centers and there is a strong assertion that Parthenium weed seeds were imported from subtropical North America as a contaminant of grain food aid during the 1980s famine and distributed with the grain [4].Afterwards it spreads rapidly in all regions of the country, along roads and railways, through grazing areas and arable lands, adversely affecting crop production, animal husbandry and biodiversity [5]. The successful


55

NCSTI-2012

spread of Parthenium in so many parts of the world including Ethiopia has mainly been attributed to its allelopathic properties, which enables it to compete effectively with crops and pasture species [6,7,8]Parthenium is considered a noxious weed because of its allelopathic effect its strong competitiveness for soil moisture and nutrients and the hazard it poses to humans. Study also indicated that Parthenium causes general illness, asthmatic problems, irritations of skin and pustules on hand balls, stretching and cracking of skin and stomach pains on humans. [9] and animals [10].

Parthenium weed seed is also a contaminant of grain, pasture and forage seeds. Hence, it results in restricted sale and movement of these produces [11]. Parthenium is also known by its environmental impacts. Because of its invasive capacity and allelopathic properties, it has the potential to disrupt natural ecosystems. It is an aggressive colonizer of wasteland, road sides, railway sides, water courses, cultivated fields and overgrazed pastures[9]. The allelochemicals released from Parthenium affecting many plant species are sesquiterpene lactones and phenolics [12]. Parthenin is the major sequiterpene lactone whereas caffeic, vanillic, ferulic, chlorogenic and anisic acids are the major phenolics[13,14,15].These two synergistically acting groups of allelochemicals signifi-cantly decrease the seed germination and subsequent growth in many crops [6,7,8]. Currently, Parthenium is one of the noxious weeds threatening crop production in Ethiopia. The rapid spread of Parthenium in Ethiopia would be a bigger risk to the expansion and sustainable production of many crops. Control of Parthenium is therefore crucial to boost productivity of many crops in the country. Hand-weeding mostly used by small-scale farmers is more difficult due to the allergic effects of Parthenium on human body. Furthermore, resource poor farmers of Ethiopia can not afford the purchase of herbicides and the use of herbicides is unsafe in terms of health and environmental considerations. Mechanical control on the other hand is rather costly in terms of labour and time requirement. Therefore, other options must be sought for sustainable Parthenium management in Ethiopia. One option is utilizing Parthenium for the purpose it is suitable, management by utilization. Parthenium composting and vermicomposting are therefore one of these options. Composting significantly reduced allelopathic inhibitory effects of Parthenium and suggest Parthenium composting with locally available plant materials as a means to reduce its allelopathic inhibitory effect and as a way of Parthenium management by utilization [16]

II.METHODOLOGY

A. Composting of Wastes

The Parthenium plants were subjected to the composting process.The Parthenium plants (10 kg) were chopped into bits of uniform size. One layer of chopped plants was spread in the tank above which a layer of soil was placed. The process of layering was carried out until it reached a height of 1 m. Water was sprinkled in the stacking process to maintain the moisture content. The tanks were kept covered with wooden planks in an undisturbed condition. The moisture content was maintained during the composting period by sprinkling water and turning it at regular intervals.

B. Vermicomposting of Wastes(Parthenium hysterophorus)

The vermicomposting of wastes was carried out in cement tanks employing the burrowing type of earthworms, Eudrillus eugeniae. The chopped Parthenium plants (10 kg) were layered in separate cement tanks. The cement tank was filled with alternate layers of organic wastes and cow dung slurry. The entire heap was sealed with a paste of cow dung and it was left undisturbed for about twenty days. The earthworms were released into the heap after moisturing it at the completion of 20 days. Then they were covered with wooden planks to prevent the earthworms from crawling out and to protect them from predatory birds. Water was sprinkled regularly to maintain adequate moisture content in the compost and body temperature of earthworms. After 60 days, a black, granular vermicompost was ready for harvest.

C. Analysis of Manurial Value of Fresh Wastes, Composts and Vermicomposts

The Parthenium plants (weed) were collected at the flowering stage, were shade dried and ground and passed through 2 mm sieve. The sieved material was used for analyses. The decomposed Parthenium were shade dried, passed through 2 mm sieve and stored under ambient conditions.The physico-chemical characters and nutrient composition of the composts and vermicomposts were analysed and compared with their fresh organic forms.


56

NCSTI-2012III.RESULTS AND DISCUSSION

The analysis of fresh, composted and vermicomposted Parthenium samples indicated significant variations in the rates of physico-chemical characters, organic constituents, macro and micronutrients, C : N and C : P ratio (Table 1).

The pH of the fresh Parthenium was slightly acidic (6.84) whereas it turned alkaline in composted (8.89) and vermicomposted (8.53) samples. The electrical conductivity of the fresh Parthenium was 0.19 dSm-1 and that of the compost and vermicompost were observed to be 0.30 and 0.28 dSm-1. The moisture content of the fresh, composted and vermicomposted Parthenium were calculated as 10.00%, 20.00% and 30.80%, respectively. The pore space volume was 38.00 ml for fresh Parthenium, 20.00 ml for composted Parthenium and 18.00 ml for vermicomposted Parthenium. The fresh Parthenium had a per cent pore space of 59.38, composted Parthenium 64.52 and vermicomposted Parthenium 58.06. The bulk density was measured as 0.31 Mg m-3 for fresh Parthenium, 0.65 Mg m-3 for composted Parthenium and vermicomposted Parthenium. The particle density of fresh, composted and vermicomposted Parthenium were 0.77, 1.82 and 1.54 Mg m-3.

The chemical characters of fresh and decomposed Parthenium was analysed and the results obtained were tabulated.The content of total free phenolics was high in fresh Parthenium (18.26 mg g-1) and reduced to 4.60 mg g-1 in composted Parthenium and 4.69 mg g-1 in vermicomposted Parthenium.The content of reducing sugars of fresh, composted and vermicomposted Parthenium were 10.09, 9.82 and 9.98 mg g-1 respectively. The content of total soluble sugars was found to be highest in fresh Parthenium (56.54 mg g-1) and decreased in composted Parthenium (23.88 mg g-1) and vermicomposted Parthenium (10.89 mg g-1). The level of non-reducing sugars was more in fresh Parthenium (46.45 mg g-1). The composted Parthenium had a content of 14.06 mg g-1 and the lowest content of 0.91 mg g-1 was observed in vermicomposted Parthenium. The cellulose content was observed as 0.46 mg g-1, 0.57 mg g-1 and 0.66

mg g-1 for fresh, composted and vermicomposted Parthenium.

The content of organic carbon (C) varied among the fresh and decomposed samples. It was 2.02% in fresh Parthenium and reduced to 0.87% in composted Parthenium and 1.03% in vermicomposted Parthenium. The organic matter was found to be 3.49%, 1.50% and 1.78% for fresh, composted and vermicomposted Parthenium. The data in Table 1 reveal that while the fresh Parthenium recorded the lowest per cent of total Nitrogen (N) (0.62%), the composted Parthenium (2.10%) and vermicomposted Parthenium (2.30%) exhibited higher values. While the fresh sample recorded the least value of 0.10% for Phosphorus (P), it was found to be 0.30% in composted Parthenium and 0.42% in vermicomposted Parthenium. The Potassium (K) content was recorded to be 0.32% in fresh and 1.20% and 1.36% in composted and vermicomposted Parthenium samples. It is elucidated from Table 1 that Zinc (Zn) and Copper (Cu) contents were lower in fresh Parthenium (178 and 76 ppm) compared to the composted (210 and 82 ppm) and vermicomposted Parthenium (235 and 80 ppm). The Iron (Fe) content was lower (76 ppm) in composted Parthenium compared to both fresh (84 ppm) and vermicomposted samples (94 ppm). Manganese (Mn) content was observed to be 125 ppm, 140 ppm and 156 ppm respectively in fresh, composted and vermicomposted Parthenium.The C : N ratio of the decomposed samples showed a steady reduction from 3.26 (fresh Parthenium) to 0.41 in composted and 0.45 in vermicomposted Parthenium .There was a drastic reduction in C : P ratio (Table 1) of the samples from 20.20 in fresh sample to 2.90 and 2.45 in composted and vermicomposted samples.

Significant differences in N, P and K contents among the organics were recorded. Highest N content was in the composts prepared from Parthenium (2.95 %) and Chromolaena (2.32%) at pre -flowering stage. This was followed by N content in the compost of Cassia (2.15%) prepared at pre flowering stage as well as poultry manure (2.02%). Least N content was noticed in FYM


57

NCSTI-2012

(0.54%) and in the compost prepared from Partuloca (0.63 %). The compost prepared from weed species at post-flowering stage contained less N, P and K as compared to pre-flowering stage. As regards to the phosphorus is concern, it was maximum in poultry manure (1.60%) followed by the compost prepared before flowering from Cassia (0.80%), Parthenium (0.82 %) and Chromolaena(0.65 %). Least phosphorus content was in FYM

TABLE 1 NUTRIENT STATUS FRESH AND DECOMPOSED Parthenium hysterophorus L

Parameters Fresh Parthenium CompostedParthenium

VermicompostedParthenium

Physical Characters

1. pH 6.84 8.89 8.53

2. Electrical conductivity (dSm-1) 0.19 0.30 0.28

3. Moisture content (%) 10.00 20.00 30.80

4. Pore space volume (ml) 38.00 20.00 18.00

5. Per cent pore space 59.38 64.52 58.06

6. Bulk density (Mg m-3) 0.31 0.65 0.65

7. Particle density (Mg m-3) 0.77 1.82 1.54

Chemical Characters

8. Total free phenolics (mg g-1) 18.26 4.60 4.69

9. Reducing sugars (mg g-1) 10.09 9.82 9.98

10. Total soluble sugars (mg g-1) 56.54 23.88 10.89

11. Non-reducing sugars (mg g-1) 46.45 14.06 0.91

12. Cellulose (mg g-1) 0.46 0.57 0.66

Organic constituents

13. Organic carbon (%) 2.02 0.87 1.03

14. Organic matter (%) 3.49 1.50 1.78

Macro and micro nutrients

15. Nitrogen (%) 0.62 2.10 2.30

16. Phosphorus (%) 0.10 0.30 0.42

17. Potassium (%) 0.32 1.20 1.36

18. Zinc (ppm) 178.00 210.00 235.00

19. Copper (ppm) 76.00 82.00 80.00

20. Iron (ppm) 84.00 76.00 94.00

21. Manganese (ppm) 125.00 140.00 156.00

22.C : N ratio

3.26 0.41 0.45

23. C : P ratio 20.20 2.90 2.45

(0.26 %) and compost of Portulaca (0.35%). Further, potassium concentration was highest in poultry manure (1.42%) and in the compost prepared from Parthenium (1.39%) at pre-flowering stage. This was followed by K content of the composts of Parthenium(1.12%) Cassia (0.85 to 0.95 % ) and Chromolaena (0.78 to 0.88 % ) prepared before flowering. Lowest K content was in FYM (0.34 %), Portulaca compost (0.58 %) and vermin compost (0.55 %)[17].These results are inconformity with the findings of [18].In general these results clearly indicated that compost prepared from the


58

NCSTI-2012weed species specially Parthenium and Cassia at pre-flowering stage was far superior in nutrients content than FYM as well as vermi compost and were comparable with poultry manure. However, N content was maximum in composts of weed species Parthenium, Cassia and Chromolaena prepared before flowering as compared to poultry manure and other organics. Among the composts prepared from the different weed species,compost of Parthenium prepared before flowering contained more N, P and K. These are in line with the findings of [19,20,21].An experiment with vermicomposting of congress grass (Parthenium hysterophorus Linn.), water hyacinth (Eichhornia crassipes) and bhang (Cannabis sativa Linn) showed a high increase in nitrogen, potassium, phosphorus anda high decrease in organic carbon, C/N, C/P ratio in the experiment having Eisenia foetida. The otherwise toxic weeds used in this experiment are thus converted into compost with higher concentration of nutrients[22].

IV.CONCLUSION

The results of the study shows that Parthenium can be successfully used for preparing organic manure and used in increasing the crop productivity as an alternative source to FYM. The decomposition in addition to providing beneficial and profitable manure also ensures quality crop production.

REFERENCES[1] Kohli, R.K, Batish, D.R, Singh, H.P. and Dogra, K. 2006.

Status, invasiveness and environmental threats of three tropical American invasive Weeds (Parthenium hysterophorus L., Ageratum conyzoides L., Lantana camara L.). Biological Invasions, 8: 1501-1510.

[2] Evans, H.C.1997. Parthenium hysterophorus: A review of its weed status and the possibilities for biological control. Biocon. N. Info., 18:89-98.

[3] Seifu, W.1990. Parthenium hysterophorus L., a recently introduced noxious weed to Ethiopia. A preliminary reconnaissance survey report on Eastern Ethiopia. East Harerge, Ministry

[4] Tamado, T and Milberg, P .2000. Weed flora in arable fields of eastern Ethiopia with emphasis on the occurrence of Parthenium hysterophorus. Weed Res., 40: 507-521

[5] Tefera, T 2002. Allelopathic effects of Partheniumhysterophorus extracts on seed germination and seedling growth of Eragrostis tef. J. Agron. Crop Sci., 188: 306-310.

[6] Singh, H.P, Batish, D.R, Pandher, J,K and Kohli, R.K. 2003. Assessment of allelopathic properties of Partheniumhysterophorus residues. Agric. Ecosy. Environ., 95: 537-541

[7] Batishm D.R, Singh, H.P, Pandher, J.K and Kohli, R.K 2005). Allelopathic Interference of Partheniumhysterophorus residues in soil. Allelo. J.,15: 267-273.

[8] Batish, D.R, Singh, H.P, Pandher, J.K and Kohli, R.K 2005b. Phytotoxic effects of Parthenium hysterophorus residues on three Brassica sp. Weed Biol. Manage., 5: 105-109.

[9] Wiesner, M, Taye, T, Hoffmann, A, Wilfried, P, Buettner, C, Mewis, I and Ulrichs, C. 2007. Impact of the Pan-Tropical weed Parthenium hysterophorus L. on human health in Ethiopia. “Utilisation of diversity in land use systems: Sustainable and organic approaches to meet human needs”, Tropentag, October 9-11, Witzenhausen

[10] Narasimhan, T.R, Ananth, M, Narayana S.M, Rajendra, B.M, Mangala, A and Subba, R.P.V.1977. Toxicity of Parthenium hysterophorus L. Current Sci., 46: 15–16.

[11] Chippendale, J.F and Panetta, F.D.1994. The cost of Parthenium weed in the Queensland cattle industry. Plant Prot. Quar., 9: 73-76.

[12] Swaminathan, C, Vinaya, R.R.S and Sureshi, K.K.1990. Allelopathic Effects of Parthenium hyterophorus L. on germination and seedling growth of a few multipurpose tress and arable crops. Int. Tree Crops J., 6: 143-150.

[14] Batish, D.R, Singh, H.P, Kohli, R.K, Saxena, D.B and Kaur, S 2002. Allelopathic effects of parthenin against Avena fatua and Bidens pilosa. Environ.Exp. Bot., 47: 149-155.

[15] Batish, D.R, Singh, H.P, Kohli, R.K, Kaur, S, Saxena, D.B and Yadav, S 2007. Assessment of parthenin against some weeds. Zeitschrift für Naturforschung 62c: 367-372.

[15] Singh, H.P, Batish, D.R, Kohli, R.K, Saxena, D.B and Arora, V .2002. Effect of parthenin – a sesquiterpene lactone from Parthenium hysterophorus on early growth and physiology of Ageratum conyzoides. J. Chem. Ecol., 28: 2169-2179.

[16] Mulatu Wakjira, Gezahegn Berecha and Solomon Tulu.2009Allelopathic effects of an invasive alien weed Parthenium hysterophorus L. compost on lettuce germination and growth. African J. of Agric. Res., 4:1325-1330

[17] Channappagoudar, B. B, Biradar, N.R, Patil, J.B and Gasimani, C.A.A. 2007.Utilization of weed Biomass as an Organic Source in Sorghum. Karnataka J. Agric. Sci., 20: 245-248

[18] Biradar, D. P., Shivakumar, K. S., Prakash, S. S. and Pujar, T. 2006, Bionutrient potentiality of Partheniumhysterophours and its utility of green manure in rice ecosystem. Karnataka J. Agric. Sci., 19 : 256-263.

[19] Stolyarenko, V. S., Kovalenko, V. E., Samoshkin, A. A.,Bondar, P. S., Pashova, V. T. and Skripnik, L. N.,1992, Growth, development and microelement content in sorghum manured with organic waste bioconversion products. Fiziologiya Biokhimiya Kul Turnykh Rastenii, 24: 476-482.

[20] Raghuwanshi, R. K. S. AND Umat Rajiv., 1994. Integrated nutrient management in sorghum (Sorghum bicolor)wheat(Triticum aestivum) cropping system. Indian Journal of Agronomy, 39 : 193-197.

[21] Chadwick, D. R., John, F., Pain, B. F., Chambers, B. J.and Williams, J. 2000.Plant uptake of nitrogen from the organic nitrogen fraction of animal manures : a laboratory experiment. J of Agric. Sci., 154 : 159-168

[22] Avnish Chauhan and Joshi, P.C.2010 Composting of Some Dangerous and Toxic Weeds Using Eisenia foetida.Journalof American Science ,6:1-6

Economic Efficiency of Export–Oriented Cattle-Fattening Farms in East Shewa Zone: The Case of Adama City

59

NCSTI-2012


And it’s Surroundings, Ethiopia

Oumer Berisso Department of Economics, School of Business and Economics,

Adama Science and Technology University, PO Box 1888, Adama, Ethiopia. E-mail: [email protected]

Abstract -This study was undertaken to measure the economic efficiency of export–oriented cattle-fattening farms in Adama city and its surroundings and to identify their deter-minants. Technical, allocative and economic efficiencies were estimated by a non-parametric Data Envelopment Analysis (DEA) method and the identification of their determinants using Tobit model. The study result shows the existence of substantial inefficiencies in production as well as efficiency differentials among farms. The mean technical, allocative and economic effi-ciencies were found to be 88, 62 and 55%, respectively. An econometric analysis based on Tobit model indicates that length of fattening period had significant negative impact on all efficiency scores, whereas training and schooling had significant positive impact on allocative and eco-nomic efficiencies. Farm head’s year of experience on cattle-fattening, extension contact and off farm income had significant positive impact on technical efficiency. Experience sharing pro-gram, year of experience and feeding frequency has significant positive impact on economic ef-ficiency. Policy measures aimed at developing better farm-level training programs focusing on the fattening period, feeding and management recordkeeping focus on enhancing farms’ access to information, and marketing systems are recommendations drawn from the study to increase productive efficiency of cattle fattening farms in the study area.

Keywords: DEA, Tobit analysis, cattle- fattening farms, technical, economic, allocative efficien-cies.

I. INTRODUCTION Like most Sub-Saharan African countries,

Ethiopian economy is heavily reliant on agri-culture and allied economic activities. This is because agriculture adds 47% to the total GDP, provides livelihood to more than 80% of the population, constitutes nearly 90% of the nation’s total export and makes most of the exchange earnings to the economy (CSA, 2007). When we take the livestock subsector, Ethiopia ranks first in Africa and tenth in the world with respect to livestock population. It is estimated at around 35 million tropical live-stock units (TLU), which includes 30 million heads of cattle, 42 million heads of sheep and goats, about 7 million equines, one million camels, over 53 million chickens, 10 million bee colonies, and 40 thousand ton annual har-vestable fish (Azage et al., 2009) in the coun-try. Livestock production contributes an esti-mated 16 percent to the total GDP and over 40

percent to the agricultural GDP (Diao et al., 2007), 15% of export earnings and 30% of agricultural employment (Staal et al., 2008). The sub-sector is the second major source of foreign currency through export of live ani-mals, hides and skins (ILRI, 2003). The sector also employs about one-third of the country’s rural population (ILRI, 2003). The recent eco-nomic policy of the government of Ethiopia under Growth and Transformation Plan (FDRE-GTP, 2011) places a great emphasis on commercialization of agriculture, diversifica-tion of production and exports, and private sector investment in order to move farmers beyond subsistence farming to small-scale market-oriented agriculture. Nevertheless, the agricultural service delivery system in general and livestock service in particular which were implemented in the past did not lead to the envisaged commercialization of agricultural producers. Hence, with the process of com-mercialization of the country’s subsistence-oriented production systems to more produc-


60

NCSTI-2012

tive and market-oriented production systems, the agricultural support service has to be trans-formed and should become responsive and innovative.

On the other hand, despite that agriculture is primarily a rural based activity in Ethiopia, the phenomenon of urban and peri-urban agri-culture (UPA) is also becoming evident. There is large population of agricultural producers with in urban and peri-urban areas. Now a days, market oriented livestock production systems are one of UPA emerging and domi-nating most urban centers in the country, vary-ing in kinds and levels. The systems involve the production, processing and marketing of milk and milk products, animal breeding and cattle fattening farm productions. Addis Ababa and Bushoftu cities and their surroundings areone of the areas that exhibit market oriented dairy production. Adama city and its sur-rounding towns are also well known by its cat-tle fattening farms as a major market –oriented, especially export -oriented livestock production activity. Thus given the current and previous governments’ economic policies that aim at improving farm productivity; productiv-ity and market performance of market oriented livestock production with export potential, detailed and systematic empirical studies on the production efficiency of market oriented livestock production in Ethiopia are scarce or non-existent, particularly of export-oriented. Accourdingly this research have been con-dacted to assess farm level technical, alloca-tive and economic efficiencies of export-oriented cattle-fattening farms and its con-founding factors in Adama city and its sur-roundings.

II. MATERIALS AND METHODS The data used for this study were col-

lected both from primary and secondary sources. The primary data was collected through farm survey, field observations and key informant interviews using structured questionnaires. The content of the question-naire was mainly emphasized on personal characteristics of farmers, farm characteristics, and access to institutions. Before embarking on the actual survey, a pre-test of the ques-tionnaire was conducted in order to identify difficult or unwanted questions, to see the re-action of the respondents to the questions, and to check the appropriateness of the question-naire.

To generate primary data for the research a two stage cluster random sampling technique

was used to select a sample of respondents. Firstly, a total list of export-oriented cattle-fattening farms was obtained after stratifying them into two groups, namely, urban, and peri-urban feedlots. Secondly, a list of` the sample respondent of cattle-fattening farms who has bulls in their feedlots during the survey period was prepared randomly from both lists. Ac-cordingly the total sample size for the study was 62 cattle-fattening farms that are selected randomly from a total population of 95 cattle-fattening farms in the study area, which has bulls in their feedlots during the survey period.

For data analysis, descriptive method, which includes frequency distributions, maxi-mum, minimum, percentages, measures of central tendencies and measures of dispersion of the sample units, were used to describe some important characteristics of the sampled farms; a mathematical (linear) programming method called DEA was used to estimate effi-ciency of production and an econometric method called Tobit was used to identify fac-tors affecting the efficiency of production.

THE MODELS The models used in the study were a two-

stage approach involving DEA and Tobit models. In this approach, a DEA problem is solved in the first stage using General Alge-braic Modelling System (GAMS) and at the second stage, the efficiency scores from the first stage are regressed upon different ex-planatory variables using Tobit model. Hence, under DEA frontier, we used the model speci-fied by equation (1) and solve the same linear programming problem (Charnes et al., 1978) for the ith farm or for each DMUsi i = 1, 2,..., n; and obtain the technical efficiency (TE) score for each of the n DMUs in the sample.

000.

min,

XxYyts

i

i

(1)

Like wise, in order to derive a measure of total economic efficiency (EE) index, the model

specified by equation (2) were used and solved the same input-based cost minimization DEA

model. (2) 000.

min

*

,

*

XxYyts

xw

i

i

xi

ti

i


61

NCSTI-2012

Then the total or overall economic efficiency (EEi) index for the ith firm is computed as the

ratio of the minimum cost to the observed cost given input prices and CRS technology (Coelli

et al., 2005), given by equation :

i

ti

it

ii XW

XWEE*

(3)

Allocative efficiency (AEi) of the ith firm is then the ratio of economic efficiency and tech-nical efficiency of the ith firm. For the ineffi-ciency factors the Tobit model (Tobin, 1958),

is specified as:

(4)

111

1

ii

ii

n

jijji

EifEEif

E

xE

Where i refers to the i-th farm in the sample; xj represents various farm-specific variables, as defined earlier; βj are parameters to be esti-mated; ηi is the error term and

i ; Ei is an efficiency measure representing technical or allocative or eco-nomic efficiency and Ei

* is the latent variable, with

),0( 2N

iii xxEE .

DESCRIPTION OF MODEL VARIABLES A. Input-Output Variables for DEA Models

Output Variable: - Yield or the output variable(y) in the production function for this study is a yield per farm which is defined as the actual quantity of the output (the total sum of live weight of all bulls in the feedlot) pro-duced per farm during the survey period, measured in kilograms.

Production Inputs for DEA Model:-The quantity and qualities of the inputs and techni-cal skills of the farmer to use the inputs prop-erly determines production or efficiency of the farm producers. Thus, for this study six inputs such as Labor(in hours), Number of Bulls(in head), Total Quantity of Concentrate Feed (quintal), Total Quantity of Roughage Feed (quintal), Length of Fattening Periods(in days), Costs of Veterinary and Medicine (in birr) were specified as production inputs and used in the production functions (equations). B. Variables for Tobit Model

The Dependent Variable for The Tobit Model: Are the technical efficiency, allocative efficiency and economic efficiency estimates

derived from DEA frontier models in the pro-duction functions.

Independent Variables for Inefficiency Factors: After economic efficiencies are esti-mated for each farm, sources of inefficiency differential among farm, besides input con-straints, were estimated using efficiency scores as dependent variable, using Tobit analysis. For this purpose twelve explanatory Variables were used for inefficiency model for this study, and are categorized in three groups as follow:

Demographic Factors: These include per-sonal characteristics of farm head such as age, education and cattle fatting farm experience of the farm heads.

Socioeconomic Factors: These are farm characteristics such as farm size (number of bulls), length of fattening period, daily feeding frequency, off/non-farm employment and ex-isting management record.

Institutional Factors: These include ac-cess to institutions such as use of credit, exten-sion service, training of the farm head on cat-tle-fattening farm and their participation on experience sharing programs.

III. RESULTS AND DISCUSSION Descriptive Results:-Descriptive statistics

were used to analyze primary data. Accord-ingly the summery of descriptive analysis of the output and input variables used in the non-parametric DEA models and of the explana-tory variables used in Tobit model are pre-sented as below. A. Descriptive Analysis of the Output and in-put Variables for the DEA Models

The Descriptive statistics result shows that the sample cattle-fattening farms produced 72.9 tons of live weight, on average. The minimum was 17.2 tons and the maximum was 220 tons. To reach the present level of production, cattle-fattening farms used ap-proximately 3376 labour hours, 205 tons of total feed, of which 129.1 tons is concentrate (by-product) feed and 75.9 tons is roughage feed on average. In addition to these inputs, farms expenses 4427 birr for veterinary and medical costs on average during the fattening period. The sampled farms kept their bulls on average for 67 days in the feedlots. The aver-age Feed intake per animal during the fatten-ing period was 1051.12 kg, with concentrate feed 662 kg and roughages 389.11 kg. Daily feed intake per animal during the fattening period was 15.75kg with concentrate feed 9.92


62

NCSTI-2012

kg and roughages of 5.83 kg. The farms were obtained a bull weighing a live weight of 373.92 kg on average during the fattening pe-riod. Thus the farmers were producing on av-erage above the country’s minimum require-ment of live weight of bulls to export from the country during the fattening period; as the live weight average is 373.92 kg in the study area while the minimum national value to export from a country is 325 kg. B. Descriptive Analysis for Explanatory Vari-ables Used in Tobit Model

Demographic Characteristics :- The sam-ple farms, the average age of the farm heads at the time of the survey was 42 years with the minimum and maximum 22 and 65 years, re-spectively. Furthermore, when we see age of the farm head in age range; 2 (3.2%) of them are less than 31 years old, 30 (48.4%) of them are in between 31-40 age, 25 (40.3%) of them are between age of 41-55, and only 5 (8.1) of them are more than 56 years old. This indi-cates most of the producers are young. When we see year of farm head experience of the sampled farms on cattle-fattening farm or/and similar activities, the sampled respondent farm head have on average 7.65 years of farm ex-periences that ranges from 1 year to 18 years of farm experience. Majority of the sample respondents, that is 53 (85.5%) have attended at least elementary school education out of which 11 (17.7%) have completed primary education, 1-6; 29 (46.8%) completed junior and secondary education, 7-12; 8 (12.9%) have graduated with diploma or certificate and 5 (8.10%) have BA/BSC degree qualification. Meanwhile, 9 (14.5%) of the sampled produc-ers are illiterate.

Socio Economic Characteristics: - Only about 29 (47%) of the sampled farm head re-ported to earn off/non-farm income from non-cattle fattening farm activities. About 48 (77.4%) the sampled farm have management records. Regarding feeding frequency, the ma-jority of them, that is, 56(90.3%) farmers were feed their bulls twice per day and only 6(9.7%) of the farmers reported that they feed their cattle three times daily.

Institutional Characteristics:- A total of 47(75.8%) of the sampled farms have reported contact with extension agents but have very few contacts with extension services in a month, only (1.6 times on average, or 1-4 times per month). Exactly half of the sampled farms have access to credit of which 25(40.3%) have used formal credit institutions,

while 6(9.7%) have used informal credit sources either from relatives, friends or local money lenders. Credit use averaged approxi-mately 365806.5 birr per farm. About 38(61.3%) of the sampled producers have ac-cess to training on cattle-fattening farm system and 18 (29%) of them have obtained experi-ence sharing out of Ethiopia.

DEA AND ECONOMETRIC MODELS RESULT

I. DEA Result: Analysis of Production Effi-ciency

A. Technical Efficiency Analysis:-The re-sults derived from DEA models as shown in table 7 indicates that technical efficiency indi-ces differ substantially among farms, ranging from 0.34 to 1.00 with 19 fully efficient farms out of the 62 sampled farms with an average efficiency score of 0.88. An average efficiency score of 0.88 implies that if the average farm in the sample was to achieve the technical ef-ficiency level of its most efficient counterpart, then the average farm could realize a 12 per-cent input reduction . A similar calculation for the most technically inefficient farm reveals input saving of 66 percent. Though the maxi-mum and the minimum technical efficiency scores differ considerably, the modal technical efficiency class is 0.91-1.00, representing 46.8 present of the sample producers indicating that the highest number of farmers have technical efficiencies between 0.90 – 1.00. Of the sam-ple farms, 79% had a higher technical effi-ciency scores than the mean technical effi-ciency. This implies that the farms are fairly technically efficient.

B. Allocative Efficiency Analysis: - The estimated allocative efficiencies differ sub-stantially among the farms ranging between 0.31 and 1.00 with the mean allocative effi-ciency of 0.62. This implies that if the average farm in the sample was to achieve allocative efficiency level of its most efficient counter-part, then the average farm could realize 38 percent cost saving. A similar calculation for the most allocatively inefficient farm reveals cost saving of 69 percent. The frequency of occurrence of the estimated allocative effi-ciencies in deciles ranges indicate that a clus-tering of allocative efficiencies in the region of 0.61-0.70 efficiencies range representing 46.8 percent of the respondents.


63

NCSTI-2012

C. Economic Efficiency Analysis:-The combined effect of technical and allocative efficiency results shows that the average eco-nomic efficiency level for the sample farms is 0.55, with a lowest of 0.20 and a highest of 1.00. This indicates the existence of substan-tial economic inefficiencies of production in the study area. However, low economic effi-ciency scores reveal that there is a consider-able room to increase agricultural output with-out additional inputs, given the existing tech-nology. If these farms operated at full effi-ciency levels, they could reduce, on average, their costs of production by 45 percent and still produce the same level of output.

II. TOBIT Estimates: Analysis of Determinants of Efficiency:

Tobit analysis was used to explain effi-ciency differentials among the sampled farms. Before undertaking Tobit model estimation and making an econometric analysis, the ex-planatory variables were checked for their multicollinearity using different batteries of tests. To test existence of multicollinearity in the hypothesized explanatory variables, VIF and TOL was computed. The results show that the entire explanatory variables have no seri-ous multicollinearity problem. Furthermore, for hetroscedasticty test, we used a hetrosce-dasticy robust method; robust standard errors are often reported in applied cross-sectional work, especially when there is heteroskedas-ticity problem (Jeffrey, 2002). Finally Tobit estimation was conducted by STATA.

ECONOMETRICS ANALYSIS The Tobit results show that except credit

amount and length of the fattening period all explanatory variables have a positive relation-ship with technical efficiencies.

The significant positive relationship of off/non-farm income with technical efficiency might be explained by the fact that off/non-farm employment may absorb under-employed labour resources, improve the experience and human capital of the farm operator and bring additional income that could be used for fund-ing farm activities, thus leading to a positive relationship with technical efficiency. More number of contacts with extension services was associated with greater technical effi-ciency. Thus extension visits were found to positively and significantly affect technical efficiency for this study. Similar results were

reported by Kaliba, A.R and C.R Engle, (2006).

Schooling is positively and significantly related to allocative and economic efficiencies, which confirms with previous studies by Coelli et al. (2002), and this result may be ex-plained by the fact that education is expected to make farm head less conservative and more receptive to new technology and innovation, which will consequently lead to higher effi-ciencies. Moreover, this result may be ex-plained by the better education level of the sampled farm head as about 86 percent of the sampled farm head have at least completed primary education thus resulted in positive efficiency relations. Feeding frequency is posi-tively and significantly related to economic efficiency, which indicated that farms with higher feeding frequency were more economi-cally efficient than farms with lower feeding frequency, as expected. Similar results were reported from previous studies (Ceyhan and Hazneci, 2010 and Ghorbani et al., 2009).

Another outcome of the Tobit model was the significant effect of credit use on economic efficiency. Although 31 (50%) of the sampled farms had access to credit from either formal or informal credit sources, credit use showed unexpected negative sign on technical, alloca-tive and economic efficiencies level signifi-cantly. As a result, farms using more credit were more inefficient. This may be due to shortage of working capital, due to high input costs and low returns on outputs, together with high credit costs, the Moral Hazard problem of credit and ‘fungibility of loan’ (or loan diver-sion).

Training had significant positive impact on allocative and economic efficiencies. Simi-lar results were reported from previous studies (Ceyhan and Hazneci, 2010). The result veri-fies the importance of capacity building and technical assistance for agriculture entrepre-neurs and producers to improve production efficiency. Farm experience in cattle-fattening has a significant positive impact on technical and economic efficiencies. This is because most of the sampled farm heads are experi-enced farmers (with average 7.65 years of cat-tle fatting farm experience) and as most of farms are own operated, daily work of farm were done by themselves, thus resulted in a positive influence on technical and economic efficiencies among the farms. The result is in line with the findings of Mbanasor and Kalu (2008). Experience sharing program was found to significantly and positively affect economic efficiency as hypothesized. The


64

NCSTI-2012positive relationship of experience sharing activity with efficiency magnifies the role of experience sharing program as one capacity building programmes in addition to training and extension service as technology adoption can be acquired through observation also. Of course 18 (29%) of the sampled farmers have obtained experience sharing programs out of Ethiopia to some African and Meddle East countries.

Length of fattening period was found to be significantly and negatively affect all effi-ciencies. This result is in line with Ghorbani et al. (2009). The negative relationship of length of fattening period with all efficiencies may be is due to the extended period of fattening pe-riod length as few of the sampled farms kept their bulls for more than three months, which in tern results in loss of live weight of bulls, loss of the standard quality of beef for export and incurs to more unnecessary feed costs and thus resulted in the negative relationships with all efficiencies.

IV. CONCLUSION AND RECOMMENDATIONS

A. Conclusion:- This study evaluated economic efficiency

of export-oriented cattle fattening farms in Adama city and its surroundings and identified their determinants, using DEA and Tobit mod-els respectively. Using detailed survey, data collected in the production year of 2010 from June to July, form 62 sampled cattle fattening farms and measures of technical, allocative and economic efficiencies were obtained. The results indicate that the mean technical, alloca-tive and economic efficiencies were 88%, 62% and 55%, respectively. This indicates the exis-tence of a substantial allocative and economic inefficiencies as well as variations in effi-ciency levels among farms. DEA results show that the cost excess owing to inefficiency of the sample farms in the study areas is, on av-erage, 45% and is mainly due to allocative inefficiency. Tobit result shows that length of the fattening period and credit amount had significant impacts on all efficiency scores, whereas training and schooling had significant impact on allocative and economic efficien-cies. Extension contact, Farm head year of experience and off/non-farm income had a significant impact on technical efficiency. Farm head year of experience, Experience sharing and feeding frequency had a signifi-cant impact on economic efficiencies.

B. Recommendations (Policy Implications):-Policy implications of this analysis are

that efficiency estimates indicate both the dis-tribution of the farms’ efficiency and its insti-tutional and socioeconomic determinants. An analysis of the determinants’ relative effi-ciency indicates which aspects of the farms’ human and physical resources must be tar-geted by public investments to improve farm efficiency. That means the identification of those factors contributing to efficiency differ-entials among farms might help formulate bet-ter policy for intervention. Result from ineffi-ciency differentials suggests that policy-makers should focus on these institutional and socio-economic factors influencing efficiency of production. They should focus on enhanc-ing farms’ access to information via better ex-tension services, marketing systems and en-couraging management record keepings. Ca-pacity building and technical assistance for agriculture producers is important to improve production efficiency. Thus training, agricul-tural extension and other capacity building programs for farms should be provided to im-prove the economic efficiency of individual farms up to at least the level of the best cattle-fattening farms. Hence training focusing on the fattening period, feeding and management record keepings are the major recommendation drown from the study to enhance production efficiency of cattle fattening farms in the study area.

REFERENCES

Azage Tegegne, Berhanu Gebremedihin, Dirk Hoekstra and Nigatu Alemayehu (2009). Rural urban linkage in market-oriented dairy development in Ethiopia: Lessons from the Ada’a dairy cooperative. Ad-dis Ababa, Ethiopia

Binswanger, H. and Rosenzweig, M. (1986). Be-havioral and material determinants of production relations in agriculture. Journal of Development Studies, 22: 503-539.

Bravo-Ureta, B.E. and A.E. Pinheiro (1997).Technical, economic and alloca-tive efficiency in peasant faring evi-dence from the Dominican Republic. Developing Economies, 35(1): 48-67.

Ceyhan, V. and K. Hazneci (2010). Economic efficiency of cattle-fattening farms in Amasya province, Turkey. Journal of Animal and Veterinary Advances, 9 (1): 60-69.

Ethiopian Indigenous Knowledge of Traditional Foods and Beverages Processing: Needs for Modern Food Processing Technology and TransformationsEconomic Efficiency of Export–Oriented Cattle-Fattening Farms in East Shewa Zone: The Case of Adama City

65

NCSTI-2012Charnes, A., W.W. Cooper and E. Rhades (1978).

Measuring the efficiency of decision making units. European Journal of Op-erational Research, 2: 429-444.

Coelli, T. and G. Battese (1996). Identification of factors which influence the technical in-efficiency of Indian farmers. Australian Journal of Agricultural Economics, 40: 103.

Coelli, T., R. Sandura and T. Colin (2002) ‘Techni-cal, Allocative, Cost and Scale Efficien-cies in Bangladesh Rice Production: A Non-Parametric Approach’, Agricultural Economics, 53: 607–626.

CSA (Central Statistical Authority), 2007. Statisti-cal Abstract. CSA, Addis Ababa, Ethio-pia.

Diao, X., Belay F., S. Haggblade, Alemayehu , S., T,. Kassu, W., and Bingxin Yu (2007).Agricultural Growth Linkages in Ethiopia: Estimates using Fixed and Flexible Price Models, IFPRI Discussion Paper No. 00695, Development Strategy and Governance Division, International Food Policy Research Institute, Wash-ington DC.

Ghorbani, A., S.A. Mirmahdavi and E. Rahima-badi, (2009). Economic efficiency of Caspian cattle feedlot farms. Asian Journal of Animal Science, 3: 25-32.

ILRI (International Livestock Research Institute) (2003). Livestock marketing in Ethiopia: A review of structure, performance and development initiatives.

Jeffrey M. Wooldridge, 2002. Econometric Analy-sis of Cross Section and Panel Data. The MIT Press, Cambridge, Massachusetts, London, England

Kaliba, A.R and C.R Engle, (2006). Cost efficiency of catfish farms in Chicot Country, Ar-kansas: The impact of extension ser-vices. Aquac.Econ.Manage. 10:223-243.

MEDaF (Ministry of Economic Development and Finance) (2011). Growth and Transfor-mation Plan, Addis Ababa, Ethiopia.

Mbanasor, J.A. and K.C. Kalu (2008). Economic efficiency of commercial vegetable pro-duction system in Akwa Ibom state, Ni-geria. Tropical and Subtropical Agro Ecosystems, 8: 313- 318.

Staal, S. J., A. P. Nin, and J. Mohammad (2008). Dairy development for the resource poor Part 2; Kenya and Ethiopia dairy devel-opment case studies PPLPI Working Pa-per No. 44-2.

Tobin, J. (1958). Estimation of relationships for limited dependent variables. Economet-rica, 26: 24–36.

Ethiopian Indigenous Knowledge of Traditional Foods and Beverages Processing: Needs for Modern Food Processing Technology and Transformations

66

NCSTI-2012

Ethiopian Indigenous Knowledge of Traditional Foods and Beverages Processing: Needs for Modern Food Processing

Technology and Transformations

Tariku Hunduma Ethiopian Institute of Agricultural Research, Ambo Agricultural Research Center, P.O.Box 37, Ambo,

Ethiopia, [email protected]

Abstract—In Ethiopia, traditional foods and beverages processing practices can generally be categorized into traditional fermentation and instant preparation of foods and beverages. Traditional foods such as Injera, Kocho, Wakalim; condiments such as Awaze, Siljo, and Qochqocha (Datta?); and beverages like Tella, Borde, Shamita, Tej and Katikala are produced through traditional fermentation while local foods such as Kollo,Nifro, Besso and beverages likeBirz are instantly produced. This age-old Indigenous knowledge has been used for generations without sufficient scientific interventions. This traditional process is to be blamed for time and energy consumption, impact on human health and environment. Almost all traditional food and beverage processing practices are done by women. Around Ambo women spent 18-19 hrs a day doing different household activities. Most of the traditional practices are extremely tedious, unhygienic, one-at-a-time/piece-by-piece process and are still at their archaic stage. For instance, traditional enset processing for fermentation is a deadly practice, done by women in group using locally made traditional wooden equipment. Tella, Borde, Shamita, Tej and Katikala making are also tedious which most Ethiopian women practice to support their household life. In Arsi-Negele town 87.3 % of the populations directly support their life by katikala distillation with wood fuel consumption rate of 76.98 kg/day/household. Health impact of traditional cooking and baking due to indoor pollution has also been documented. Environmental impact due to deforestation is getting worse as about 200,000 hectares of forest are cleared for cultivation and wood fuel every year. With the current idea of entrepreneurship, most Ethiopian women are creating their small business either individually or in group. Now it is very common to get commercial Injera, Dabbo,Kollo, Yebekollo Tibs, Yebekollo Kikil, Yedinich Kikil than ever before. Sadly, all are nearly at their archaic practices. What to do then? We need to understand this indigenous traditional knowledge and modify/improve/transform through science and technology. Here, an Ethiopian Tej, Sheba Tejfrom Brotherhood Winery and automated Injera making machine which produces 1000 Injera/hris shining light on technology development and transformation of these traditional processing. With due respect to the past efforts made by all concerned bodies, I dare say, yet we are lagging behind to transform/modify these indigenous traditional practices. Now, with the increasing number of Universities, research institutions and attentions given to the development of Science and Technology in the country, it is time to emphasis on it. Here, synergetic work of food microbiologist, food technologist, mechanical engineers and others is very important for this common goal.

Keywords- Beverages, Foods, Indigenous Knowledge, transformations, Ethiopia.

I. Introduction Indigenous knowledge (IK) is defined as a

knowledge that an indigenous (local) community accumulates over generations of living in a particular area or community. A number of terms are used interchangeably to refer to the concept of IK, including Traditional Knowledge (TK),

Indigenous Technical Knowledge (ITK), Local Knowledge (LK) and Indigenous Knowledge System (IKS) [1].The scopes of indigenous knowledge include traditional foods and beverages processing, traditional herbal medicine, ritual practices, household items preparation, environmental disaster management, agricultural production system, health care practices and the likes. This paper deals with indigenous

Ethiopian Indigenous Knowledge of Traditional Foods and Beverages Processing: Needs for Modern Food Processing Technology and Transformations Ethiopian Indigenous Knowledge of Traditional Foods and Beverages Processing: Needs for Modern Food Processing Technology and Transformations

67

NCSTI-2012

knowledge of traditional foods and beverages processing practices and discussed on needs for modern food processing technology and transformations.

Ethiopia, Home for Indigenous Knowledge

Ethiopia is endowed with ethnic diversity manifested in cultural diversity and a variety of indigenous knowledge such as traditional foods and beverages processing practices. This knowledge is usually passed from generation to generation through traditional socialization processes without sufficient scientific interventions and documentation. In Ethiopia, traditional foods and beverages processing practices can generally be categorized into traditional fermentation and instant preparation of foods and beverages. Varieties of traditional foods/primary food products such as Injera (leavened pancake from Tef, Eragrostis tef),Dabbo (varieties of bread), Kocho (fermented primary food product of enset, Ensete ventricosum), Wakalim (fermented meat sausage) are produced through traditional fermentation. Condiments such as Awaze (fermented red sweet pepper, Capsicum annum with varieties of spices added), Siljo (fermented safflower, Carthmustinctorius and faba bean, Vicia faba flours), and Qochqocha (Datta?) (fermented small chili pepper, Capsicum frutescene at harvest green maturity with varieties of spices added) are also produced through traditional fermentation. Beverages like Tella (a fermented home-made beer from cereals with alcohol content of around 2% to 4%), Borde (a fermented beverage from cereals with low alcoholic content), Shamita (also a fermented beverage from cereals with low alcoholic content), Tej (a fermented honey wine with alcohol content around 6.98 % to 10.9% ),and Katikala/Araki (a fermented and distilled home-made beverage from cereals with alcoholic content around 45%)) are as well produced by traditional fermentation [2-9]. Local foods such as Kollo (roasted cereals and pulses), Nifro (boiledcereals and pulses), Besso (roasted and powdered barley),Genfo (porridge), kitta (unfermented bread), Chuko (barley roasted, powdered and spiced into paste), Kinche (coarsely milled barely/wheat, boiled and spiced), Ashuk (roasted and slightly boiled faba bean grain), Dabbokollo (wheat dough sliced into pea size and roasted),Yebekollo Tibs (roasted green maize cob),Yebekollo kikil (boiled green maize cob), Yedinch kikil (boiled potato tuber), Kitfo (minced and spiced meat), Yesiga Kikil (boiled and spiced meat), Shiro (roasted and powdered pulses),

berbere (dried, and powdered red sweet pepper),different spices, varieties of Wot (sauce) including Dorro (chicken) Wot, Shiro Wot [10] are instantly produced for consumption. Beverages like Birz(an unfermented honey wine and assumed non-alcoholic) are also instantly produced for consumption. There are also lots of traditional foods and beverages around the country not yet investigated and known to others.

II. Significance of Indigenous Knowledge of Traditional Food

and Beverage Processing Indigenous knowledge of traditional foods and

beverages processing has great role in sustaining the livelihoods of the poor. It has great significance: (1) in improving shelf life/food safety through different indigenous knowledge such as drying, smoking, salting and fermenting; example Kuanta (salted and smoked meat),Dirkosh (sun dried and roughly powdered Injera),Chuko, Besso, Dabbokollo, Kocho, Wakalim and dairy products; (2) to make use of non-utilized and underutilized foods (diversify rural food); example Moringa, Mushroom, Amaranthus and enset are commonly consumed in some parts of the country; (3) in diversifying recipes by converting an agricultural produce into varieties of food products; example barley can be prepared as Injera, Chuko, Besso and beverage like shamita; (4) in minimizing yield losses that may occur during later stage of harvesting and storage by converting an agricultural produce into a food product at its green harvest maturity stage; example, Qochqocha (Datta?), Yebekollo Tibs, Yebekollo Kikil; (5) to make food products available year round (enhance food availability in periods of scarcity); example Chuko, Kuanta, Dirkosh, Qochqocha (Datta?), Kocho and dairy products; (6) in increasing availability and accessibility; (7) in addition to the aforementioned, it is income generating technology for most poor Ethiopian households (cf. [2-9] and [10] for part III). Here, we can imagine how many of Ethiopian households are dependent on their traditional knowledge of foods and beverages processing to directly support their life and sustain their family. Tella, Katikala, Borde, Shamita, yebekollo tibs, Yedinch tibs, Kollo vendors are good examples.

However, the fundamental roles of indigenous knowledge of traditional foods and beverages processing in sustaining the livelihoods of the poor have often been neglected and most of the process (if not all) are still at their archaic


68

NCSTI-2012

technology. As a result, this indigenous knowledge is full of challenges.

III. Challenges, Needs for Modern Food Processing Technology and Transformations

This indigenous knowledge is blamed for many challenges such as time and energy consumption, impact on human health and environment, unhygienic conditions. Here are some examples: A. Traditional Enset Processing for Fermentation

Enset, sometimes called “false banana” is a herbaceous and monocotyledonous plant growing to 4 -11 meter in height with its pseudostem dilating at the base to a circumference of 1.5 to 3.0 m [11]. Over 80% of enset produced in the country is grown in the south and southwestern parts of the country [12]. Enset food products are used as staple and/or co-staple food for about 20% of the Ethiopian population [13] and is particularly cultivated for its carbohydrate rich food products.

The corm and pseudostem of enset plant is processed for traditional fermentation in an earthen pit into primary food product, kocho.Traditional enset fermentation is an age-old technique in enset growing regions of the country and is still used without sufficient scientific

modification of the process. The traditional knowledge of enset processing for fermentation normally includes knowledge of how to identify mature enset plant, preparation of pit and processing area, pulverization and decortication, bulla extraction, preparation of fermentation

enhancer, storage of fermenting mass in a pit, continuous mixing and check-up of fermenting mass and determination of stage of

Table 1 Constraint and technology considerations for enset traditional processing for fermentationMain processing stage Constraint Technology considerations

Mature ensetidentification

Harvest maturity signs differ from locality to locality Nutrient content of the plant may not correctly coincides with the local signs being used

Scientific identification of uniform harvest maturity signs that coincides with high nutrient containing stage/age

Harvesting/up-rooting Non-mechanized/locally done by local tools Labor intensive/tedious and time consuming

Mechanization of harvest through tools and equipment

Pulverization and decortications

Non-mechanized/ done by locally made wooden equipmentDone from seating position with one leg raised and also on long standing position and is unhygienic Labor intensive/most tedious and time consuming stage of enset processing

Mechanization of pulverization and decortications that may ease the process and also improve sanitary conditions

Bulla extraction Extraction done either by feet in perforated sacks or by hand using sieve on bucket Tedious, time consuming and unhygienic

Mechanical extractor that ease means of extraction and improve sanitary conditions

Long time natural fermentation in an earthen pit with continuous check-up, mixing and pit or leaves changing.

Not controlled fermentation Tedious, time and energy consuming unhygienic Liable to microbial contamination Not timely for needy households

Identifying important fermentative microorganisms Developing starter culture Transforming natural pit fermentation into controlled environment fermentation, (vessel/ motorized fermentor with impeller or machine-mixer) Improve sanitary conditions and shorten time of fermentation

fermentation [14] (Fig. 1). This traditional knowledge of is generally owned by women [15] and practiced by locally made wooden equipment [16] Constraint and needs

Enset fermentation is long time fermentation from a few weeks to one or more years [17-19]. The process is generally considered tedious, time and energy consuming, unhygienic and reduces quantity and quality of food products [20] (Table 1).Attempts Done So Far in Improving Enset Processing for Fermentation

To ease this tedious, time and energy consuming traditional process, lots of efforts have been made by different concerned bodies/institutions (Fig. 2).


69

NCSTI-2012


70

NCSTI-2012

However, this back breaking exercise still

looks for better processing equipment and food

technology in general. The women involved in

the operation recommended for improvement

of the overall enset processing methods in

order to save energy, time and to make it more

hygienic [21].

B. Preparation of Ethiopian Traditional Foods, Beverages and Condiments

Preparation of almost all Ethiopian traditional foods, beverages and condiments are very tedious, time and energy consuming.


71

NCSTI-2012

They are not only tedious, time and energy consuming but also have health and environmental impacts. These traditional foods and beverages such as Injera, Dabbo, Kollo, Yebekollo Tibs, Yebekollo Kikil, Yedinich Kikil,Talla, Tej, Katikala/Araki, Borde, Shamita and the like are routinely prepared both for household consumption and retail. Traditional condiments such Qochqocha (Datta?) are still processed using traditional stone-mill.

Traditional procedures for preparation of these traditional foods, beverages and condiments include steps like traditional grinding, dehulling, traditional stone milling, sorting, sieving, roasting, kneading dough and baking, distillation and the like. Ethiopian women spent lots of time a day doing these household activities. For example women around Ambo spent 18-19 hrs a day doing different household activities [15]. Some of these activities such as traditional grinding of Gesho (Rhamnus prenoides) leaves/stems and traditional dehulling of various cereals using traditional wooden mortar and pestle,

Mewqecha for most traditional beverage preparation are back breaking exercises. Preparation of bread such as Injera is a hard exercise from grinding cereals into flour to baking. Katikala/Araki preparation from its early stage to distillation is a long and tiresome activity. Health and environmental impact of traditional cooking and baking due to indoor pollution and deforestation has also been documented. For example, the indoor air pollutants level during Katikala/Arakidistillation exceeded air quality guidelines set by the World Health Organization standards [22]. In Arsi-Negele town alone about 87.3 % of the populations directly support their life by katikala distillation (Fig. 3) with wood fuel consumption rate of 76.98 kg/day/household [22]). It was reported that cooking and baking accounts for over 50% of all primary energy consumption in the country and 75% of the total energy consumed in households [23]. Currently, the rate of deforestation is estimated to be 200,000 hectares per year as forest is cleared for cultivation and wood fuel [23].

.

With the current idea of entrepreneurship[25-26], most Ethiopian women are creating their small business either individually or in group and most of our traditional foods which were commonly used in home is getting popularity on market. Now it is very common to get commercial Injera, Dabbo, Kollo, Yebekollo Tibs, Yebekollo Kikil, Yedinich Kikil and others than ever before. Tella, Tej, Birz, Borde and

Shamita are very popular beverages and are commercially available in every traditional beverage retailer houses. Millions of households are sustaining their family life through selling these traditional foods and beverages. According to [3], over 2 million hectoliters of Tella is thought to be produced annually in households and Tella vending houses in Addis Ababa. Very recently the


72

NCSTI-2012

popularity of Yegebs kollo (best example is Elsa Kollo) at commercial level even in supermarket is a newly generating entrepreneurship idea. Sadly, all are nearly at their archaic practices. For example, most Kollo producers use traditional wooden dehuller, mortar and pestle, Mewqecha (Fig.4) to dehull the seeds. Study done somewhere else[27], indicated that use of mechanical dehulling is advantageous as it gives more dehulled grain yield with uniform dehulling of grains than traditional dehulling, and reduces the drudgery of processing a large quantity of grain and thus save time.

They also perform roasting one-at-a-time/piece-by-piece which consumes time and energy and also leads to lack of uniformity in the final food product as the temperature is not

controlled. Quality and consistency of quality of the final food products have great market implications [28] and one needs to give attention to such factors in doing such business. Ethiopian condiments such as Awaze, Siljo andQochqocha (Datta?) are the most frequently used condiments [3] and have great potential to be commercialized. Compare our traditional condiment with tomato ketchup, the imported and commercially available condiment (Fig. 5). But the challenge of traditional processing such as milling using traditional stone-mill and grinding using the traditional wooden mortar and pestle, Mewqecha is still tedious, time and energy consuming. Table 2 presents some processing steps, constraints and technology considerations for some processing steps of the traditional methods.

Table 2 Constraint and technology considerations in some Ethiopian traditional foods and beverages processing Some processing steps Constraints Technology considerations Gesho (leaves /stem) grinding

Done using traditional wooden mortar and pestle, MewqechaInconsistency in quality of grinding one-at-a-time/piece-by-piece process tedious, time and energy consuming

Mechanization of grinding (mechanical grinder)


73

NCSTI-2012

Some processing steps Constraints Technology considerations Continued…

Dehulling cereals such as in preparation of Kollo, Shamita

Done using traditional wooden dehuller, mortar and pestle, Mewqechaone-at-a-time/piece-by-piece process tedious, time and energy consuming Lack of uniformity in dehulled grains, inconsistency in dehulled grains poor dehulling quality, highly broken

Mechanization of dehulling (mechanical dehuller)

Roasting such as for Kollo preparation using Mitad, traditional flat clay pan or metal pan

one-at-a-time/piece-by-piece process Lack of uniformity in roasted grains (great market concern), inconsistency in quality of Kollosevere indoor pollution High fire wood consumption

Simplified electric/gas oven Continuous oven with uniform/ controllable temperature and conveyor belt for mass roasting

Mass deep roasting, Asharo such as for Tellapreparation using mostly flat metal pan

One-at-a-time/piece-by-piece process Contribute to inconsistency in quality ofbeverages producedSevere indoor pollution High fire wood consumption

Simplified electric/gas oven Continuous oven with uniform/controllable temperature and conveyor belt for mass roasting

Kneading dough, and bread baking such as Injera using mitad

one-at-a-time/piece-by-piece process tedious, time and energy consuming unhygienic Inconsistency in quality of the food products severe indoor pollution High fire wood consumption

Mechanization of kneading and baking. cf. “Injera making machine” below

Katikala/Arakidistillation

one-at-a-time/piece-by-piece process tedious, time and energy consuming Inconsistency in quality of the beverage produced severe indoor pollution High fire wood consumption

Transformation of Katikaladistillation into modern controlled distillation system

Condiment preparation such as Qochqocha (Datta?)

Done by traditional stone mill Unhygienic tedious, time and energy consuming undertaken by natural fermentation Inconsistency in quality of the final products Liable to microbial contamination

Transformation of condiment preparation in general Identifying important fermentative microorganismsDeveloping starter culture Transforming the natural fermentation into controlled environment fermentation Improve sanitary conditions

Continued…

IV. Exemplary Technologies in Transforming the Indigenous Knowledge

A. “Injera making machine”

Our customary pancake, Injera, is made through tedious, time and energy consuming traditional process. The process starts from flour production to baking. Now, in most part of the country, milling for flour production is commonly done by electrical mill, otherwise by traditional stone-mill. Then the flour is mixed with water and allowed to ferment through natural

fermentation for about two to three days [3] and is then ready to bake into large flat pancakes on traditional flat clay pan and rarely on an electric stove.

Now, this traditional processing of Injera has been dramatically transformed into modern food processing technology. This recent technology has addressed the issue of energy efficiency of the current method of Injerabaking, one- at- a- time process, sanitary, health and environmental impact. This technology has come into reality in the United


74

NCSTI-2012

States of America when the first automated “Injera making Machine” that can produce 500 unites/hr (first generation) and 1000 units/hr (second generation) was designed and manufactured (Fig. 6) by an Ethiopian Prof. Wudneh Admassu, a chemical engineer. The second generation machine is put in place in Washington DC in the summer of 2004. This machine can produce 1000 unit/hr consistently and uniformly 24 hours a day using available

electricity, which can be enough to feed up to 8,000 to 10,000 people. Therefore the Injera machine can easily be installed and operated wherever electric power is available. The Injera made using this technology has a shelf life of at least 5 to 6 days due to minimal contact by bare hand, which is another improvement over the old way of making Injera, which becomes moldy within 2 to 3 days of shelf life [23].

(Source: [23]

B. Sheba Tej: America’s Favorite

Ethiopian Honey Wine

In Ethiopia Tej has been known as one of our

famous and favorite traditional beverages for

centuries. But not commercialized at large

scale level except household venders have

made it available on market to support their

daily subsistence. Now Ethiopian Tej called

Sheba Tej in United Stated of America from

Brotherhood Winery is becoming popular

beverage. Sheba Tej made from pure organic

honey is now part of the premium wine list of Brotherhood Winery (Fig.7)


75

NCSTI-2012

Please read the story how Ernest McCaleb come across this entrepreneur idea of presenting this ‘Nectar of the Gods’ to the world. “America’s oldest winery began producing one of the world’s oldest wines after an African American entrepreneur, Ernest McCaleb, met and initiated a joint collaboration with Brotherhood Winery. McCaleb focused on the production and distribution of organic Ethiopian honey wine. A chance meeting with an Ethiopian in Paris gave rise to his eventual introduction to Ethiopian honey wine. Having a great passion for Africa, its diversity, traditions, and history, McCaleb continued on his entrepreneurial quest and established Sheba in 2003 with the sole purpose of producing authentic honey wine according to ancient Ethiopian traditions. To that end, he arranged for three generations of Ethiopian women—a mother, her daughter and granddaughter—to travel from Addis Ababa, Ethiopia to New York’s Brotherhood Winery to demonstrate how Tej is prepared. Cesar Baeza, an internationally-renowned Chilean wine master and new owner of Brotherhood Winery, studied how this first batch of Sheba Tej was made. The careful end product was a naturally fermented, organic drink with a pleasing golden yellow hue—an ancient, spicy, semi-dry, full-bodied wine. The aroma of honey and wild flower permeated the air, and the Tej was joyously tasted by Baeza and the employees of Brotherhood Winery in conjunction with a hearty meal of Injera and Wot prepared by the three Ethiopian women.

Since then, Sheba Tej, produced at Brotherhood Winery has won awards at international honey wine festivals, and is distributed in many stores across the U.S. and the Caribbean. “Since I’ve begun doing this,” McCaleb says, “I’ve learned more about this rich history, and as I give tasting sessions I have become even more inspired. This is beyond the commercial success. It’s about pride and heritage, which those women taught us when they came to Brotherhood Winery.” ”

From these two excellent examples you can now imagine how these ancient indigenous traditional heritages have been transformed into modern industrial production level.

Source: [29]

V. Future Direction and Conclusion For generations, indigenous knowledge of

traditional foods and beverages processing has been serving indigenous peoples in sustainable living. However, this knowledge is still underutilized resource in the development process of Ethiopia. One big challenge in utilizing this knowledge is lack of awareness, negligence and undermining. In the process of poverty alleviation programme, increasing agricultural productivity in terms of yield increase per sedoes not bring food security unless food processing practices that could increase recipes diversification, year round availability, accessibility and affordability (agro-industrial and agribusiness development) should be taken into consideration in parallel to yield increase. In this regard, improving Indigenous knowledge of food and beverage processing can be used as good opportunity in reducing poverty and eliminate hunger and malnutrition as this indigenous knowledge has a lot to do with them, though lack of attention toward understanding and improving this knowledge remains a Challenge. In Ethiopia, undoubtedly, there are lots of potential Indigenous knowledge of food and beverage processing practices which are not yet addressed and even their existence is not yet known to others. We can imagine how a single chance meet has provoked Ernest McCaleb to transform Tej to industrial production level with an entrepreneur idea of commercializing it. Ernest McCaleb, Brotherhood Winery has learned the knowledge of Tej production from Ethiopian women whom he has transported them to New York’s Brotherhood Winery to demonstrate the art of making Tej. We Ethiopians are here for generations along with this and other such traditional arts and have not been thought about improving and make use of these arts.

After long generations of tedious, time and energy consuming process of Injera making, a significantly transforming technology has come into reality. This can be a footprint for


76

NCSTI-2012

the many unreached and even untouched indigenous heritages of the country for better way of life and economic benefit. Besides, if such indigenous knowledge get understood and improved through science and technology, it can also provides linkages between agricultural producers and processors creating good opportunity for value addition on agricultural raw materials. In Ethiopia, currently, the idea of entrepreneur is now getting momentum than ever adding value to most of the agricultural produces, and this needs support from science and technology.

To maximally make use of the indigenous knowledge, it is imperative to secure the existing gaps such as lack of awareness, ignorance and undermining. Understanding and formal documentation of this heritage is very essential for further improvement/modification/transformation of the art through science and technology. With due respect to the past efforts made by all concerned bodies, I dare say, yet we are lagging behind to transform/modify this indigenous traditional practice. Now, with the increasing number of universities, research institutions and attentions given to the development of Science and Technology in the country, it is time to emphasis on it. Here, synergetic work of food microbiologist, food technologist, mechanical engineers and others is very important for this common goal.

References

[1] UNEP. http://www.unep.org/ik/Pages.asp?id=About%20IK[2] Ketema, B., Tetemke, M. and Mogessie, A.

“Antimicrobial susceptibility patterns of lab isolated from wakalim, a traditional Ethiopian fermented sausage”. Journal of Food Safety 30, pp. 213–223, 2010.

[3] Mogessie A. “A review on the microbiology of indigenous fermented foods and beverages of Ethiopia”.Ethiop. J. Biol. Sci. 5(2): 189-245. 2006.

[4] Girum, T., Eden, E. and Mogessie, A. “Assessment of the antimicrobial activity of lactic acid bacteria isolated from Borde and Shamita, traditional Ethiopian fermented beverages, on some food-borne pathogens and effect of growth medium on the inhibitory activity”. Internet Journal of food safety. 5: 13-20.2005.

[5] Kebede, A., Langsrud, T, Fekadu, B., Narvhus, J. A. “The effects of technological modifications on the fermentation of borde, an Ethiopian traditional fermented cereal beverage”. Journal of Food Technology in Africa, 9 (1): 3-12.2004.

[6] Kebede, A., Fekadu, B., Langsrud, T., and Narvhus, J. A. “Indigenous processing methods and raw materials of of borde, an Ethiopian traditional fermented beverage”. Journal of Food Technology in Africa, 7 : 59-64. 2002.

[7] Ahmed, I., Tetemke, M. and Mogessie, A. “Some microbiological and biochemical studies on the fermentation of ‘awaze’ and ‘datta’, traditional Ethiopian condiments”. International Journal of Food Sciences and Nutrition, 52: 5-14. 2001.

[8] Kelbessa, U., Alemu, F., and Eskinder, B. “Effect of natural fermentation on nutritional and anti-nutritional factors of tef (Eragrostis tef)”. Ethiop.J. Health. Dev., 11: 61-66. 1997.

[9] Kelbessa, U., Alemu, F., and Eskinder, B. “Natural fermentation of enset (Ensete ventricosum) for the production of Kocho”. Ethiop. J. Health. Dev., 11: 75-81. 1997.

[10] Mesfin W. “Ethiopia, change in dietary behavior and food security”. United Nations Educational, Scientific and Cultural Organization. 1985.

[11] Taye, B. 1993. “An overview on enset research and future technological needs for enhancing its production and utilization”. In: Tsedeke Abate, Hiebsch C, Brandt S. A. and Seifu Gebremariam (eds). Enset-Based Sustainable Agriculture in Ethiopia. Proceedings from the International Workshop on Enset held in Addis Ababa, Ethiopia, 13-20 December 1993.

[12] Terefe’ B. “Enset research in Ethiopia: 1985-1993”. In: Tsedeke Abate, Hiebsch C, Brandt S. A. and Seifu Gebremariam (eds). Enset-Based Sustainable Agriculture in Ethiopia. Proceedings from the International Workshop on Enset held in Addis Ababa, Ethiopia, 13-20 December 1993.

[13] Brandt, S. A., Spring, A., Hiebsch, C., McCabe, J. T., Tabogie, E. Diro, M., Wolde-Michael, G., Yntiso, G. Shigeta M., and Tesfaye, S.. “The “tree Against Hunger”. Enset-Based Agricultural Systems in Ethiopia”. American Association for the Advancement of Science with Awassa Agricultural Research Center, Kyoto University Center for African Area Studies and University of Florida, Washington, DC, USA, 66 pp. 1997.

[14] Tariku, H. and Mogessie, A. “Effect of altitude on microbial succession during traditional enset (Ensete ventricosum) fermentation”. International Journal of Food, Nutrition and Public Health, 4(1): 39-51. 2011.

[15] Belay, A., Hunduma, T. Fekadu, E., Negisho, K. and Ali, A. “Gender-based analysis of production system in ambo”. pp. 71-87. In: Chiche, Y. and Kelemu, K. (eds.). Proceedings of the workshop on gender analysis in agricultural research held in Addis Ababa, Ethiopia, 27-29 November 2006. 2008.

[16] Tariku, H. “Effect of altitude on microbial succession during traditional Enset (Ensete


77

NCSTI-2012

ventricosum) fermentation”. MSc thesis. Addis Ababa University, Ethiopia. pp 89. 2008.

[17] Kelbessa, U., Ayele, N. and Melaku, U. “Traditional enset-based foods: survey of processing techniques in Sidama”. In: Tsedeke Abate, Hiebsch C, Brandt S. A. and Seifu Gebremariam (eds). Enset-Based Sustainable Agriculture in Ethiopia. Proceedings from the International Workshop on Enset held in Addis Ababa, Ethiopia, 13-20 December 1993.

[18] Liyuwork, Z. “ Kocho processing in southern and southwestern Ethiopia”. In: Tsedeke Abate, Hiebsch C, Brandt S. A. and Seifu Gebremariam (eds). Enset-Based Sustainable Agriculture in Ethiopia. Proceedings from the International Workshop on Enset held in Addis Ababa, Ethiopia, 13-20 December 1993.

[19] Sandford, J. and Helen, K. “The effect of gender on resource contribution, decision making and influence: A comparison between enset, tef and maize”. In: Tsedeke Abate, Hiebsch C, Brandt S. A. and Seifu Gebremariam (eds). Enset-Based

Sustainable Agriculture in Ethiopia. Proceedings from the International Workshop on Enset held in Addis Ababa, Ethiopia, 13-20 December 1993.

[20] Seifu, G. “Enset research in Ethiopia: 1976-1984”. In: Tsedeke Abate, Hiebsch C, Brandt S. A. and Seifu Gebremariam (eds). Enset-Based Sustainable Agriculture in Ethiopia. Proceedings from the International Workshop on Enset held in Addis Ababa, Ethiopia, 13-20 December 1993.

[21] Firew k. “Research experiences on the development of enset processing equipment”. (Inpress)

[22] Nejibe M. “Environmental impact of ‘katikala’ production in arsi-negele woreda, central rift valley of Ethiopia”. MSc thesis. Addis Ababa University, Ethiopia. pp 88. 2008.

[23] www.Zelaleminjera.com) [24] Endalew, A. “Socioeconomic analysis of katikala

production and consumption in arsi-negele woreda of oromia region, ethiopia”. MSc thesis. Addis Ababa University, Ethiopia. pp 95. 2008.

[25] ዶ/ር ወሮታው በዛብህ፡፡ ኢንተርፕረነርሽፕ ራስንና ሀገርንለማበልፀግ ድብቅና ልዩ ችሎታዎን ፈልገው በማግኘትእንድጥሩና ስከታማ እንድሆኑ የሚያግዝ መጽሐፍ፡፡ ጂኒየስየስልጠናና የምክር አገልግሎት ማዕከል፡፡ አዲስ አበባ ፤ኢትዮጵያ፡፡ 142 ገጽ ፡፡ መጋቢት 6 ቀን 1998 ዓ.ም

[26] ኢምፐርቴክ ኢትቶጵያ ቢዝነስ ሶሳይት (ኢ.ኢ.ቢ.ሶ)፤ አዲስአበባ ፤ ኢትዮጵያ፡፡ የብልጥፅግና ቁልፍ፡፡ 184 ገጽ፡፡ 1999 ዓ.ም

[27] Reichert, R.P., and YOUNG, C.C. “Dehulling cereal grains and grain legumes for developing

countries. II. Chemical composition of mechanically and traditionally dehulled sorghum and millet”. Cereal chem. 54:74. 1977.

[28]Fellows, P.J. and Axtell, B. (Eds). Setting up and running a small flour mill or bakery. Opportunities in food processing series. Wageningen: ACP-EU Technical Center for Agriculture and Rural Cooperation (CTA), 2004. 250 pages, ISBN 92-9081-276-1

[29] “Tadias Megazine”. May 6th, 2007


78

NCSTI-2012


Migbar Assefa Lecturer, Department of Mechanical and Industrial Engineering, Technology Institute,

Hawassa University, P.O.Box 05, Hawassa, Ethiopia, email: [email protected]

Abstract— Fatigue induced crack growth in mechanical components is a very complicated phenomenon and the major problems in the prediction and maintenance of mechanical structures. Fatigue failure usually occurs due to the initiation and propagation of surface or near-surface cracks, which are frequently assumed to be elliptical or semi-elliptical in shape for numerical modelling. Closed form solutions for the stress intensity factors (SIFs) are available for simple crack geometries in three dimensions; however, for arbitrary-shaped cracks in finite specimens, numerical methods are the only option to model three-dimensional fatigue crack growth. This paper presents current issues and developments related to Extended Finite Element Method (X-FEM) by reviewing recently available literatures. The review will give more emphasis on fatigue crack growth modelling of engineering materials using X-FEM and partition of unity finite element method (PUFEM) which help in encapsulating various research outcomes in a structured manner. In this review the following key questions will be addressed: what has been done so far in this area of study, what are the recent progresses towards the subject and what are the future trends and research potentials related to the topic.

Keywords- Review, X-FEM, Meshfree Methods, Fracture Mechanics

I. INTRODUCTION

Fatigue crack growth prediction is still very much an empirical art than a science, despite being a relatively old subject having nearly one and half centuries of history described as in a number of books and review papers. In the early days, SN curves were used to design fail-safe structures for infinite life. As more economic and reliable design is required, many comparative studies have been performed to justify upgrading of current fatigue design S-N curves, to provide a better basis for design, or to propose improved methods [24]. This is because the empirical constants of SN curves are derived from constant amplitude cyclic tests and hence are not representative of the random spectrum load that many mechanical structures are exposed to [23]. Moreover, it has been determined that the experimental fatigue lives of specimens and components subjected to random amplitude loading can be well below the fatigue lives predicted by the SN data. Hence a better design strategy is required.

With the introduction of fracture mechanics,

fatigue crack growth comes out to be one of the major concerns in the development of the damage tolerance concepts in the aerospace industries. During the last few decades, numerous papers have been published on fatigue life and fatigue crack growth prediction under constant and variable amplitude loading. Paris, Gomez and Anderson [26] introduced stress intensity factor for the correlation between the crack growth rate, da/dN, and the stress

intensity factor range, ΔK. This paper was considered as one of the milestones in this area. In this work the authors adopted the K-value from the analysis of the stress field around the tip of a crack as proposed by Irwin in 1957.

Advanced engineering systems in recent days

demand the use of computer aided tools. In such tools, computational simulation techniques are often used to model and investigate physical phenomena in engineering. The simulation requires solving the complex differential equations that govern the phenomena. Usually, such complex differential equations are largely solved using numerical methods, like Finite Element Method (FEM) and finite difference methods (FDM). In these methods, the spatial domain where the partial differential governing equations are defined is often discretized in to meshes.

The finite element method (FEM) is certainly

one of the most popular and powerful numerical tool for studying the behavior of a wide range of engineering problems. One of the important applications of FEM is the analysis of fracture problems like, crack propagation. Fundamentals of the present form of the linear elastic fracture mechanics (LEFM) came to the existence practically in naval laboratories during the First World War [31]. Since then, LEFM has been successfully applied to various classical crack and defect problems, but remained relatively limited to simple geometries and loading conditions.


79

NCSTI-2012

As compared to the standard finite element method, X-FEM enables improved approximations of non-smooth solutions such as those including jumps, kinks, and singularities. This is achieved by providing a concept for enriching the approximation space such that a priori known solution properties are included. Typical applications of the X-FEM are found in fracture mechanics, biomechanics, and two-fluid flows. Many current research activities extend the application fields of the X-FEM and investigate new formulations and special issues of enriched approximations [38]-[39].

X-FEM enables the accurate approximation of solutions that involve jumps, kinks, singularities, and other locally non-smooth features within elements. This is achieved by enriching the polynomial approximation space of the classical finite element method. X-FEM focused on methods for problems with discontinuous gradients, singularities, and other non-smooth features. These methods are of particular importance for this class of problems because they tremendously facilitate their solution: they avoid the need for the finite element mesh to conform to these features, and for evolutionary problems they avoid the need for remeshing [38].

II. X-FEM-BASIC REVIEW While the finite element method is well

developed and robust, it is not particularly well suited to model evolving discontinuities. As a specific instance of a Galerkin method, it relies on an element structure to construct an approximating space. The construction of a discontinuous space with finite elements requires alignment of the element topology with the geometry of the discontinuity. This in turn demands a regeneration of the mesh as the discontinuity evolves, resulting in projection errors and a considerable computational cost [9]-[10].

Over the course of the past several years, much

attention has been focused on developing new approximations in the Galerkin framework which do not require a mesh for their construction, this method is known as meshfree method. The concepts in meshfree methods can be traced to the Smoothed Particle Hydrodynamics (SPH) method and element-free Galerkin (EFG) method [3].

Meshfree methods have been applied to

problems in applied mechanics for which the use of a traditional finite element mesh presents significant difficulties. These include the modelling of quasi-static and dynamic fracture in solids by both the EFG and SPH methods.

Both quasi-static and dynamic fractures are

local phenomena, in that the crack geometry is

confined to a relatively small portion of the domain. An alternative to meshfree approximations concerns the development of finite elements which are capable of representing inter-element discontinuities. These elements are then used locally about the crack geometry to represent the arbitrary discontinuity.

Ted Belytschko et al [4] presented an overview

of meshfree approximations based on moving least-squares, kernels and partitions of unity. It is shown that the three methods are in most cases identical except for the important fact that partitions of unity enable p-adaptivity to be achieved. Methods for constructing discontinuous approximations and approximations with discontinuous derivatives are also described. The authors also pin pointed several issues in implementation and fast ways of constructing consistent moving least-square approximations. In this work many of the advantages of both meshfree and assumed strain methods can be realized in the partition of unity framework, where local enrichment functions are incorporated into the approximation in a straightforward fashion. This framework is viewed to be more flexible than the assumed strain approach, in that it involves a direct modification to the displacement approximation.

The basic mathematical foundation of the

partition of unity finite element method (PUFEM) was discussed by Melenk and Babuska [20]. They illustrated that PUFEM can be used to employ the structure of the differential equation under consideration to construct effective and robust methods. The global solution of PUFEM has been the theoretical basis of the local partition of unity finite element method, to be called later the extended finite element method.

The term Extended Finite Element Method (X-FEM) has been originally proposed by Belytschko and Black [5] some 12 years before, in this paper the authors adopted the partition of unity concept to model crack growth, by locally enriching a finite element approximation with the exact near tip crack fields. A key feature of this work was the incorporation of discontinuous enrichment functions, and the use of a mapping procedure to model arbitrary discontinuities. The incorporation of the near-tip fields also provided for accurate stress intensity factor calculations on relatively coarse meshes. Unfortunately, the mapping technique is awkward for long cracks, and the asymptotic near-tip functions are not the best choice to represent the displacement discontinuity far from the crack tip. As a result, some level of remeshing is implemented as the crack grows.

Later Dolbow et al. [10]-[11] and Moes et al.

[21] improved the method and called it the extended


80

NCSTI-2012

finite element method (X-FEM) by introducing a wonderful technique called enrichment that includes the asymptotic near tip field and a Heaviside function H(x). Since then, the X-FEM has gained great attention and interest from the scientific community in the different arenas of engineering.

N. Sukumar et al, [34] extended the asymptotic

near tip field and a Heaviside function proposed by Dolbow et al [10] and Moes et al [21]. In this work Sukumar and his collogues described an Extended Finite Element method (X-FEM) for three-dimensional crack modelling and arbitrary branched and intersecting cracks. By adding special functions, finite element space has been enriched to account for the crack by using the notion of partition of unity. Stolarska et al. [33] represented the crack location, including the location of crack tips by using the level set method (LSM) and the extended finite element method to compute the stress and displacement fields necessary for determining the rate of crack growth. This combined method requires no remeshing as the crack progresses, making the algorithm very efficient and also the level sets enable to address the most important X-FEM issues “Where and how to enrich?” and they simplify implementation aspects of the X-FEM.

N. Moes et al, [22] developed a method for the

analysis of arbitrary cracks in three-dimensional bodies. The method developed by the authors combined the extended Finite element method, which constructs arbitrary discontinuities through a discontinuous partition of unity with level set methods.

The method used by Moes and his collogues was based on the extended finite element method, in which the crack discontinuity is introduced as a Heaviside step function via a partition of unity. To improve the accuracy of this method, the authors introduced branch functions with the inclusion of asymptotic near-tip fields for all elements containing the crack front. In addition to Moes et al, Ted Belyteschko et al, on their review, core enrichment functions are not readily available for problems involving nonlinear or anisotropic materials; such problems can be modeled by using only the step function enrichment.

B.L. Karihaloo and Q.Z. Xiao [7] thoroughly

reviewed the recently developed techniques for modelling cracking within the finite element (FE) framework which use meshes independent of the crack configuration. The authors addressed traditional FE method with the partition of unity method for modelling individual cracks, intersecting or branching cracks, as well as cracks emanating from holes or other internal interfaces. In this work the authors also discussed in detail about numerical integration for the enriched elements, linear

dependence and the corresponding solution techniques for the discretized system of equations, as well as the accuracy of the crack tip fields.

N. Sukumar et al. [35] proposed novel

numerical paradigm for three-dimensional crack propagation of planar cracks. This new technique couples the X-FEM to the fast marching method (FMM) the later adopted a second order upwind finite difference scheme. In the X-FEM, a discontinuous function and the two-dimensional asymptotic crack-tip displacement fields were added to the finite element approximation to account for the crack using the notion of partition of unity. In this work, the planar three-dimensional crack was represented by two level set functions: one for the crack front and the other for the crack plane. For three-dimensional crack modelling, a discontinuous function was used to model the interior of the crack surface, and functions from the two-dimensional asymptotic crack-tip displacement fields were used for the crack front enrichment. These enrichment functions were added to the finite element approximation within the context of a displacement-based Galerkin formulation. The authors also addressed the computational geometry issues associated with the representation of the crack and the enrichment of the finite element approximation and stress intensity factors (SIFs).

Pedro M.A. Areias, Ted Belytschko [27] developed a pragmatic approach to the extended finite element method for general three-dimensional problems. During the course of numerical testing, it was found that despite the known accuracy defects of the underlying element, the softening and cracking results are very sharp and quite accurate, even with coarse meshes. Shaofan Li and Cerup B. Simonsen [30] used meshfree method to simulate ductile crack growth and propagation under finite deformation and large scale yielding conditions. In this work the authors developed a new parametric visibility condition and its related particle splitting procedure to simulate ductile crack propagation in three dimensional objects by automatically adapting the evolving strong continuity or fracture configuration due to an arbitrary crack growth in the materials. The numerical computations show that the proposed procedure can accurately simulate crack growth and propagation in ductile materials undergoing finite deformation and large scale yielding.

Stephane Bordas and Brian Moran [32]

described how X-FEM coupled with level set methods can be used to solve complex three dimensional industrial fracture mechanics problems through combination of an object-oriented research code and a commercial solid modelling/finite element package. This paper briefly described how object-oriented programming shows its advantages


81

NCSTI-2012

to efficiently implement the proposed methodology. The methodology presented by the authors was successfully applied to a real world aerospace component, which proved its simplicity and accuracy and believed to be a promising tool for damage tolerance assessment of complex three-dimensional components.

T. Rabczuk et al, [37] reviewed different crack

tracking techniques in three-dimensions applicable in the context of partition of unity methods, particularly mesh free methods. Detailed descriptions of a local crack tracking procedures have been simulated in this paper. A crack tracking procedure was proposed in detail and implemented in the context of the extended element-free Galerkin method (XEFG). In this work several three-dimensional cracking examples were compared to other results from the literature or the experimental data and show good agreement.

Jeong Hoon Song [16] exploited the versatility

of Extended Finite Element Method to simulate dynamic fracture in plane and thin shell structures to solve quazi-brittle fracture problem. The method is implemented by using 4-node quadrilateral plane and Belytschko-Lin-Tsay shell element, which have high computational efficiency. This method is applied to simulate several experiments involving dynamic fracture and nonlinearities. In this PhD work the author demonstrated that the failure modes in the experiments are good agreement with the proposed method and can be used for general applications of dynamic fracture.

Haim Waisman [15] proposed a combined

analytical approach with the extended finite element method (X-FEM) to extract the Strain Energy Release Rates within the classical stiffness derivative technique. The author presented two ideas, in the first case the crack is mesh independent and in the second case the asymptotic crack tip field is embedded in the mathematical formulation of the stiffness matrix. By employing these properties the authors showed that the derivative of the stiffness matrix with respect to the crack extension has been computed in a closed form and on the fly during the analysis. Thus the virtual crack extension, and the error inherent in the finite difference scheme of the classical stiffness derivative technique is completely avoided.

Yazid Abdelaziz et al, [1]-[40] presented a

detailed overview and recent progress of the extended finite element method X-FEM in the analysis of crack growth modeling. The authors also summarized the important milestones achieved by the finite element community in the arena of computational fracture mechanics.

S. Natarajan et al [29]-[36] proposed the new numerical integration technique to integrate the discontinuous and singular integrands appearing in the X-FEM stiffness matrix. The proposed method eliminates the need to sub-divide elements cut by strong or weak discontinuities or containing the crack tip. In this paper few examples with known analytical solution from linear elastic fracture mechanics has been illustrated to check the effectiveness of the proposed method is illustrated. It is seen that with mesh refinement, both integration techniques provide convergence of the SIFs to the analytical SIFs.

Thomas-Peter Fries and Ted Belytschko [38]

presented an overview of the extended/generalized finite element method with emphasis on X-FEM methods for discontinuities and high gradients. The authors examined the similarities of PUM, GFEM and X-FEM, along with some of the differences in the early literature.

D. Motamedi and S. Mohammadi [8] adopted

extended finite element method for modelling the crack and analyzing the domain numerically. Orthotropic crack-tip (near-tip) enrichment functions, which can be applied to all types of orthotropic materials, are implemented. The authors used Heaviside and near-tip enrichment functions in the framework of the partition of unity for modelling crack discontinuity and crack-tip singularities within the classical finite element method, as proposed by N. Moes et al [22]. In this procedure, elements that include a crack are not required to conform to crack edges. In this work, mixed-mode stress intensity factors (SIFs) were determined based on evaluation of the dynamic J-integral, to determine fracture properties of the domain and it is found that results obtained by the proposed method, are in good agreement with other numerical methods.

Jian-Ying Wu [17] presented a unified

theoretical framework of enriched finite elements for modelling cohesive cracks using the variational multiscale method. In this work the kinematics (i.e. coarse and fine scale displacement and strain fields) and statics (i.e. coarse and fine scale equilibrium equations) are thoroughly investigated in both continuum and discrete settings.

E. Giner, et al [12] presented modelling of

LEFM problems that imply crack face closure and contact using the extended finite element method (X-FEM) aiming at its application to fretting fatigue problems. An assessment of the accuracy in the calculation of KII is performed for two different techniques to model crack face contacts in X-FEM. The first strategy consists of including additional standard 1-D elements along the crack faces to


82

NCSTI-2012establish the contact using a standard algorithm. The extra nodes corresponding to the 1-D elements are linked to the underlying X-FEM displacement interpolation by means of constraint equations between the degrees of freedom. The second approach is much more elaborated and accurate than the former and involves the satisfaction of the contact conditions in an integral sense under a so-called segment-to-segment approach along the crack faces. The authors implemented both strategies in the framework of the commercial code Abaqus, so as to permit its practical application to fretting-fatigue problems and authors found that segment-to-segment approach leads to optimal convergence rates of the error in KII.

E. Giner, et al [13] presented an efficient

procedure to predict fatigue lives in fretting fatigue problems. This is accomplished by means of a combined initiation–propagation approach in which the extended finite element method (X-FEM) is used. The authors verified this procedure by modelling several fretting fatigue tests available in the literature. The application of the X-FEM enables to numerically evaluate the stress intensity factors (SIFs) for cracks of different lengths emanating at the end of the contact zone and to estimate the propagation life corresponding to each of the tests.

This propagation life is combined with the initiation life calculated using a multiaxial fatigue criterion.

The predicted lives are then compared with the

reported experimental lives, showing that the consideration of the crack-contact interaction through the numerical models tends to improve the life estimation when compared with a fully analytical approach. The procedure can be applied to more general fretting problems for which analytical solutions are not available. For a series of experimental tests reported in the literature, results show that the use of X-FEM to predict the propagation phase tends to improve the life estimation when compared to the weight function method. The study shows that the proposed methodology provides consistent results. The authors suggested that this procedure can be applied to fretting configurations found in engineering practice for which analytical solutions are usually not available.

T.P. Fries and M. Baydoun [38] have presented

a new description of cracks in the frame of the X-FEM that maintains the advantages of mostly separated explicit or implicit crack descriptions. The approach used by the authors is applicable in two and three dimensions. S. E. Mousavi, E. Grinspun and N. Sukumar [28], analyze complex crack problems in elastic media using harmonic

enrichment functions in a higher order extended finite element implementation. The numerically computed enrichment function of a crack was the solution of the Laplace equation with discontinuous Dirichlet boundary condition along the crack, and its interaction with branches or other cracks is realized by imposing vanishing Neumann boundary conditions along those cracks. In this paper a nested subgrid mesh has been presented in the Laplace solve with a rasterized approximation of a crack, which simplifies the numerical integration no partitioning of finite elements is required.

M. Baydoun and T. P. Fries [19], studied

propagation criteria in three dimensional fracture mechanics within the Extended Finite Element framework. The crack in this paper is described by a hybrid explicit-implicit approach as proposed by T.P. Fries and M. Baydoun in [39]. In this approach, the crack update is realized based on an explicit crack surface mesh which allows an investigation of different propagation criteria. In contrast, for the computation of the displacements, stresses and strains by means of the X-FEM, an implicit description by level set functions has been employed.

III. CONCLUSION: This paper dedicated to the review of newly

developed extended finite element method (X-FEM) by giving more emphasis on fatigue crack growth modelling of engineering materials. The study shows X-FEM has great potential to solve crack growth with arbitrary and complex paths, as the crack growth being described independently of the mesh. This is one of the main advantages of the method that it avoids any need for remeshing or geometric crack modelling in numerical simulation, while generating discontinuous fields along a crack and around its tip. The combination of X-FEM with level sets considerably enhanced the capability of X-FEM, since this makes possible the description of cracks. In addition this, X-FEM enables one to control the powerful level set methods that have been developed to track the evolution of cracked surfaces

IV. REFERENCES [1] Abdelaziz Yazid, Nabbou Abdelkader, Hamouine Abdelmadjid, “A state-of-the-art review of the X- FEM for computational fracture mechanics,” Applied Mathematical Modelling, 33, 4269–4282, 2009 [2] Belytschko, T., J. Fish, and B. Engleman, “A Finite element with imbedded localization zones,” Computer Methods in Applied Mechanics and Engineering 70, 59-89, 1988. [3] Belytschko, T., Y. Y. Lu, and L. Gu, “Element-free Galerkin methods,” International Journal of


83

NCSTI-2012 Numerical Methods in Engineering 37, 229-256, 1994. [4] Belytschko T., Y. Krongauz, D. Organ, M. Fleming and P. Krysl, “Meshfree methods: An overview and recent developments,” Computer Methods in Applied Mechanics and Engineering, 139, 3-47, 1996. [5] Belytschko, T. and T. Black, “Elastic crack growth in Finite elements with minimal remeshing,” International Journal of Numerical Methods in Engineering 45 (5), 601-620, 1999. [6] Belytschko, T, Robert Gracie and Giulio Ventura, “A Review of Extended/Generalized Finite Element Methods for Material Modelling,” Department of Mechanical Engineering, Northwestern University, USA. [7] B.L. Karihaloo and Q.Z. Xiao, “Modelling of stationary and growing cracks in FE framework without remeshing: a state-of-the-art review,” Computers and Structures 81, 119–129, 2003. [8] D. Motamedi, S. Mohammadi, “Dynamic analysis of fixed cracks in composites by the extended finite element method,” Engineering Fracture Mechanics 77, 3373–3393 2010. [9] Dolbow, J.E., “An extended finite element method with discontinuous enrichment for applied mechanics,” PhD dissertation, Theoretical and Applied Mechanics, Northwestern University, USA, 1999. [10] Dolbow, J. and T. Belytschko, “Numerical integration of the Galerkin weak form in meshfree methods” Computational Mechanics, 23, 219-230, 1999. [11] Dolbow, J., N. Moes, and T. Belytschko,“ Discontinuous enrichment in Finite elements with a partition of unity method,” Finite Elements in Analysis and Design 36, 235-260, 2000. [12] E. Giner, M. Sabsabi, F.J. Fuenmayor , “Calculation of KII in crack face contacts using X-FEM,” Application to fretting fatigue, Engineering Fracture Mechanics 78, 428–445, 2011, a. [13] E. Giner, C. Navarro, M. Sabsabi, M. Tur, J. Dom´nguez, F.J. Fuenmayor, “Fretting fatigue life prediction using the extended finite element method,” International Journal of Mechanical Sciences, 53 217–225, 2011, b [14] Gravois A., Moes N., and Belytschko T, “Non- planar 3D crack growth by the extended finite element and level sets Part II: Level set update,” International Journal for Numerical Methods in Engineering, 53 : 2569- 2586, 2002. [15] Haim Waisman, “An analytical stiffness derivative extended finite element technique for extraction of crack tip Strain Energy Release Rates,” Engineering Fracture Mechanics 77 3204–3215, 2010. [16] Jeong Hoon Song, “Computations of the Dynamic Fracture of Quasi-Brittle Plane and Shell Structures by the Extended Finite Element Method,” PhD dissertation, Theoretical and Applied Mechanics, Northwestern University, USA, 2008. [17] Jian-Ying Wu, “Unified analysis of enriched finite elements for modelling cohesive cracks,” Computer

Methods in Applied Mechanics and Engineering, 200 3031–3050, 2011. [18] Liu, Gui-Rong, Mesh Free Methods Moving Beyond the Finite Element Method CRC Press, 2003. [19] M. Baydoun and T.P. Fries, “Crack propagation criteria in three dimensions using the X-FEM and an explicit- implicit crack description,” submitted to International Journal of Fracture, 2012. [20] Melenk, J. M. and I. Babuska, “The partition of unity Finite element method: Basic theory and applications,” Computer Methods in Applied Mechanics and Engineering 39, 289-314, 1996. [21] Moes, N., Dolbow, J. and Belytschko T, “A finite element method for crack growth without remeshing,” International Journal of Numerical Methods in Engineering, 46, 131-150, 1999. [22] Moes N. Gravouil A., and Belytschko T., “Non- planar 3D crack growth by the extended finite element and level sets Part I: Mechanical model,” International Journal for Numerical Methods in Engineering, 53: 2549- 2568, 2002. [23] Mulugeta Haile, Tzi-Kang Chen, Michael Shiao, and Dy Le, “Crack growth behavior in preloaded metallic nested-angle plates under flight load spectrum” Experimental and Applied Mechanics, Volume 6, Proceedings of the 2011 Annual Conference on Experimental and Applied Mechanics series, Volume 9999, 3-11, 2011. [24] N. Pugno et al., “A generalized Paris’ law for fatigue crack growth,” Journal of mechanics and physics of Solids, issue 54, pp. 1333–134954, 2006. [25] O.O. Ochoa and J.N. Reddy, Finite Element Analysis of Composite Laminates, Kluwer Academic Publishers, the Netherlands, ISBN-7923-1125-6, 1992. [26] Paris P. C., Gomez M. P. and Anderson W. E., “A rational analytical theory of fatigue”. The Trend of Engineering, 13, PP. 9-14, 1961. [27] Pedro M.A. Areias, Ted Belytschko, “Analysis of three-dimensional crack initiation and propagation using the extended finite element method,” International Journal for Numerical Methods in Engineering, 63: 760–788, 2005. [28] S. E. Mousavi, E. Grinspun and N. Sukumar “Harmonic enrichment functions: A unified treatment of multiple, intersecting and branched cracks in the extended finite element method,” International Journal for Numerical Methods in Engineering , 2010; 00:1–32. [29] S. Natarajan, D. R. Mahapatra, S. Bordas, “Integrating strong and weak discontinuities without integration subcells and example applications in an X-FEM/GFEM framework,” International Journal for Numerical Methods in Engineering, 83(3), 269–294, 2010. [30] Shaofan Li and Cerup B. Simonsen, “Meshfree Simulations of Ductile Crack Propagations,” International Journal for Computational Methods in Engineering Science and Mechanics, 6:1–19, 2005. [31] Soheil Mohammadi, Extended Finite Element Method for Fracture Analysis of Structures, Blackwell/ Wiley Press (UK), 2008.


84

NCSTI-2012 [32] Stephane Bordas and Brian Moran, “Enriched finite elements and level sets for damage tolerance assessment of complex structures,” Engineering Fracture Mechanics Volume: 73, Issue: 9, Pages: 1176-1201, 2006. [33] Stolarska, M., Chopp, D.L., Moës, N. and Belytschko, T. “Modelling crack growth by level sets in the extended finite element method,” International Journal for Numerical Methods in Engineering, 51, 943–960, 2001. [34] Sukumar, N., Moës, N., Moran, B. and Belytschko, T. “Extended finite element method for three dimensional crack modelling,” International Journal for Numerical Methods in Engineering, 48,1549–1570, 2000. [35] Sukumar, N., Chopp, D.L. and Moran, B., “Extended finite element method and fast marching method for three-dimensional fatigue crack propagation”, Engineering Fracture Mechanics, 70, 29–48, 2003. [36] Sundararajan Natarajan, “Enriched Finite Element Methods: Advances & Applications”, PhD Thesis, Institute of Mechanics and Advanced Materials, Theoretical and Computational Mechanics, Cardiff, Wales, U.K.2011. [37] Timon Rabczuk, Stéphane Bordas, Goangseup Zi, “On three-dimensional modelling of crack growth using partition of unity methods,” Computers and Structures, Vol. 88, issue 23-24, pp 1391-1411, 2008. [38] T. P. Fries and Ted Belytschko, “The extended/generalized finite element method: An overview of the method and its applications,” International Journal for Numerical Methods in Engineering, Volume 84, Issue 3, pages 253– 304, 2010. [39] T.P. Fries and M. Baydoun, “Crack propagation with the X-FEM and a hybrid explicit-implicit crack description,” International Journal of Numerical Methods in Engineering, page early view. (2011) [40] Yazid Abdelaziz, Abdelmadjid Hamouine,“A survey of the extended finite element,” Computers and Structures 86, 1141–1151, 2008.

Design of Molding Die for Shoe Sole

85

NCSTI-2012

Design of Molding Die for Shoe Sole Mesfin Seid Ibrahim1

Department of Industrial Engineering, Ethiopian Institute of Technology-Mekelle, Mekelle University,

P.o.Box 3040, E-Mail: [email protected], Mekelle,Ethiopia.

Abstract — The mold and part design of plastic parts for injection molds is a complicated process. This paper presents the design of plastic injection mold for producing molded shoe soles using injection molding machines. It covers from design of a shoe sole to modeling and analysis of molding die using cutting age modeling and simulation software. The paper starts from the scratch design of shoe sole bottom pattern by selecting appropriate sizing system and footwear style, and goes on to rib design and finally fully featured shoe sole design using CATIA®. After the shoe sole is designed, part simulation analysis is undertaken to explore gate location analysis, fill and pack analysis and cooling quality analysis using Moldflow® Adviser Software. Getting all results accepted in part analysis phase, the mold for the designed shoe sole is designed by considering the melt delivery systems design using Autodesk Inventor Software.The results of the study enable us to understand that mold design is one of the expensive design areas that need virtual prototyping prior to producing the actual prototype. Thus, all the results showed us that CAD software is vital, while CAE software enables to predict and overcome problems.

Keywords- Plastic Injection Molding, Mold Design, Shoe Sole, CATIA, Autodesk Moldflow Adviser, Autodesk Inventor Professional.

I. INTRODUCTION

Nowadays, the technology of the tool and die fabrication in plastic injection is one of the world's fastest growing industries. Plastic has, quite literally, become the cornerstone of our society. We make so many things from plastic that is hard to imagine what our lives would be like if it was never invented. Plastic is now used in almost every application, ranging from household articles to space travel, from transportation to packing, from medicine to toys, from bridge building to sports.

According to shoe sole website www.jayelco.com, molded soles are made from rubber or rubber combined with other materials. In this paper shoe soles that can be manufactured only using injection molding are under consideration.

Generally, injection molding is a process that forms the plastic into a desired shape by melting the plastic material and forcing the plastic material under pressure into the mold cavity. The shape of the plastic that is desired is achieved by cooling in thermoplastic or by chemical reaction for thermosetting [1].

In order to build a low cost and processable mold for thermoplastic product, it requires deep knowledge in polymer processing eld particularly in rheology studies. Thus, many polymer product manufacturers take easier approaches by undergoing trial and error method in the early stages of manufacturing new product design. Most of the time the results are quite satised, but it is too time consuming. Consequently, small and medium scale manufacturers are unlikely to introduce new

products in short time. Some commercial Computer Aided Engineering (CAE) software designers have taken the opportunity to develop injection molding ow analysis simulation software to help predict and overcome the problems [2].

Several professional CAE software in injection molding are available in the market such as Moldow® Adviser and Moldex3D®. The injection molding simulation software helps to study the ow patterns of polymer melt inside the mould during injection, packing and cooling processes [3]. The output results can be used as guidance to design mold with correct operating parameters and the most important is the cost expenses in building a mold.

Mold design plays the most important roles to produce good quality products either in mechanical properties or appearance. However, mold design and fabrication is a costly and high technology process because it uses Computer Aided Engineering (CAE) software to analyze and simulate the plastic parts, Computer Aided Design (CAD) software to design the complicated plastic product and Computer Aided Manufacturing (CAM) to do the programming fabrication to run the computer numeric control (CNC) for milling or lathe [1].

II. METHODOLOGY

A. Shoe Sole Sizing system and Style Selection There are various approaches to design a mold for

a shoe sole. In this paper the general procedure followed in designing shoe sole mold is illustrated


86

NCSTI-2012

Moreover, the French size system (Paris Point) detail in comparison with British and American sizing systems is presented in Table 1.

Table 1. Sizing Systems Comparison

So taking 41 pp of the French size system as mean

size for male sole size the detail dimensions of concern are calculated as follows.

The standard insole length for the selected 41 pp is 41 pp*2/3 = 27.33 cm = 273.33 mm as shown in Fig. 2 above. In addition, the ball width is about 92.2 mm as indicated in professional shoe fitting [6].

After the standard geometry of the shoe sole is determined, the shoe sole CAD model is developed as shown in Fig. 5. The shoe sole CAD model is then be used in the part simulation software, Autodesk Moldflow® Advisor, to study the gate location, fill + pack analysis and the cooling quality analysis results. B. Detail analysis of the machine, mold and plastic

materialAfter the CAD model of the shoe sole is

developed, the part analysis is carried out on Moldflow® Advisor simulation software. The machine that the mold is going to be designed for is static injection machine for the production of one-color soles model SP 345-3 EXT manufactured by MAIN GROUP s.p.a (Italy). Table 2. Injection molding machine specification

Parameters Total Weight 5500 kg

Stations and Injectors No 3 independent stations and 3 resp.

Mold Closing (Clamp) force

600 KN = 66 ton

Press Opening Stroke 210mm Mold-Holder Dimensions

343*400 mm

Standard Mold Dimensions

300*400 mm

Plasticization capacity of each injector

45Kg/h

Injection Volume 750 cm3 Screw diameter 66 mm Screw Ratio(l/d) 14:1 Injection Pressure 300 bar = 30 MPa


87

NCSTI-2012

Table 3. Properties of a typical Aluminum

Chemical Composition Composition (wt.-%) A.A

No. Cu Cs Fe Mg Mn Si Ti Zn Others 3,4365

1.2-2.0 0.18-0.28 0.5 2.1-2.9 0.3 0.4 0.2 5.1-6.1 Ti-Zr = 0.25

Physical Properties Density Yield

Strength Ultimate Tensile Strength

Modulus of Elasticity

Coefficient of thermal Elasticity

Thermal Conductivity

A.A. No.

Kg/cm3 N/mm2 N/mm2 MPa W/(moC) W/(K.m) 3,4365

2.82 410-530 480-610 72,000 23.7 153(0-100oC)

The most common type of material for shoe sole

production with the above machine is A.A 3,4365 Aluminum [7] because of its suitability for this purpose as shown in Table 3. Out of the limited number of plastic materials for shoe sole production (Rubber, Thermoplastic Rubber, Light Thermoplastic Rubber, Polyvinyl Chloride, Light Polyvinyl Chloride, Thermo Polyurethane, Polyurethane) the consideration depends on the machine type, the mold material type and processing conditions.

Thus, considering the SP 345-3 EXT one-color injection molding machine 0 and AA 3, 4365 Aluminum mold material, the most suitable plastic materials is found to be Thermoplastic Rubber (TPR/E)as shown in Table 4. C. Part Simulation Analysis Results

The analysis sequence is the predefined sequence of analyses that we want to run and it depends on the analysis technology (dual domain or 3D) that is used, and the molding process that is selected. The analysis technologies that are available depend on the analysis type that is selected [9].

There are many types of analysis for most of the injection molding processes ranging from design adviser analysis to sink mark analysis. But for the purpose of practicality, frequently used analysis types on Autodesk® Moldflow® Adviser are:

Gate Location Analysis

Fill and Fill+Pack Analysis Cooling Quality Analysis

These four types of analyses have can be seen in detail with various types of analyses which are described in the consecutive parts.

Table 4. TPR Property

Gate Location Analysis

The Gate Location analysis is used to recommend

injection locations for the shoe sole. The gate region locator algorithm determines and recommends suitable injection locations based on criteria such as the part geometry, flow resistance, thickness, and molding feasibility and it produces the gate location analysis result. This result rates each place on the shoe sole CAD model for its suitability as an injection location and is used as a preliminary input for a full Fill + Pack analysis [9].

Fig. 6 shows results of the shoe sole gate location analysis from Moldflow® Adviser. It shows that the most suitable and recommended gate location is the center region of the part indicated with reddish color.

The four points compares the suitability of locations in the recommended region, and as can be seen point 4 is the most suitable location with 99.5% of degree of suitability. Fill Time

The Fill time result shows the position of the flow front at regular intervals as the cavity fills; it is also to show how the plastic material flows to fill the mold. From that we know that the short shot (part of the model which did not fill) part will be displayed.

Material TPR /TPU/TPO Manufacturer Tonen Chemicals Density 1.0404/1.2182 g/cc Melt Temperature 215 oC Mold Surface To 45 oC Pressure Max 180 MPa


88

NCSTI-2012

From that result one can also understand how the weld line and air trap will form [1].

Fig. 7 shows the plastic fill time result for the

entire part starting from the gate location region indicated in blue to other extremes with red color. Thus, the shoe sole will be filled fully in about 2.74 seconds. Confidence of fill

It shows the probability of plastic filling a region within the cavity under conventional injection molding conditions [9].

This result is derived from the pressure and temperature results. The confidence of fill will display in three colors: green, yellow, and red. Showing in green will definitely fill, in yellow may be difficult to fill or may have quality problem, in red will be difficult to fill or will have quality problem and in translucent will not fill (short shot).

The result shown in Fig. 8 below depicts that the shoe sole designed could be filled 100% as the region shows green color. Thus, the shoe sole mold will not definitely face short shot in the process of injection.

Quality Prediction Result

The Quality prediction result is used to estimate the quality of the mechanical properties and

appearance of the part. It is derived from pressure, temperature and other results. The colors displayed in the quality prediction result are red (will definitely have quality problems), yellow (may have quality problems) and green (will have high quality).

But for the result with some yellow, the confidence of fill result has to be checked. As the quality prediction analysis result shows in Fig. 9, yellow region is found. This may have quality problems if the confidence of fill is not 100%. However, in this analysis confidence of fill result shows 100% with green color as seen in Fig. 8. This assures that the shoe sole model can be easily filled and part quality should be acceptable [3].

Injection Pressure

The Injection pressure result shows the maximum injection pressure value obtained before the velocity/ pressure switch-over occurs during the filling phase [9]. At the beginning of filling, the pressure is zero, or 1 atm in the absolute pressure scale, throughout the mold. The pressure at a specific location starts to increase only after the melt front reaches that location. The pressure continues to increase as the melt front moves past, due to the increasing flow length between this specific location and the melt front.

The pressure difference from one location to another is the force that pushes the polymer melt to flow during filling. The pressure gradient is the pressure difference divided by the distance between two locations.

Like water flowing from higher elevations to lower elevations, polymer always moves in the direction of the negative pressure gradient, from higher pressure to lower pressure; therefore, the maximum pressure occurs at the polymer injection locations and, the minimum pressure occurs at the melt front during the filling stage, as shown in Fig.10.


89

NCSTI-2012

As the injection pressure analysis result shows in

Fig.10 above, the maximum injection pressure for the shoe sole is 13.1MPa and it occurs at the gate location.

Time to reach ejection temperature

The Time to reach ejection temperature result shows the amount of time required to reach the ejection temperature, which is measured from the start of fill. If the part has not frozen by the end of the cycle time provided, a projected time to freeze is displayed in the result [9].

For Molds designed without micro switches it needs more cooling time than the analysis result shows. Thus by adding the cooling time, the part keeps the desired shape when ejected from the mold.

This result takes into account the dynamics of the packing phase, and where new hot material enters the cavity. This new hot material affects the cooling time.

From the time to reach ejection temperature result in Fig.11, four points are taken randomly to observe the ejection time in the different parts of the shoe sole. As it can be seen from point 1 the time to reach ejection is 10.11 sec which implies the region is relatively thin. And progressing in the same way thicker regions like point 4 takes the maximum time to ejection of about 64 sec.

Air traps

1 2

An air trap occurs where the melt traps and

compresses a bubble of air or gas between two or more converging flow fronts, or between the flow front and the cavity wall. Typically, the result is a small hole or a blemish on the surface of the part. In extreme cases, the compression increases the temperature to a level that causes the plastic to degrade or burn.

Air traps are often due to converging flow fronts caused by racetrack or hesitation effects, or by non-uniform or non-linear fill patterns. Air traps are acceptable if they occur on a surface that does not have to be visually perfect.

The air traps analysis result is shown in Fig.12. As indicated in points 1 and 2, some air traps are observed on the surface of the shoe sole model. However, these kinds of air traps are acceptable if they occur on a surface that does not have to be visually perfect. D. Mold Development

After doing all Moldflow simulation for the product, mold design will be proceeded based on the simulation results. The shoe sole mold designed in this paper is a two plate mold. The standard two plate mold parts are ejector system (ejector pin), back plate, cavity, sprue, stationary (core) and moving (cavity). However, for most of the molded shoe soles, the two plate molds don’t have some of the parts listed here. The very important parts of two plate shoe sole mold for injection molding are stationary plate, moving plate, runner and gating system. The design of these parts is discussed in the subsequent sections.

The experience in shoe sole industries with Main Group SP 345-3 machine shows that parts are removed from the stationary (core) plate manually. This is because of the flexibility nature of shoe sole plastics (like TPR), which would be difficult to put ejector pins to eject parts.


90

NCSTI-2012In designing core and cavity the important inputs

related to it are to determine the shot capacity and clamping force.

Shot Capacity

The shot capacity is the full amount as a weight or volume of material injected during molding from the screw. This is usually given as a shot capacity for polystyrene, and will vary with material. The shot size is the amount of material required to fully fill a molding tool [10].

Shot Capacity of Injection Unit =

(1)

Shot capacity = 750 cm3*300 bar /1000 = 225 Nm

Clamping Force

When the machine is started up with a new mould, the internal mould pressure required is not yet known. Naturally, from experience, it may be possible to extrapolate a figure, which depends on the type of plastic and the component format. However, practice has continually shown that it is initially quite sufficient to use a guide value of 2.5-5 KN/cm2 to calculate a projected molded component surface[10].

The mould closing force is the force required to physically shut the tool. The clamping force is the force required to hold the tool shut when the material is injected and is much higher than the closing force.

Clamp. F = Proj. Area * Specific Clamp. F (2)

The projected area can be defined as the area of the shadow cast by the molded part cavity when it is held under a light source, with the shadow falling on a plane surface parallel to the parting line [11]. The projected area for the designed shoe sole is found from the CAD design in CATIA®. Other values are Volume=1.558*10-4 m3 and density (ρ) = 1.04g/cc for TPR then the weight= ρ (density)* v (volume) = 162 grams, (which can be taken 190g with some allowance). Material: TPR Specific clamping force: 2.5 to 5 KN/cm2 (Average = 3.5 KN/cm2), Molded part's projected area: 425 cm2

Clamping force = 3.5 x 425 = 1487.50 (KN)

Core and cavity Creation

The shoe sole model designed in CATIA is imported to Autodesk® Inventor® Professional to undertake the mold design process. The mold design

process starts with defining workpiece setting which in our case is taken as 400mm*300mm*100 mm from the injection molding machine standards shown Table 2.

Then the runoff surface, parting line and extruding direction are defined as per the proposed settings.

After that the core and cavity are generated

automatically following the given parameters of runoff surface, parting line and extrusion direction as shown in Fig. 13. Runner Design

The weight of the shoe sole is found to be about 190g and the runner length about 140 mm, the runner diameter for this two plate mold is found to be about 10.547 mm by using the following equation [12].

(3)

The runner diameter could also be found from table given for different materials in Mold design basics [12]. The average runner diameter could be taken as 10 mm.

Thus, the most suitable runner type for shoe sole mold is found to be standard runner type with two impression runner layout. Gate Design

The gate is the link between the part and the runner system and it is suggested that it has to be as short as possible. There are many types of gates, among this the most suitable type for two plate molds is tunnel gate.

The base wall thickness of the shoe sole as indicated above is 3.5mm, thus the diameter of the gate has to be 0.3w to 0.7w [13]. Taking an optimum percentage of 0.55w, the diameter of the gate to which it is attached to the sole cavity is calculated to be about 2mm. i.e. 0.55*3.5mm ≈ 2mm.

This result is compatible with the default gate diameter suggested on Autodesk Moldflow and is acceptable result as shown in Fig.14.


91

NCSTI-2012

VI. CONCLUSION The design of the shoe sole mold die has been

undertaken by following critical shoe sole design procedures, from modeling shoe sole using CAD software and analysis of the model using Autodesk® Moldflow software to generation of mold parts using Autodesk® Inventor software.

The French/Continental (pp) sizing system with a value of 41 pp having a moccasin shoe sole style for male shoe sole is selected. Aluminum is found to be the most suitable mold material for Thermoplastic Rubber (TPR) plastic injection process. From this the shoe sole design is developed, the best gate location is found and quality results are predicted.

From part analysis results the fill time is about 2.74s with 100% confidence of fill, 13.1 MPa of injection pressure and 64.34s time to reach ejection temperature. And the runner length and gate diameter for the mold are found to be 140mm and 2mm respectively.

The part simulation analysis would enable to achieve the desired prototype by predicting the possible problems and suggesting solutions. It was possible to generate the molding halves and design melt delivery systems based on the inputs from tooling software and eliminates basic design error.

Finally, a shoe sole mold that could produce two soles at a time is designed successfully. The part simulation results could serve as guidance to manufacture the shoe sole mold with correct operating parameters to the shoe sole injection molding machine.

REFERENCES

[1]. Wong, C.T., Shamsuddin Sulaiman, Napsiah Ismail & A.M.S. Hamouda, Design and

Simulation of Plastic Injection Molding Process, Universiti Putra Malaysia Press, Vol. 12 No. 2 (2004) pp 85-99.

[2]. W. Wan AbduRahman, L.T. Sin, A. Rahmat, Injection molding simulation analysis of natural ber composite window frame, journal of materials processing technology 197 (2008) pp. 22–30

[3]. L. D. Bank, Dave Klafhen, R. Smierciak “Why Plastic Flows Better in Aluminum Injection Molds”, Aluminum Injection Mold Co., Plastic Processing Consultant, Alcoa Forged and Cast Products.

[4]. L.Kong, JYH. Fuh, A windows native 3D plastic injection mold design system, journal of materials processing technology 139 (2003) pp. 81–89

[5]. TLD Asia Psific (2000), Footwear Design and Manufacture, 1st Edn

[6]. Ross T., W. A Rossi, (1993), “Professional Shoe Fitting”, National Shoe Retailers Association, USA

[7]. Georg Menges, Walter Michaeli, Paul Mohren, (2000), “How to Make Injection Molds”, 3rd Edn, Hanser Publishers, USA.

[8]. Wifama, http://used-footwear-machinery.com/product_info.php/products_id/208, Oct 2011.

[9]. Moldflow Adviser Advanced 2012 Help files, Autodesk, Inc. (2011).

[10]. V.Goodship, (2004), Practical Guide to Injection Molding, Rapra Technology Limited, UK.

[11]. Rosato D. V., Donald V. Rosato, Marlene G. Rosato (2000). “Injection Molding Handbook”, 3rd Edition, Kluwer Academic Publishers.

[12]. MTN KALIP SAN. LTD. STl., “Mold Design Basics”, http://www.mtn.com.tr/engindex.php, Sep 2011.

[13]. John P. Beaumont, (2004), “Runner and Gating Design Handbook: Tools for Successful Injection Molding”, 1st Edn, Hanser Publishers, USA.


92

NCSTI-2012

Dynamic Formulation of 3 PRS PKM Based on Screw Theory And Newton-Euler’s Approach

Hassen Nigatu 1

Department of mechanical engineering, Adama Science and Technology University, Adama, 1888, Ethiopia, E-mail: pr.@adama_university.net

Abstract— many applications in the field of production automation, such as machining, assembly

and material loading require machines that are capable of high speed and acceleration. Since 1965 Stewarts’ work the parallel robot proved that to be strong supplement to the serial manipulators. Parallel robots are able to work on many tasks with much better performance. The study of robot dynamics is necessary for its mechanical design and synthesis, providing the force that must be resisted by joints, links and actuators. The model for the inverse dynamics is developed in this work is based on Newton-Euler’s approach and is a capable of calculating the force a long the driving direction during its operation. Inverse position analysis is provided by vector loop method, followed the inverse velocity and acceleration are performed with developing Jacobian and new hessian matrix for them by screw theory and Lie algebra. Based on the kinematics, the equation of motion of a manipulator formulated for each moving link and platform. This formulation performed by synchronizes the velocity and acceleration equation, which formulated in a very simplified form using screw theory, with Newton –Euler’s equation of motion of the manipulator. Therefore we found the equation of all the reaction force, constraint moment and driving force equation in a very integrated and unified form. Furthermore the solid model and its simulation is provided by using UG-NX-5 in order to visualize the system before manufacturing the real system and mathematical formulation solved by using software MATLAB.

Keywords: parallel kinematic machine, Jacobian, Hessian, UG-NX-5, 3PRS, Dynamics

I. INTRODUCTION There is a plentiful research on lower mobility

parallel kinematic machines (PKMs), most of which examines the machining applications of the parallel robot.

Recently, there has been an increasing trend in the research of parallel manipulators, as industries want to make the current machines smaller, more energy efficient and precise. For instance, Y.G Li, T.Huang and H.T .Liu[1] “a general approach for formulating dynamics of lower mobility parallel manipulators. Moreover, Joshi, S., Liu, H., Chetwynd, D.G.,Li,Z., [2]“formulated the generalized jacobian analysis of lower mobility manipulators, Dressler, I., Robertson. And Johansson, R.[3] “has designed the accuracy of kinematic and dynamic models of a Gantry-Tau parallel kinematic robot, Li Y G, Song Y M, Feng Z Y, et al.[8] has derived the Inverse dynamics of 3-RPS parallel mechanism by Newton-Euler formulation in Chinese and Liu, H., Huang, T.,Chetwynd, D.G.[11], An Approach for Acceleration Analysis of Lower Mobility Parallel Manipulators. Since the industrial revolution, there has been an ever increasing to improve product quality and reduce manufacturing cost parallel kinematic machines are applied on various area such as, For high speed and high precision machining

center, for machining purpose, air plane simulators and For pick and place, assembly and carrying load purpose in higher industries like aircraft, electronics and automotive factories and robotic assistance surgery.

I I. CAD MODEL KINEMATIC DESCRIPTION A Position analysis The architecture of 3PRS PKMs is showing on

fig.2 which is composed of moving platform, a fixed base the three supporting limbs with identical configurations. Each limb connects the fixed base to the moving platform by prismatic, revolute and spherical joint respectively and the prismatic joint is the active joint which is actuated by the linear actuator servo motor.

The considered machine is q 3-DOF PKM, which can be showed mathematically by using mobility criterion.

j

iifjnM

116 , where M is the DOF, n

is number of links in the system, j is number of joint and fi is the number of DOF of the ith joints. For this manipulator n=8,j=9,fi=3 for spherical joint and fi=1 for prismatic and revolute joints. Therefore the above equation will give as the system has 3 DOF.

31131986M 3DOF (1)


93

NCSTI-2012The vectors and reference frames are described

also in the figure for the sake of analysis, as shown in the figure a fixed Cartesian reference coordinate frame O(x, y, z) is attached at the center point O of the fixed triangle base platform ΔA1A2A3. And the moving coordinate system p(u, v, w)is attached on the moving platform a point P Which is a center of ΔB1B2B3. For simplicity without losing generality, let x-axis be aligned toward OA3 and the u-axis pointing along the direction of 3PB .From the above expression a1,a2,a3 are the vector of fixed base platform from the center point O to Ai and b10,b20, b30 are the vectors of the moving platform from the center point P to Bi.

Fig.1. 3D CAD model of 3PRS PKM

Fig.2. kinematics architecture of 3PRS Parallel manipulator

The vector loop equation can be written in the following form based on geometric arrangement of

the machine components.

0,iBA

i bRpP Where, (2) TzyxP "'' zxzRotR This means that the first rotation

about z axis (ϕ), the second rotation will followed by about x axis (ψ) and the final rotation about y axis (θ). This yields the following rotation matrix in equations (3).

The machine being analyzed here is a spatial

mechanism. A complete description of position and orientation of the upper platform with respect to the fixed reference frame needs six variables. These six variables have been chosen as the element of Tzyx .

"'' ZXZRR

CCSSSSCCCCSSSCCCS

SSCSSSCSCSCCRB

A (3)

Then the constrained equations formulated as follows along x- and y- axis.

212

ccbx (4)

212

ccby (5)

y

z

A

1B

2B

3B

iS ,1

iS ,2

irS ,1

1A 3

irS ,2

ii sl ,3

ii sd ,1

p

O

'p

u

v

w

ia

0ib

Once if we found the entire constraint equations we need to find the parasitic rotation motion ϕ by back substation of all element of the rotation matrix in to R12= R21.

x

After some computations and replacement this will give as

ϕ=ψ (6) Now we can formulate the prismatic motion

parameter “d” by vector loop method based on fig.2. Then we got the following final form formula

after huge mathematical computations.

22,1

2,1 iii

Tii

Tii bpsplspd (7)

B velocity analysis In this section the velocity analysis of 3PRS have

done by applying the principle of screw theory. There are five joint screws associated with each limb. The first joint is the only actuated joint and the remaining joints are passive. The instantaneous twist of the moving platform can be expressed as a linear combination of five screws.

iiiiiiiiii sssssdps ,5,5,4,4,3,3,2,2,1,1 ˆˆˆˆˆ (8) Where


94

NCSTI-2012

0

ˆ 1,1

ss i , ,

2

2,2ˆ s

slbs ii

i

3

3,3ˆ s

sbs i

i

4

4,4ˆ s

sbs i

i

jis ,̂

s2

, (9)

Where is a unit vector along the joint Jth joint of the ith limb. These five screws form five system for which a one system of reciprocal screw exists. This reciprocal screw lies on the intersection of the two planes. The first plane is perpendicular for the prismatic joint axis and the second plane is containing both the revolute and spherical joint. This reciprocal screw denoted as is zero pitch screw passing through the center of spherical joint and parallel to .

5

5,5ˆ s

sbs i

i

i,

irs ,1̂

i

iiir s

sbs

,2

,2,1ˆ (10)

By taking the inner product of both sides of the instantaneous with the constraint wrench we found the constraint Jacobian . )( CJ

(11)

TT

TT

TT

C

sbssbssbs

J

3,223,2

2,222,2

1,211,2

This matrix represents the constraint imposed by the revolute joint. An additional basis screw which is reciprocal to the passive joint of the ith limb can be identified as zero pitch screw passing through the center of spherical joint. This reciprocal screw represents wrench of actuation and it is normal to the previous system. This can be expressed by

(12)

i

iiir s

sbs

.3

.3,2ˆ

Take the orthogonal product of this reciprocal

wrench for both sides of the twist screw. Then we found Jx.

(13) iiriir dsssps ,2,1,2 ˆˆˆˆ

(14)

TT

TT

TT

x

sbssbssbs

J

1,311,3

1,311,3

1,311,3

Equation 13 can be rewrite three times, once for each limb and it yields.

qJpsJ qx Where the inverse Jacobian is is:- qJ

(15)

3,13,3

2,12,3

1,11,3

000000

ssss

ssJ

T

T

T

q

The actuation Jacobian, is responsible to relating the Cartesian velocity with joint rate , is a

function of and . qJ

xJ aJ

c

a

JJ

J Where, q

xa J

JJ

16

000

)(

)(

)(

)(

)(

)(

3

2

1

0

3,233,2

2,222,2

1,211,2

3,13,3

3,33

3,13,3

3,13,3

2,12,3

2,32

1,12,3

2,11,3

1,11,3

1,31

1,11,3

1,3

d

d

d

v

sbs

sbs

sbs

sssb

ssss

sssb

ssss

sssb

sss

n

TT

TT

TT

T

T

T

T

T

T

T

T

T

T

T

T

Equation 16 is a generalized Jacobian that relates

the velocity of joint rate to the velocity of the moving platform.

C Acceleration analysis For the acceleration analysis of this 3PRS parallel

manipulator we used a new approach [11] named Hessian matrix to achieve an explicit and compact form equation.

],,,,,,[, ,5,4,3,2,1 iiiii stastastastastaiJa (17)

i

iiiiic n

nlsbstcJ

,1

,1,3,1, for i=1,2,3 (18)

Where is constraint Jacobian for a limb. icJ ,

Here

ii

iii ss

sss

.5,3

.5,3,4 and for i=1,2,3 ii Rss ,2,5

iciai JJJ ,, , Jacobian for each limb. (19) The Hessian matrix can be found from the derivative of screws in the Lie bracket form. This yields...

This equation puts the screw

derivative ii

Tiii qHqqJA **(

iq) and its coefficient side by side.

i.e. is coefficient and is screw derivative in Lie bracket form.

iH

(20)

ic

iaciai H

HHH

,

,,

0When we put all these screw derivatives independently yields the following three equations.


95

NCSTI-2012

,1, ,2, ,1, ,3, ,1, , ,

,2, ,3, ,2, , ,

,

, 1, , ,

ˆ ˆ ˆ ˆ ˆ ˆ

ˆ ˆ ˆ ˆ

ˆ ˆ

i

i

i i

ta i ta i ta i ta i ta i ta n i

ta i ta i ta i ta n i

a i

ta n i ta n i

H

0 $ $ $ $ $ $

0 0 $ $ $ $

0 0 0 $ $

0 0 0 0

,1, ,2, ,1, ,3, ,1, ,6 ,

,2, ,3, ,2, ,6 ,

,

,5 , ,6 ,

ˆ ˆ ˆ ˆ ˆ ˆ

ˆ ˆ ˆ ˆ

ˆ ˆ

i

i

i i

tc i tc i tc i tc i tc i tc n i

tc i tc i tc i tc n i

c i

tc n i tc n i

H

0 $ $ $ $ $ $

0 0 $ $ $ $

0 0 0 $ $

0 0 0 0

,1, ,1, ,1, ,6 ,

,

, , ,1, , , ,6 ,

ˆ ˆ ˆ ˆ

ˆ ˆ ˆ ˆ

i

i i i

ta i tc i ta i tc n i

ac i

ta n i tc i ta n i tc n i

H

$ $ $ $

$ $ $ $

(21)

(22)

(23)

The matrix in eq. (20) is Hessian with 666 matrix for each limb contain both the actuation and constraint joints and eq.(21,22,23) are the element of general Hessian matrix.

I I I. DYNAMICS OF 3DOF PRS PKM A D

machine mathematical

ew principle by

erse

parallel

4 shows that the vector and scalar

CAD model of a limb.

igure 4. Schematic representation of a limb

tion matrix of the ith limb can be writing as follows.

(24)

using Euler’s angle rotation computation method.

ynamic formulation The vital step in the robotic formulation and analysis is to understand about the dynamical behavior, which allows knowing the force and torquing demand in each mechanical element during the operation. Another purpose of formulate and analyze the dynamics of parallel manipulator is to provide fast and easy equation to solve the control algorithm. In order to solve these problems in detail that related with dynamics of machine, there are different approaches that have been proposed at different time, Such as Lagrange, N-E and virtual work principles. This work comes with a ncombining screw mathematics and N-E approach in order to utilize the advantage in both methods. The work in this paper focused on the invdynamics, i.e. knowing the data related to the moving platform (position, velocity and acceleration) and looking for the information on the limbs and joints (motor torque and reaction force) by using the combination method of these two widely used and complicated mathematics. In order to find the dynamics of the manipulator we need to decompose the parts and formulate the velocity and acceleration equation for each joint of the limb and each limb of the manipulator. Figure 3 andrepresentation in one typical limb geometric

configuration with their global coordinate located at the fixed platform. Analyzing one limb and the platform is good enough to understand the dynamical behavior of all other limbs since all limbs have identical configuration and also since the manipulator has symmetric configuration.

Figure 3. Design parameter representation on the

F Assume each leg is connected to the slider by the

revolute joint such that it cannot rotate about z’ axis, the orientation of limb i with respect to the fixed base will by roation of ϕi about z axis and followed by roation of θi about y axis. Hence the rota

ii cs 0The above rotation matrix simply obtained by

iiiii

iiiii

iR ssccsscscc

A


96

NCSTI-2012

Here we need to find the screw axis for the fixed length limb from the geometric configuration of a limb.

iyix

iyiy

ixix

i

ss

bayl

baxl

s

,2

3,2

3

,3

1

1

1 (25)

From schematic diagram fig.(2) we obtain the position analysis equation in vector sum form as follows

(25a) 0,,3,1 iiiiii bpslsda Hence, by differentiating the right hand side of

eq.(25a) the velocity at contact point Bi, the point between fixed length link and platform (spherical joint), will be formulated as follows.

0,ii bvpvb (26) Transforming eq.(26) to the ith limb frame yields

iAi

ii vbRvb , where

Ti

AA

i RR The velocity at the spherical joint can be

formulated in terms of angular velocity of the ith limb by taking the derivative of left hand side of eq. (25a) with respect to time.

iii

ii

ii

iii

ii

iii vbslslsdsd 33,1 (27) Cross multiplying both sides of eq.(27) by si yields.

i =i

0

1,

,

xii

yii

i

vb

vb

d+

0

1,

,

xii

yii

i

vb

vb

l

(28)

(29) rhssdar iiiii ,7,1,1

(30) riserr iii ,6,1,2

By differentiating eq.(29) we found the velocity mass center of the fixed length arm as follows.

= iiv ,1

ixii

i

vbld 0

01 (31)

Again by differentiating eq.(30) we also found the velocity of mass center of the slider.

xiii

ii

ii

iii

i

vblld

lddlev

,

,2 00

(32)

The acceleration of ball point Bi expressed in the fixed reference frame is obtained by differentiating eq.(26) with respect to time. oippoippib bbvv ,,. (33) Express eq.(33) in terms of the ith limb frame to transform from manipulator coordinate frame to local limb coordinate frame as follows.

(34) ibAibi vRv

i

.. In order to find the angular acceleration of the ith

limb differentiate the right hand side of eq.(27) and cross multiplying by isi .this yields.

0

2

2

1

i

ixi

iyi

ixi

i

iyi

izi

iyi

ii

i

dvbvb

vb

dvbvb

vb

d

+

0

2

2

1

i

ixi

iyi

ixi

i

iyi

izi

iyi

i lvbvb

vb

lvbvb

vb

l

(35)

Once we formulate the angular acceleration of the

ith limb, the acceleration of the center of mass of the fixed length link that connect with the platform by spherical joint and the limb body, that is the link containing the prismatic motion, can be drive by derivate eq.(31and 32).

xiii

ii

i

bvldv

,

,1 00

1

(36)

xiii

ii

ii

iii

i

bvlld

lddlev

,

,2 00

(37)

After we formulate all velocity and acceleration of center of mass of the link, we need to find the dynamics equations of the manipulator. To simplify the analysis we decompose the manipulator in to three identical limbs and platform. We combine the fixed length link and limb body in to subsystem and formulate the dynamic equation for the subsystem.

Euler’s equation of motion can be written about axis of revolute joint Ci .

ici

ici h

d

tdn (38)

iCi

iCi

ii

ii

ii

ii

ici hhvsemvsrimh ,2,1.2,62.1,31 (39)

Where = , = are the angular momentums of the fixed length link and the vertical link containing the sliding joint about their respective center of mass, and where and are the inertia matrix of the fixed link length and limb body of sliding joint about their respective center of mass and expressed in the ith limb frame. Differentiate eq.(39) with respect to time we found :

iCh ,1 i

ii

i I ,1 iCh ,2 i

ii

i I ,2

ii I ,1 i

i I ,2

ici h

dtd = ii vsrm .1,31 ii + ii vsem .2,12 ii +

ii

ii I ,2i

ii

ii

ii

i I ,1 ii

ii

ii

ii II ,2,1 (40)

Since the orientation of link changes as the link moves, causing and vary. The time rate of change of angular momentum includes both the

ii I ,1

eration ter

ii I ,2

m ( i Iangular accel & ) and i

ii ,1 i

iI ,2


97

NCSTI-2012

(gyroscopic torque term ii

ii ,1

and

ii I

ii

ii

ii I ,2 ).

gR AA

ii,6

(40)

and

ci gc (41)

in to eq(38).

semgRs iAA

ii 2,3mfsl ib

ii

ii 0,,3

Where xibi

ibi ff ,0,0, Ag 00

cii

xii

i

yii

i

gssemrmfbl

mfbl

2

1,0

,0

Then substit

n ii

ci1

Tzib

iyib

i ff ,0,,0,

Tcg

ici n =

s

cii

cii

emgss

gsr

2

1

ute eq(40) and eq(42)After some substitution and further simplification by using the fact that the product of inertia are all equal to zero we obtain xi

i fb , andyi

i fb ,.

ixi

i fb 1

l,0 xii

i vrm ,11 - cic s egmsrgm 21

il1 yi

iiyy , i

yii

iyyi

xii IIve ,2,,1,22 m

(43) xii

ixxi

xii I ,,2,

ulating the dynamic

is the force applied on the

taken about the center

P (46)

0i

ng

c

(47)

zi

zi

zi

z

fb

fb

,0

,0

,0

,0 (48)

ixxi

yii

yii Ivemvrm ,1,22,11

i

i

lfb 1

yi ,0

(44) The next step should rm

eq

orm

m latform by the

moment of

reaction force exerted on the movi

AR BR

0i

B fb

be fouation of motion of the moving platform. These

equations of motion are expressed in the fixed frame. Since the reaction forces

ii fb obtained in the

ith limb frame, they should transf in to the fixed frame before substituting it in the motion equation.

3

Ap

Api

A vmgmfb (45) 1

0i

p

Where 00 ii

iA

iA fbRfb

oving p ith limb expressed in the fixed frame A.

The resulting Bnp

mass of the moving platform and expressed in the moving frame B is

pB n =

3B

ii

Bi

B Nfbb 1

0

Where B = denotes the 0ifb 0

ii

Bi

AA

B fbRfbR

platform by the ith limb at the spherical joint and then expressed in the moving frame B.

B = TA , ssscccsscscc

scsscscccsscsssc

=

iyi

ixi

iui

iyi

ixi

ivi

iyi

ixi

iui

ii

yii

xii

vi

fbafbafbab

afbafbab

afbafbab

fbafbafbab

23,022,021,0

13,012,011,0

33,032,031,0

33,032,031,0

pvpwpupuuppvpvzwp

pwpwpvpuuppwpyvp

pwpwpvpvvppwpuup

z

v

uB

IIIIIIIII

nnn

,,

,,

,,

(49)

Assume that are principal axis of the moving platform. Substitute eq.(48) and eq.(49) in to eq.(46) and use the fact that all the product of inertia are equals to zero. Then we found.

vu, and w

3

1i

zii

yii

xii

vi fbafbafbab ,033,032,031,0

pwpwpvpvvppwpuup II

+

=PuNB I ,,

3

1i

zii

yii

xii

ui fbafbafbab ,033,032,031,0

pwpwpvpuuppwpy II

+

=

PvB N

vpI ,,

3

1i

pzi

iyi

ixi

ivi

zii

yii

xii

ui

Nfbafbafbab

fbafbafbab

,013,012,011,0

,023,022,021,0

= pvpwpupuuppvpvwp III ,, The final and the main target of this formulation is

to get the actuating (driving) force of the slider along z axis. Therefore once the reaction force at the spherical joint are found, the driving force can be obtained by summing up all the force acting on the ith fixed length link along the zi axis passing through the center of spherical joint. Also the constraint moment on the revolute joint along xi axis as follows.

(50) izi

iczii vmcgmfbf .1110

(51) xioi

iyii fblM ,,

B Numerical simulations 020,800.0,6699.0;1794.0;2807.0 zlba

OO 3600

382.0;075.0;083.0

01.37;17.31;135.69 21

rehmlmlmp

ilI ,1,

096899.0810274.0

859692.0

ilI ,2,

102787.0185955.0

20125.0

02793.100073393.000002978.1

pI

TA g 87.900 FA , T100


98

NCSTI-2012

stststT 4783.4,2857.4,1926.0,3852.0 321

Let’s put the path and motion rules of the platform as follows. The trajectory is selected to represent a typical motion that might be used during a practical application of the manipulator.

max 1

1 2

max 2 3

2sin 0

02sin

πψ t t tT

ψ t t t tπψ t t t t

T

(51)

maxmax 1

max 1 2

2 maxmax 2 3

2cos 02 2

2cos

2 2

ψT πψ tπ T

ψ t ψ t t t

π t t ψTψ t t tπ T

t t

(52)

2max

max 1

maxmax 1 2

2

max 2

2 3max max

2

2sin 02 2

4

2sin2

2 4

ψT πψ t t tπ T

ψψ t T t tψ t

T πψ t tπ T t t t

ψ ψt t T

t

t

1

k

(53)

Take the time derivative of eq.(4) and eq.(5) results T

ccv zkwherekJpv ],,[ And (54)

C2 1 C 0.5 S2 S 0S2 1 C 0.5 C2 S 0

0 0v

a ψ θ a ψ θa ψ θ a ψ θ

J

The time derivative of eq.(54) results

(55) HvkckcckJvv T

Where is matrix with (i=1, 2, 3)

is derivative of three independent

variables. The derivative of Jv yields three dimensional matrix with three layers as given below.

Hv

z

333

c

iHv ,

ck

,1

2 S2 1 C C2 S 0C2 S 0.5 S2 C 0

0 0Hv


0

0

(56)

,2

2 C2 1 C S2 S 0S2 S 0.5 C2 C 0

0 0v


H (57)

(58)

000000000

,3Hv

The angular velocity vector of the platform zyx

can be expressed by standard matrix operation for

T],,[ x operator.

kcJ (59)

Where, S S C 0

C S S 01 C 0 0

Jω

ψ θ ψψ θ ψθ

Taking the time derivative of eq.(59) it results (60) cHkcckJ T

Where 3 3 3Hω �

,

is also a three dimensional matrix with Hω i i ( 1, 2,3 ) being its ith layer;

,1

C S S 0S C 0 0

0 0Hω

ψ θ ψψ θ

0

,3Hω θ

,2

S S C 0C C 0 0

0 0Hω

ψ θ ψψ θ

0 0 0S 0 00 0 0

0

As we know the inverse acceleration analysis formulation from different level of understanding the independent joint variables are given and we need to find the active joint variable acceleration and the moving platform acceleration.

Therefore is given. Acceleration of the prismatic joint ( ) can be evaluated using the equation formulated below with the help of manipulator geometry in put parameter.

ckT

aaaa qqqq ],,[ 3,2,1,

From the equation of motion of the platform (for i=(1,2,3) and where A is

“k” layer of ii

Tiii qHqqJA

96 matrix ) and the principle of reciprocal screws by multiplying both side of the above equation with after some computation we found JA . Where , ,

is the acceleration of three active sliding joints

on the ith limb followed by

irir ss ,2,1 &

qHqq T

aa qq ,2, Taqq ]0[

aq

Taa qq [ 1,

6

]3

66H matrix known as the Hessian matrix of 3 PRS parallel kinematic machine.

tTT

t sHJJsJAq So here in this equation we have derived the mainly important part of solving acceleration analysis of 3 PRS.

izi

iczii vmcgmfbf .1110

xioi

iyii fblM ,,

When the platform of the manipulator moves according to the preceding rules, the velocity and acceleration of the actuated joints, the linear


99

NCSTI-2012

velocity and acceleration of the reference point P’, and the angular velocity and acceleration of the platform versus time can be evaluated using the proposed approach with the software MATLAB for solving all the formulation provided.

0 0.5 1 1.5 2 2.5 3 3.5 4 4.5-0.02

-0.015

-0.01

-0.005

0

0.005

0.01

0.015

0.02

time (sec)

linea

r vel

ocity

(m/s

)

linear velocity of platform

vxvyvz

(a)

0 0.5 1 1.5 2 2.5 3 3.5 4 4.5-0.2

-0.15

-0.1

-0.05

0

0.05

0.1

0.15

time (sec)

linea

r acc

elra

tion

(m/s

)

linearaccelration of the platform

dvxdvydvz

(b)

0 0.5 1 1.5 2 2.5 3 3.5 4 4.5-0.2

-0.15

-0.1

-0.05

0

0.05

0.1

0.15

time (sec)

linea

r vel

ocity

of t

he s

lider

(m/s

)

dq1dq2dq3

(c)

0 0.5 1 1.5 2 2.5 3 3.5 4 4.5-0.8

-0.6

-0.4

-0.2

0

0.2

0.4

0.6

time (sec)

angu

lar v

eloc

ity (m

/s)

angular velocity of platform

wxwywz

(d)

0 0.5 1 1.5 2 2.5 3 3.5 4 4.5-4

-3

-2

-1

0

1

2

3

4

5

time (sec)

angu

lar a

ccel

ratio

n (m

/s)

angular accelration of the platform

dwxdwydwz

(e)

0 0.5 1 1.5 2 2.5 3 3.5 4 4.5-0.8

-0.6

-0.4

-0.2

0

0.2

0.4

0.6

0.8

time (sec)

linea

r acc

eler

atio

n of

pris

mat

ic jo

int (

m/s

2 )

linear acceleration of sliding joint

ddq1ddq2ddq3

(f)


100

NCSTI-2012

0 0.5 1 1.5 2 2.5 3 3.5 4 4.5800

850

900

950

1000

1050

time (sec)

driv

ing

forc

e in

N

linear linear driving force of actuator

fz1fz2fz3

(g)

0 0.5 1 1.5 2 2.5 3 3.5 4 4.5-500

-400

-300

-200

-100

0

100

200

300

400

500

time (sec)

Con

stra

int f

orce

in N

m

constraint force of actuator

M1M2M3

(h)

Figure 5 Linear velocity (a) and acceleration (b) of the moving platform, linear velocity of the actuation joint(c) and angular velocity (d) of the moving platform, angular acceleration(e) of the moving platform and linear acceleration(f) of the sliding joint, linear driving force (g) of the sliding joint and constraint moments(h).

I V. CONCLUSION

Throughout this work, the CAD model for 3PRS machine is buildup, the inverse position has been analyzed, inverse velocity and acceleration is derived in detail by using screw theory with developing new Hessian matrix. This machine designed to have translational motion along z-axis and rotational motion about x and y axis with pure 3DOF motion. Followed the inverse dynamics studied extensively in detail and simple form by applying N-E approach and screw theory for the motion equations. In the analysis of velocity and acceleration the simulation result have been completed also later the unit force applied on the actuator along z-axis the simulation performed in order to get the actuator driving force.

The analytical result shown in fig.5 (a, b, c, d, e, f g & h) has solved by using the software MATLAB by writing the algorithm for N-E approach. The result shown if fig.5 has verified by using the software UG-NX-5 simulation of CAD model (except the constaint moment )by comparing with the result that we found from N-E approach. Therefore from the comparision of the results we realized that there is no significant difference between the analytical and simulation result.

VI. REFERENCE

[1] Zhang, F.; Zhang, D.; Yang, J., and Li, B., 2005, "Kinematics and Singularity Analysis of a 3-DOF Parallel Kinematic Machine.", IEEE. 29 July-1 Aug. 2005.

[2] Joshi, S., Liu, H., Chetwynd, D.G.,Li,Z.,

“generalized jacobian analysis of lower mobility analysis of lower mobility manipulators,” IEEE transaction on robotics and automation, submitted .

[3] Dressler, I., Robertson. and Johansson, R.,2007, “Accuracy of kinematic and dynamic models of a Gantry-Tau parallel kinematic robot,” Proceedings of IEEE International Conference on Robotics and Automation, pp.883- 888

[4] Tsai, L.-W., 2000, “Solving the inverse dynamics of a Stewart-Gough manipulator by the principle of virtual work,” ASME Journal of Mechanical Design, 122(3): pp. 3-9.[5] Li ， Y.M. ， Xu ， Q.S., 2005, “Kinematic Analysis and Dynamic Control of a 3-PUU Parallel Manipulator for Cardiopulmonary Resuscitation” International Conference on Advanced Robotics, pp.344-351

[6] Wisama Khalil and Sylvain Guegan, 2004, “Inverse and Direct Dynamic Modeling of Gough–Stewart Robots,” IEEE Transactions on Robotics, 20(4):pp.754-762.

[7] SUN Tao1, SONG YiMin1†, LI YongGang2 & LIU LinShan1 “Dimensional synthesis of a 3-DOF parallel manipulator based on dimensionally homogeneous Jacobian matrix”

[8] Li Y G, Song Y M, Feng Z Y, et al. “Inverse

dynamics of 3-PRS parallel mechanism by Newton-Euler formulation (in Chinese)”. Acta Aeron Astron Sin, 2007, 28: 1210–1215

[9] Carretero J A, Nahon M, Gosselin C M, et al. Kinematic analysis of a three-DOF parallel mechanism for telescope application. In: Proceedings of the 2007 ASME Design Automation Conference, Sacramento, California, 1997

[10] Lee, K. M., Shah, D. K., 1988, “Dynamic analysis of a three-degrees-of-freedom in- parallel actuated manipulator,” IEEE Transactions on Robotics and Automation, 1988, 4(3): 361-367


101

NCSTI-2012[11] Liu, H., Huang, T., Chetwynd, D. G., “An Approach

for Acceleration Analysis of Lower Mobility Parallel Manipulators,” Proceedings of the ASME 2010 International Design Engineering Technical Conference & Computers and Information in Engineering Conference, DETC2010-28020.

[12] Tsai. Lung-Wen, “Robot analysis: the mechanics of serial and parallel manipulators,” J.WILEY & SONS, INC., 1999

[13] Rico, J. M., Duffy, J., 1996, “An application of screw algebra to the acceleration analysis of serial chains,” Mechanism and Machine Theory, 31(4): pp. 445-457.

[14] Huang, Z., Zhao, Y. S., Zhao, T. S., 2006, “The Advanced Spatial Mechanism,” Beijing: The High Education Press.

[15] Meng-Shiun Tsai a,*, Ting-Nung Shiau a, Yi-Jeng Tsai a, Tsann-Huei Chang, “Direct kinematic analysis of a 3-PRS parallel mechanism,”

Mechanism and Machine Theory 38 (2003) 71–83 [16] G. Pond, J. Carretero, “Kinematic Analysis and

Workspace Determination of the Inclined PRS Parallel Manipulator” proceeding of the 2004 ROMASY, Montreal, Quebec, Canada.

[17] Zhu, S. J., Huang, Z., Guo, X. J., 2005, “Forward/reverse velocity and acceleration analyses for a class of lower-mobility parallel mechanisms,” Proceedings ASME Design Engineering Technical Conferences and Computers and Information in Engineering Conference, pp. 949-955.

[18] Sameer A. Joshi, Lung-Wen Tsai*, “Jacobian Analysis of Limited-DOF Parallel manipulators”

[19] Y.G. Li , H.T. Liu, X.M. Zhao , T. Huang ,*, Derek G. Chetwynd, “Design of a 3-DOF PKM module for large structural component machining” Mechanism and Machine Theory 45 (2010) 941–954

[20] B.Dasgupta, P.Choudhury, “A general strategy based on Newton-Euler approach for the dynamic formulation of parallel manipulator”

[21] Yangmin Li*, Qingsong Xu, “Kinematic analysis of a 3PRS parallel manipulator,”robotics and computer-integrated manufacturing 23(2007)395-408

[22] Clavel R.DELTA, “a fast robot with parallel geometry” proceeding of 18th international symposium on industrial robot, Lausanne; 1988.p.91- 100

[23] Lee KM, Arjunan S. “A three degree of freedom micromotion in parallel actuated manipulator” IEEE Trans robot automat 1991 634-641 [24] Jian-Li “Design of 3-DOF Parallel Manipulators for

Micro-Motion Applications” master’s thesis in University of Ontario Institute of Technology 2010

[25] Ting-Nung Shiau and Meng-Shiun Tsai “Research on

Kinematic and Dynamic Characteristics of a 3-PRS

Parallel Mechanism” Dissertation of Doctor of

Philosophy in National Chung Cheng University June

2007

[27] Wang, J.; Gosselin, C.M., 2004, “Kinematic Analysis and Design of Kinematically Redundantly Parallel

Mechanisms,” ASME J. Mech. Des., 126(1), pp.109- 118.

[28] Mohamed, M.G; Gosselin, C.M., 2005, “Design and Analysis of Kinematically Redundant Parallel Manipulators with Configurable Platforms” IEEE Trans. Rob.,21(3), pp. 277-287.104

[29] Zhang, D. and Wang, L., 2005, "Conceptual Development of an Enhanced Tripod

Mechanism for Machine Tool", Robotics and Computer-Integrated Manufacturing 21, no. 4-5 pp. 318-27.

[30] Bi, Z. M. and S. Y. T. Lang. "Kinematic and Dynamic Models of a Tripod System with a Passive Leg." IEEE/ASME Transactions on Mechatronics 11, no. 1 (02, 2006): 108-111.

[31] Chablat, D. and P. Wenger. "Architecture Optimization of a 3-DOF TranslationalParallel Mechanism for Machining Applications, the Orthoglide." IEEE Transactions on Robotics and Automation 19, no. 3 (06, 2003): 403-410.

Ergonomics Aspect of a Vehicle: The Case of Bajaj

102

NCSTI-2012


Mengist Hailemariam1, Awol MohammedDept. of Mechanical and Vehicle Engineering, Adama Science & Technology University, P.O.Box 1888, Adama, Ethiopia

1 Corresponding Author, E-mail: [email protected]

Abstract — The three wheeler-four stroke automobile, Bajaj, is one of the common means of public transport and carrying freight in Ethiopia and the rest of the world. The number of Bajaj is increasing in most cities and has created employment opportunity for many people while most drivers complain about physical fatigue as a result of inappropriate Ergonomics considerations. This paper attempts a descriptive study on the effect of driving Bajaj to get insight into issues and potential improvements for the systemfrom ergonomics point of view. Bajaj drivers from Adama City were participated in the study. Data are collected through observation, interview, and questionnaire. Critical parts of the Bajaj are evaluated and the drivers’ jobs are broken down into tasks and then analyzed. The results of the Ergonomic survey and task analysis imply that the design of Bajaj impact safety and comfort of the drivers and results series back and foot pain in the long run. Based on findings, Ergonomics solutions are proposed to improve the ergonomic practices so as to alleviate the users or Bajaj drivers’ difficulties and risks.

Key words- Ergonomics, Bajaj, Task analysis, Ergonomic solutions

I. INTRODUCTION

Ergonomics is the study of the interaction between people and machines and the factors that affect the interaction. Its purpose is to improve the performance of systems by improving human-machine interaction [1].Ergonomics finds a better fit between people and the things they do, the objects they use, and the environments in which they live, work, travel, and play [2]. Research has shown that the application of Ergonomics principles and programs in any workplace resulted in fruitful outcomes beyond their costs [3]. The goal of Ergonomics in the workplace is to improve efficiency, quality, and job satisfaction by making routine and repetitive tasks more comfortable and easier to do. By lowering the fatigue factor and human error it reduces stress, both physical and psychological. This in turn leads to higher productivity. Ergonomics must be an integral part of design, manufacturing, and use. This applies to transportation system such as the three wheeler-four stroke automobile, Bajaj which is the common means of public transport and carrying freight in Ethiopia. The three-wheel design of Bajaj provides a better road grip than a bicycle. It also provides gasoline savings when compared to a four-wheeler taxi[10]. The design of Bajaj has a number of Ergonomics related problems. If these problems do not get solutions, many Bajaj drivers could be subjected to serious back and foot pain.

Following [2], the knowledge of how the study of posture, repetitive motion, and workspace during Bajaj driving and the design Bajaj parts, controls, switches and displays affect the Bajaj drivers is critical to better understand Ergonomics of Bajaj as they relate to the drivers’ needs. This enables to design an improved driver-Bajaj interface which is more compatible with the task and the driver. This study examines the Ergonomics risk of driving Bajaj and proposes Ergonomics solutions so as to alleviate Bajaj drivers’ difficulties and risks. Accordingly, a descriptive study has been conducted and Bajaj drivers from Adama City have participated in the study. There are about 3000 Bajajs working in Adama city. Task description and analysis has been done based on data collected through observation, interview, and questionnaire to check the risks and to understand drivers’ reaction towards Bajaj driving and to get insight into issues and potential improvements for the system. The rest of the paper has been divided in to three sections. Section II presents task description of selected Bajaj parts to identify drivers’ risks. Section III discusses analysis of the tasks leading to risks and the Ergonomic solutions. Section IV presents the conclusion.

II. TASK DESCRIPTION

A driver of Bajaj controls, observes, and manipulates the overall system of the Bajaj and every decision is left for the driver. In order to evaluate Bajaj as a system ergonomically, relevant parts of the Bajaj such as Steering


103

NCSTI-2012

bar, Foot break, Seat and controls, displays and switches are analyzed.

A. Steering bar A Steering bar is manipulated by the Bajaj driver to guide the Bajaj to the required direction and the rest of the steering system such as front wheel. That is the steering system responds to the driver inputs. Major parts of the steering bar are illustrated in Figure 1.

Figure 1: Parts of the steering bar

The steering bar is not adjustable and a repeated loading and twisting of the wrist is required during driving.

B. Foot break It is observed that almost all the Bajaj drivers put their leg on a foot rest with size 150x100x30mm, Figure 2(a).

Figure 2 (a): The foot break

But, drivers put their foot on a steeped floor while pressing the foot break. Besides, there is no enough space or clearance between the end/edge of the driver seat and the foot rest. Therefore, the driver is required to turns his/her knee to 900 to press the break as illustrated in figure 2 (b).

Figure 2 (b): Foot position

C. The Seat A Bajaj has fixed seat and fixed back rest having the size and work spaces tabulated in table 1. The drivers sit for long period of time.

Table 1: The size of Bajaj seat

D. Controls, Displays and Switches The control, displays and switches, Figure 3 provide direct physical interface between the driver and the Bajaj system.

Figure 3: Controls, Displays & Switches

On/Off switch is located under the dash board and has three positions: OFF, all electrical circuits off except the stereo, ON, all circuits on and PARKING, parking light and hazard operation. Light switch, Figure 4 has three positions: Top, all lights “OFF”, Middle, parking and tail lamps will be “ON” and Bottom, head, tail and parking lamps will be “ON”. Wiper motor switch has two positions: “ON” and “OFF” wiper motor while Horn button is used for sound horning.

Figure 4: Right hand switch

Side indicator switch, Figure 5 has three positions: Center-all side indicator remains “off”, Left hand (LH)-

Name of the parts Size

Seating depth (Distance between the dashboard/steering bar & the backrest)

70cm

Seating height 37cm Knee height 45cm Seat length 40cm Back seat inclination 950

Distance between the dash board & seat 30cm Steering handle height from the floor 71cm

Gear change lever Speedometer

Accelerator grip

Speedometer

Tell tale display

Start button

Right hand switch

Left hand switch

Hazard switch

On/OffSwitch

Light switch

Hornbutton

Wiper motor

Steeped floor

Foot rest

Foot break

Gear change twist grip

Dash board

The driver puts his foot on the

steeped floor and turns to the right


104

NCSTI-2012

front and rear, LH side indicator will flash and Right hand (RH)- front and rear, RH side indicator will flash.

Figure 5: Left hand switch

Light switch/Tell tale display, Figure 6 consists of four light switches or displays. It is placed on the dash board in front of the driver at 70 cm from the floor which requires forward inclination of the head and trunk of the driver.

Figure 6: Tell tale display

III. TASK ANALYSIS & SOLUTIONS

A. Task Analysis: Findings The most commonly used method for Ergonomics analysis, Task analysis [4] and [5] is used to study whether the interface of Bajaj and the driver is compatible or not and to get insight into issues and potential improvements for the system .The findings of the task analysis is discussed as follows.

The finding of analysis revealed that

• 26% of the respondents know the potential risk of the awkward posture during Bajaj driving.

• 86% of the respondents agree that they are not comfortable if a passenger sits beside them.

• 73% of the respondents feel back pain and foot pain after being a Bajaj driver.

• 54% of the respondents drive 12 hours a day while 46 % drive for 7 hours a day.

• Only 40% of the respondents take rest in between driving and 93% of them take rest after 3 hours of continuous driving.

• 93% of the respondents agree if the position of the tell tale board is changed.

• 54% of the respondents agree that the position of the foot rest/pedal is not comfortable and they feel back and foot pain because of the turning position.

Speed change gear

High beam

Working condition Like all sedentary tasks, a Bajaj driver sits for an average of 8 hours a day. Evidence suggested that sitting at work, in itself, is quite harmless [6]. However, prolonged sitting at work (more than 95% of the day) such as driving is associated with back pain [7]. According to [8], this risk increases for those who drive for 20 hrs per week or more. The Bajaj drivers participated in this study sit for more than 8 hours a day implying that they are at risk.

Seat The seat back should fully support the driver’s back. But in the existing Bajaj design, the backrest height is too short to support the back of the driver to shoulder height. Back rest angle is about 50 from the vertical, which is too small to give the driver comfort during driving. Besides, the actual distance between the back rest and the steering bar/dashboard is 70cm. Therefore, to hold the steering handle the driver should extend his/her arms almost to maximum for more than 8 hours a day. Hence, there is high risk of static (no movement) and dynamic loads (movement) on the arms and wrists of the driver. Moreover, drivers are forced to turn to one side because the working space (distance between the dashboard & seat) is about 30cm, where there is no enough clearance/ resulting in high risk of back pain. The steering handle height from the floor (71cm) is also lower than the standard size (86cm) and not adjustable which make the driver uncomfortable.

Turning posture In order to provide space for the passenger who seats beside the Bajaj driver and due to the position of foot rest (Pedal), drivers practice turning posture of the back and the foot from the natural posture to some degree during driving, Figure 7. This awkward posture of driving leads to serious back and foot pain to all Bajaj drivers.

Figure 7: Turning position

Repetition

Dimmer switch

Low beam Side indicator

Battery charging indicatorHigh beam

indicator

Sideindicator

Neutralindicator

The driver is positioned at an angle to

the seat base


105

NCSTI-2012 Repetition refers to how many times you repeat an activity. The more frequently you perform forceful movements in awkward postures, the greater the risk of injury. For instance, there is a high risk of injury as result of repeated twisting of the accelerator grip.

Temperature and Noise There is no risk in temperature while driving Bajaj. Regarding the noise, since the driver of Bajaj uses horn which has 80-90 db it may cause risk of hearing if it repeated many times and cognitive behavior of the driver.

Whole body vibration Since there are a lot of up and downs in Adama town especially out of the main road the risk from whole body vibration is high.

B. Ergonomic Solutions Based on the task analysis, Ergonomics solutions are proposed under engineering, administration and work practice/driving headings.

Engineering Solutions The least amount of pressure on the back occurs when your seat back is 1000-1100 (150-300 from the vertical) so that you are slightly inclined. The backrest should be vertically adjustable and should have a height of 540 mm. The existing Bajaj seat has less size and angle. The standard seating depth as shown in Figure 7 is 28 inch (60cm), but in the existing Bajaj design the actual distance between the backrest and dash board is 70cm. The maximum reach distance is 27 inch or 68 cm. The actual distance between the back rest and the steering bar in the existing Bajaj design is 65cm. Therefore, to hold the steering handle the driver should extend his arms almost to the maximum. This variation is a risk to the driver hands and wrists.

Figure 7: Boundaries for vertical reaches for grasping objects [9].

The foot break should be redesigned considering values indicated in Figure 7 which requires an increase in the length or workspace of the Bajaj. To improve the right hand switch, the color of the switch has to be changed from

black to green so that the driver can easily identify the switches, controls and displays as illustrated in Figure 8.

Figure 8: Improved right hand switch

The same is true for the left hand switch (Figure 9). The Tell tale display position is not in the appropriate location for the driver to see it in normal sitting position. It requires bending to forward. Hence, the Tell tale display can be improved as shown in Figure 10 where the driver can observe signals easily without bending.

Figure 9: Improved left hand switch

Figure 10: Improved tell tale display

Administrative Solutions Since Bajajs are imported products, the concerned body should not allow importing with the existing design unless they are modified considering appropriate Ergonomics principles. Loading of passengers beside the driver should be also prohibited.

Work Practice Solutions • Even the best vehicle will only be truly beneficial if the

drivers take some responsibility for their own protection.

• Driving skills, emergency procedures and even customer-care may be delivered to each driver as standard.

• Specific training on issues of posture, how and why to make adjustments should be provided.

Light switch

Wiper motor Horn

button

Hazard switch

Side indicatorDimmer switch


106

NCSTI-2012

• Awareness should be created among the drivers on the potential risk of driving in awkward posture specially without taking rest.

IV. CONCLUSION

The three wheeler-four stroke Bajaj has a number of problems related to Ergonomics which need to be solved. To address this, there is a need to redesign the existing Bajaj which includes the design of the seat, the switches, displays and controls. It does not cost too much to solve the identified Ergonomics related problems of Bajaj as compared to the potential risks to the drivers.

REFERENCES

[1] Scott Openshaw, Ergonomics and Design a Reference Guide, All Steel Inc, 2006. [2] S. Pheasant, Body Space: Anthropometry, Ergonomics and the Design of Work, Taylor and Francis, 2004. [3] Martin, Helander, Guide to Human factor and Ergonomics, Taylor and France, 2006. [4] R.S. Bridger. Introduction to Ergonomics, Taylor and Francis, 2003. [5] A. Shepherd, Hierarchical Task Analysis as a Frame work for Task Analysis, In Task Analysis, Annett, J. and Stanton, N.A. (Eds.), Taylor and Francis, 2000.

[6] Hartvigsen J, Lebeuf-Yde C, Lings S, et al . Is sitting- while-at-work bad for your low back? A systematic, critical literature review, Scandinavian Journal of Public Health, (2000) 28:230–239.[7] Hoogendoorn W, Bongers PM, de Vet HCW, Douwes M, Koes BW, Miedema MC, AriensGAM, Bouter LM Flexion and rotation of the trunk and lifting at work are risk factors for low back pain. Spine, (2000), 25: 3087– 3092.[8] Porter JM, Gyi DE, Robertson J (1992) Evaluation of a tilting computer desk. In Contemporary Ergonomics, EJ Lovesy (Ed.),Taylor and Francis, 1992.[9] E. Grandjean, Fitting the Task to the Man: An Ergonomic Approach, Taylor and Francis. (1982).[10] http://minticetea.wordpress.com/tag/rickshaw/

Feasibility Study of Pumped Storage System for Application in Amhara Region, Ethiopia

107

NCSTI-2012


Mastewal Alemu Tilahun Department of Mechanical Engineering, Bahir Dar University, City: Bahir Dar, Postal code: 1225, Country: Ethiopia

email: [email protected]

Abstract — The major, almost 89% of the electrical production is from hydro power plant. Intermittent nature of the energy production and their inefficiency to meet peak load demands are the basic problems in renewable energy sectors. Tana Beles hydropower plant is the largest hydropower plant which starts to work in May, 2010 with an investment cost of $500 million and capacity of 460 MW. The project is planted in Amhara region using the water source of Lake Tana. To make this large and very necessary renewable energy resource sustainable using energy storage system will be vital. This study will figure out a pumped storage system for the hydropower plant for additional power production and for the sustainability of the water resource. The feasibility of the system will be considered technically and economically for the hydropower plant. Key words- Pumped storage system, Tana Beles hydropower plant, Amhara region, Ethiopia, Benefit of storage system

I. INTRODUCTION Ethiopia’s electric energy generation (KWh) according

to the data of EEPCO is 51.16 per capita. And also the installed capacity performance (MW) in year 2010 is 2060. Hydro covers 88%, Diesel 11% and Geothermal 1%; the diesel power plants are isolated systems and 60 MW is rented for facial time. From hydro power plants Tana Beles covers 22%. The total number of customers who are connected to get electric energy from the system amounts to1, 896,265 which is from 75.8 million population size of Ethiopia. The total energy generation and sales is 3.981 and 3.264 TWh respectively.

There is considerable resource of hydropower in most part of the country. The 88%- 90% [EEPCO, 2011] of Ethiopia’s Electric generation is based on hydroelectric sources.

Fig.1. Ethiopia Energy Production Share by Type [EEPCO, 2011]

II. TANA-BELES HYDROPOWER PLANT

Tana Beles is one of the largest hydropower plant which contributes 22% from the total power production of the country. The hydropower plant is new as the same time it is the largest from available hydropower plants.

Upper Reservoir

Lake Tana Full Storage Level: 1,787.0 masl Minimum operating level: 1,784.0 masl

Penstocks No branches: 4 Total length of the four branches: 155.0 m Diameters: 5.6/2.8/2.2 m ( max … min )

Turbines (4 x 115 MW) Vertical axis Francis Type Rotating speed: 375.0 rpm Number of units: 4 Runner Discharge diameter: 1.9 m Specific speed [m*kW]: 96 Rated head 40.0 m3/s design flow: 315.0 m Output at Rated Head: 115.0 MW

Tailrace Tunnel Invert level: 1,460.7 masl at ch 12+928 Length: 7.108 km Outlet portal with 1,450.0 m invert level of the portal 1,452.1 masl tail water level at design flow (Q= 160 m3/s)

The net head will be about 311 m and the rated discharge is 160m3/s. According to Salini and Pietrangeli (2006) the power plant will operate at a plant factor of 0.48, so that the average outflow from Lake Tana will be as high as 77m3/s. This will inquire 70% of the average natural outflow of Lake Tana. From the four turbines; the three are used at normal condition for the production of power and the fourth one is used as standby turbine for the time of maintenance or for power production at the time of high lake level. When there is a substantial risk of the water level dropping below the minimum operation level; the operation of only one turbine or complete shutdown of the plant may arise.

The minimum operation level of hydropower plant is 1784 masl and the maximum is 1987masl. The fluctuation on the lake level will result in a significant


108

NCSTI-2012

amount of reduction in energy. For example, the minimum level of 1784 masl and a minimum level of 1784.75 masl; when this two compared there is a reduction in energy of less than 2.5% however.

A. Lake Tana

The lake has 3000 km2 surface area and the depth of the lake is around 9.5 m which is about 14m below the average water level. For the hydropower plant operation a minimum of 1784 masl operation level is designed. The lake has multi functions like fisheries, navigation, tourism, and for environmental conditions.

1. Fisherie; it is a very important activity for the people around the lake. 25 different species and 15 endemic species are available on the lake.It has a catchment capacity of 15,000 tonnes which is around 43Kg/ha/year.[ BCEOM,1996a]. The plantation of this hydropower plant may harm the habitats (breeding areas) of the fish; which may show the decline of fish population.

Fig. 2. Lake Tana Photo 2. Navigation: there are lot of islands and isolated areas

where the only transportation method is based on the lake. Transportation of people, goods; for markets, schools, health facilities for almost basic needs of the rural areas. The minimal level of the lake should have to be 1785 masl. According to the Transport and Navigation Enterprise in Bahir Dar below this level navigation is restricted. This will not concede with the minimum working level of the Tana Beles hydropower plant which is 1784 masl. There was historical problem ones in June 2003; when the level of lake was 1784.26. At the level 1784.45 masl ships were not able to have a way on the lake because of the concentration of rocks. This did lead to a big problem on markets, schools and health services for those island people. Since the minimum level of the hydropower plant is much lower than the minimum level in which the ships can move; it has high risk on the transport and navigation works done on the lake. If the minimum operation level increases to 1784.75 there will be a reduction of about 25% in the storage capacity and energy production.

3. Tourism: There are 37 islands and from those 19 have ancient monasteries and churches on them. These historical places are most visited places for tourists. And Tis Issat fall (smoking water) is also most visited area. This is a place where large income is generated from tourists. Around 30,000 Ethiopia, 10,000 foreigners [Ethiopian tourism beuro, 2009]

per year visit the place. Because of the hydropower plant the beauty of the fall and the transport to the Lake Tana monasteries will fall to high risk.

4. Environment: the temperature of the cities around the lake is comparably very high. At summer time, the sun shines burningly. The wind from the lake is very important to minimize the hotness of the cities. And also from construction point of view many habitats will be distracted.

The target of the government is to:

- Maximize the power production - Make sufficient water available for the cropping

patterns - Assure the minimum level of the lake not to

limit the only available way of transport - Comforting and increasing the attraction of the

historic places for tourist - Working on the well-being of species of the lake

shores; limit the lake variation - Avoiding high water level in the lake; to

eliminate risk of flood. - The above listed agendas were tried to be

minimized on the plantation of the hydropower plant but still the water level will be the great problem because of the climate change our globe dealing with.

III. CURRENT SITUATION OF THE PLANT As the data from the power plant report indicates

there are two extremes, maximum flow or designed flow rate which is 160m3 /s and the minimum flow, 77 m 3 /s.

Fig. 3. Generated power for a year

The maximum generated power available is only for

five months. The other seven months generate under half of the designed capacity. The maximum generation periods are on the rainy season of Ethiopia. For the months; February, March, April, May, June, July and December there is a great decrement in generating capacity of the power plant.

Since these months are in summer season, it is facial time for the country where every sectors work fulltime. Which means the energy demand will reach its maximum. Since the power generation is not getting reach to the demand, there are days of blackouts again


109

NCSTI-2012and again. The available power is not sufficient enough to fulfill the current demand of the grid. And also the continuous shut down and half load working condition will definitely harm the parts of the plant

For the hydropower plant, four Francis turbines are instead with a capacity of 115 MW which will be ideal for most of the time in a year. This minimized capacity must be enhanced by applying different mechanisms.

IV. PUMPED STORAGE SYSTEM FOR TANA BELES HYDROPOWER PLANT

The problems in the site can be minimized or removed by the application of pumped storage systems. The efficiency and the capacity of the plant will increase and as the same time the sustainability of the project will be assured.

Two options can be applied for the application of the system: in the first case, the water can be returned to the upper reservoir (lake) then to the lower reservoir to generate power. Since all the turbines are Francis turbines they can work as a pump at the same time. Lower reservoir will be prepared for the purpose of storing water as a lower reservoir. The power needed to pump the water will be covered from the peak power production of the plant and/or from intermittent renewable resources like wind and solar- PV hybrid systems. The second option will be to build other mini hydropower plants using the water after the Tana Beles hydropower house; re-using the water again and again instead of leading to the Beles River will enhance the capacity of the water. Water is a resource of the globe which obliges us to be respected and utilized efficiently. This large amount of water leaving the turbine must be in generation of power. The water will be collected after the first power house then pumped to the higher reservoir and the same operation will continue. The power need for the pump will be covered by the peak load generation of the hydropower plant and/ or renewable intermittent energy resources.

The proposed pumped storage system is planned to work for 7 hours, the other 17 hours of the day the conventional hydropower plant work in its normal mode. The maximum volume outflow within 17hours in the rainy and dray season is 9,792,000 m3 and 4,712,400 m3. The installed power plant got two purposes which are power generation and irrigation this implies that all the volume of water in the lower reservoir is not to be returned back to the upper reservoir.

A. First Scenario, Five Months Where the Power Plant Work in Full Load; Q=160m3/s

For this study 24 hours flow was analyzed by considering full capacity of the plant where, flow rate is 160m3/s, and total power capacity of 460MW. The analysis was made using excel; from each hourly collected volume of water 40% will be released to the surrounding for irrigation purpose. Since there are vast agricultural plants proposed around the sites which expect large amount of water from the hydropower plant.

The maximum volume of water available in the lower reservoir after working for 17 hours is 5875200m3.

Flow rate The expected flow rate of water in the pumping mode;

to pump the water back to the upper reservoir for 7hrs; Q= (1) Where Q is volumetric flow rate, V, volume and t is time. The plant turbine flow rate; 40 must be taken as flow rate for the pump too, so as the pumping mode synchronizes with the generation. Six reversible Francis pump-turbine systems are necessary to pump back the water to the upper reservoir. The water in hours of pumping-generating mode will be diverted to outside for irrigation purpose. Each pump/turbine will have a flow rate of 38.857m3/s.

Pump Power for the First Scenario

Power input= (2) Where: ρ (Kg/m3) stands for the density of the fluid to be pumped; water in this case, g is for acceleration due to gravity (m/s2) and H, Q stands for both gross head (m) and volumetric flow rate (m3/s) respectively and the efficiency of pump considers all components water conductors, pump, motor, transformer which totals 78.0-90.02 %Frank et al, 2011].

Putting all values into the above formula, the pump power needed amounts to 141.844 MW. Total power consumed to pump the water to the upper reservoir taking the gross head of the hydropower plant amounts to 141.844MW for each flow rate, each pump, and total for the six pump/turbine amounts to 851.064MW.

B. Second Scenario, Seven Months, Where Plant Flow Rate Becomes 77m3/s

In these dry months the expected flow rate for the hydropower plant is 77m3/s which will lead to a lower water level in the lower reservoir, as the same time a great demand for electricity in peak hours. The government use diesel power plants to full fill peak hour demand. By considering the demand, in this scenario 24 hours gathered water will be pumped to the upper reservoir. So the recorded volume of the lower reservoir after 24hr is 5682600m3.

Flow rate The same relation is used as the first scenario and the

respective flow rate for the total volume of water considered will be 225.5m3/s. The same number of pump/turbines will be used and flow rate for each will be 37.58m3/s.

Pumping power Taking the gross head of the hydropower plant and

efficiency of the pump as before, the pump power for each will be 137.182MW. And total power which will be consumed for the six installed systems will be

823.094 MW. Scenarios can be compared and concluded as below:

QHg


110

NCSTI-2012

Table 1: Summary of the two scenariosCases Scenario 1 Scenario 2 No. of months 5 7 Available seasonal Flow rate (m3/s)

160 77

No. of working hours to collect the water to lower reservoir (h)

17 24

Available water in lower reservoir (m3)

5875200 5682600

Number of pump/turbines needed

6 6

Flow rate of pump/turbine (m3/s)

38.857 37.58

Pump power for each (MW) 141.844 137.182 Total pump power needed (MW)

851.064 823.094

V. AVAILABLE POWER FOR THE PUMP The power demand for the six pumps amounts

851.064MW each need 141.844MW to pump the water to the upper reservoir. Pumped storage system is power demanding system; it demands power to work. To acquire this power input, we have different methods such as power from the installed hydropower plant in the time of off peak hours, power from the grid in off peak hours and Wind, solar hybrid power.

A. Solar and Wind Potential

The hydropower plant is located 11°49′10″N 36°55′08″E11.81944°N 36.91889°ECoordinates and daily solar radiation and wind speed data available is taken from online sources at SWERA monthly renewable energy data; and wind speed_nasa_low [ SWERA, 2011].

Wind maximum speed is almost 3.7m/s and it is the most frequent potential available. The scaled annual average solar radiation is 5.84kWh/m2/d and annual average wind speed is 3.7m/s which is measured 50 meters above ground. Wind turbine used to be designed to reach its rated power within 10-12m/s wind speed which will not be possible in case of Tana Beles.

These values of wind and solar will made this option not feasible.

B. Power from the Grid

The grid holds the main active power sources of the country Tana Beles, Gilgel G. II, Tekeze, Gilgel G. I, Melka W. Finchaa, Koka, Awash II, Awash III, Aluto, Awash 7 Kilo. From these sources most are hydro. The active load from each power plant is considered for 23 days of a month and 24 hours of each.

The figure below starts at midnight from minimum load and starts to fall down till early morning till 07:00. Then start to rise to peak hour load and continues its cycle. The load varies between 384.75 MW and 903MW in each day with almost the same pattern. The very minimum power demand is after midnight till morning.

Fig. 4. Total Load Variation on the Grid for 23 Days

The greater active load is during June. Between the first minimum and last maximum points there is a load variation of 518.25 MW. The available power from sources also doubles the maximum active load of the grid.

The relation between peak load of a day, minimum load of a day and the capacity of the hydropower plants connected to the grid is shown below.

Fig. 5. Minimum, Peak Load and Capacity of Hydropower Plants in the Grid from Bottom to Top

Enormous amount of off-peak load is available in the grid; this load will be used for peak hours by applying pumped storage system. For peak hours the government used to use diesel power cycles but it is not efficient enough to fulfill the demand. Since June got the maximum distribution than other months, it will be used for further calculation.

C. Specific Share of Tana Beles Hydropower Plant in the Grid

The power plant capacity is 460MW as indicated in earlier parts of the work, but as indicated in the power graph it is not working in its full capacity. Especially starting from 23:00 (11.00 PM) the power drops to its minimum which is 100.80, 109.30, 107.20, 110.30, 109.30, 108.00, 158.10 starting from 00:00, 01:00, 02:00, 03:00, 04:00, 05:00, 06:00 respectively. The overall minimum and maximum power distribution and the capacity of the power plant is as shown in the below graph.


111

NCSTI-2012

Fig. 7. Power Generation Variation for Tana Beles Hydropower Plant

The variation between the capacity of the plant and the current working condition can be used as available power to be used for the pumped storage system. The hydropower plant will work in its full efficiency.

V. LOAD ARRANGEMENT FOR PUMPING POWER

The pumping hours start from 0:00- 6:00 and the load on the grid amounts 495.42MW, 458.62MW, 447.25MW, 445.21MW, 447.24MW, 472.00MW, and 521.21MW respectively. From the current peak load of the grid which is 913.93 MW there is a variation of 418.51MW, 455.31MW, 466.68MW, 468.72MW, 466.69MW, 441.93MW, and 392.72MW respectively. This can be directly used as a resource power for the pump with the addition of the power from the hydropower plant. The ICS system of EEPCO has 1758.2 MW of capacity from the hydropower plants. This power can be utilized efficiently in time of off-peak using this system in a profitable manner. Taking the constant middle line in between the pump working hours and others, the variation of load looks like as below.

Fig. 6. Average Load on the Grid Starting from Mid Night

As shown in the above load graph, there is a major power difference within 24 hours framework. For the seven points starting from 00:00 at midnight the load on

the grid is very little as compared to the maximum power distribution on the grid in time of peak hours. If the grid produces 913 MW for this pumping hours the pumped storage system will work efficiently. And also the maximum power share of Tana Beles hydropower plant in the grid in time of pumping is 158.10 MW, therefore there will be an excess of 302 MW, and the rest (549. 064MW) will be covered from the grid.

VI. LAY OUT OF THE PROPOSED PUMPED STORAGE SYSTEM

Fig. 8. Layout of Tana Beles pumped storage system

For the pumped storage system installation most of the

necessary parameters are already there. Upper reservoir, penstock, turbine, generator and tailrace are already in place. Two reversible pump/turbines will be added, four motor for the already placed Francis turbine will be installed. Lower reservoir will also be prepared with maximum volume design volume.

VII. BENEFITS The benefits raised from the application of the pumped storage system can be summarized as:

1. Maximize the power distribution for peak hour demand

2. Increasing water availability and performance of Tana Beles hydropower plant in dry seasons

3. To replace the diesel power systems with renewable energy in time of facial season

4. Increasing the quality and reliability of the electric service

To analyze the benefits, the system is going to be categorized under End-user/utility customer category [James et al., 2004].


112

NCSTI-2012Table 2: Summary of the benefits of pumped storage system

CRITERIA ADVANTAGE

Water availability

Since the water is going to be recycled, there will be power production in continuous manner

End user cost avoid

315.379 $/kW

Efficiency The system is planned to charge for 7 hours; with 81%

Reduced demand charges

7048.5252 $/ kW

Electric service reliability

Gives reliability for electric power services

Power quality Increases

Avoided peak generation cost

14.994$/kW-month

Emission Totally avoided

Fig. 10. Power distribution after the installation of pumped storage system

VIII. CONCLUSION

As the values of the table indicate applying the system

for the hydropower plant is beneficiary.

Tana Beles hydro power plant is one of the plants with high rated output, but the sustainability of the plant is dependent on the water level of Lake Tana. Because of climatic changes and natural flow the level of the lake is not constant. Especially in time of dry seasons it almost decreases by half. This variability makes the power plant less efficient than the designed and dreamed goal. This study checks the proposed problem and did suggest a pumped storage system for the power plant.

After the installation of the pumped storage system, the average power on the grid increases in time of both off-peak and peak hour. The pump power will be covered from the grid and the hydropower plant.

The proposed pumped storage system will have six reversible pump/turbines. Water will be collected for 17 hours in time of rainy season, 5 months and for 24 hour in dry season which is 7 months. The power available from the pump will be bought from the grid and Tana Beles hydropower. The pumped storage system will be beneficial in cost wise and in the coverage of peak hour power demand.

REFERENCES Frank S. Barnes, Jonah G. Levine., 2011. Large Energy Storage Systems Handbook (Mechanical and Aerospace Engineering Series). 1 Edition. CRC Press. Thomas, Kao, Robert., 1998. Standard Handbook of Powerplant Engineering. 2nd ed. New York: McGraw-Hill. EEPCO Beles Multipurpose Project (2006)

Fig. 9. Power distribution after the installation of pumped storage system

Figure 9 shows from bottom to top, power in MW of Tana Beles hydropower plant, Tana Beles power generation capacity, power distribution of the grid, and power demand for the pump for 24 hours.

International Hydropower Association., 2010. Benefits of pumped storage. [ONLINE] Available at: http://www.hydropower.org/psd/articles/introduction.html. [Accessed 22 December 10].

. Power for the pump is going to be taken from the grid and full load of Tana Beles plant.

Charging will be applied beginning from 00:00 to 7:00 and discharging at any favorable time for 6 peak hours. The pumped storage system will help the country by totally removing diesel power plants. This will add environmental and economical benefits. The power quality and quantity is going to be improved.

The pumped storage system will cover the power needed in time of peak hours as shown in the below figure. The above is hydropower plant with pumped storage system and the other one is pure hydropower.


113

NCSTI-2012


Abera Endesha Bekele, Ajit Pal Singh

Department of Mechanical & Vehicle Engineering, School of Engineering & Information Technologies

Adama Science & Technology University, P.O. Box 5008, Adama,Ethiopia

E-mail: [email protected]

Abstract — The globalization of the Ethiopian economy has thrown a great challenge to the Ethiopian cement industries in respect of productivity, quality, cost, delivery etc. The internal production process and supporting infrastructure should be such that it can compete successfully in the locally and global market with better flexibility and delivery. Cement enterprise is classified as continuous production process industry. Malfunction of any one machinery effects the other production machinery. Analysis of production processes is one of the ways to improve the productivity of the enterprise by identifying the bottlenecks/root causes. This research paper deal with rotary kiln machinery downtime, high energy (fuel), and refractory bricks consumption at clinker production area is analysed for improvement of productivity. By using Ishikawa (fish bone) diagram causes of root problem are identified and action plan suggestions are specified to reduce the hindrance. The overall equipment effectiveness (OEE) of rotary kiln was calculated and found to be less (i.e. 40.2%) as benchmarked with world class manufacturing OEE value (i.e. 85%). In the present cement enterprise studied, results revealed that maintenance is the root cause problem for the production process. If scheduled preventive maintenance (during and after) is followed up, following benefits are realized viz., smooth and continuous flow of materials, delivery scheduled is improved, production cost reduces, dollar value (i.e. 1,882,927 USD per year approximately) can be saved (in case of fuel, refractory bricks), which result in overall productivity improvement. Key words- Productivity, Cement production process, Ishikawa/Fishbone diagram, Overall equipment effectiveness, Rotary kiln.

I. INTRODUCTIONThe cement industry is ideal example of the

continuous industry sector and it will be to demonstrate that the production process analysis is applicable to all different production area [1]. There are numerous challenges facing the cement industry in today’s competitive environments; one of the major challenges is the capability of the cement industry to adopt and introduce the improvement approaches and techniques by which the overall enhancement can be achieved [2].

The need for improving the efficiency of the cement production line is widely acknowledged in order to reduce the energy consumption, downtime rates production and financial loss to satisfy high levels of market demand where the demand for cement is mostly second substance behind water [3].

Continuous industry and different organization types can be improved through eliminating or minimizing downtime of production machinery to reduce non-value added activities within the production line and energy consumption [4].

The production manager must find ways to improve efficiencies to increase production. This may be accomplished by increasing the speed of current production lines by investigating the root causes for the problem in the enterprise in each production line to reduce downtime, or upgrading

individual machines to improve machine availability.

Productivity can be measured either by partial productivity, ratio of output to a particular input, or by total factor productivity, ratio of output to weighted sum of all inputs. Partial productivity may sometime give a distorted picture of productivity growth. For example, if energy requirement per unit of output decreases then energy productivity will improve.

Clinker production area is the most critical cement production process area. It is also the largest capital expenditure area, the greatest source of fuel consumption and the main area of dust emission.

Kiln is the heart of the cement industry, as the critical step of “clinkerisation” takes place here. The kiln is the major consumer of thermal energy and also one of the major electrical energy consumers in the cement industry. Any inefficiency in the kiln section will reflect directly on the whole industry production process irrespective of the inefficiencies of the other parts of the industry.

Hence the kiln section needs to be systematically and seriously concentrated upon to achieve maximum energy efficiency [5].

Clinker production is the most energy-intensive stage in cement production, accounting for over 90% of total industry energy use. Also energy costs can account for to 40% of the total cost of cement production. For good quality clinker production,


114

NCSTI-2012less fuel consumption and smooth operation of kilns there must be maximum possible combination of the oxides in the kiln [6].

Also the kiln generates the largest quantities of dust and gases. It is well known that the nature and quantity of dust and gases from kilns depend on the characteristics of raw materials, fuel, process, burning conditions, kiln dimensions, system used, etc., which in turn govern the choice of the dust collection system and its efficiency.

The largest air pollutants in cement industries are the particulate emissions, which consist of carbonates, silicates, aluminates, fluorides and alkali halides, emitted through gasses at temperature of 120-350oC.

The quantities of air and gas have to be cleaned before being discharged into the atmosphere. Besides the particulate emissions, these gaseous pollutants play an important part in the air pollution. The gases from the kilns are generally identified as CO, CO2 and nitrogen oxides [4].

Improving of downtime is done by proper implementation of preventive maintenance and thereby increases the productivity. Productivity, at the organization level, may be considered as a measure of how well the company satisfies the customer utility [7].

To satisfying the customer need the enterprise shall produce planed production from each production machine [8]. Therefore, by investigation of production processes in Mugher cement enterprise (MCE), Mugher, Ethiopia, show how well an enterprise is doing and ways of improving the productivity.

II. METHODOLOGYIn this study, both primary and secondary

information sources were used. Interview, discussion and questionnaire techniques were used for gathering primary information. The secondary data sources used were technical documents and annual reports that will assist to get information concerning the study. Different journals, books, and websites were reviewed to strengthen the study with the fact and evidences of the science and technology. Data collected through direct observation and internal data collection.

A. Data presentation and analysis Analysis of the data is made by calculating the

OEE, utilization of machinery, throughout, and by making fishbone diagram. The data gathered were analyzed by using Microsoft Excel 2003 to make conclusion from the result.

Bar charts and line chart were used to show the trends of product efficiency, productivity, and machinery downtime. Wastes of raw milled materials during cement production process, downtime of machinery and root cause for the

problems were investigated and technically applicable solution was presented in the study. B. Steps of analysis production process Understand the process - This usually is the first step in order to be able to understand what are the inputs, outputs, steps and tasks that comprise the productive process. Visual inspection and observation will always provide a better understanding than theory or a process diagram alone. Collect data - Observation, production data collection, customer surveys, sales and marketing information can contain useful data in analyzing a process. Process/analyses data - This is the biggest and an important step in process analysis, this can be done in a variety of ways depending on the data available, complexity of the process and resources available to perform the analysis.

III. ANALYSIS OF PRODUCTION PROCESSFOR SEMI-FINISHED PRODUCT

Clinker product is a semi-finished product used for producing cement with some additive raw material. Clinker production process is the energy intensive production process. Rotary kiln is the production machinery used for producing clinker. It is found between raw mill and cement mill production machinery.

Capacity and efficiency of cement production enterprise is depends on the capacity and efficiency of clinker production machinery. Downtime of rotary kiln affects the two production machinery due to production flow is continues production process. Improvement availability, performance, efficiency and quality of rotary kiln used to improvement of other two production machinery and production volume and productivity of the MCE.

This study also focuses on clinker production process area to overcome the problems and to improve the productivity.

Rotary kiln in MCE have 65m long with internal diameter of 3.8m, external diameter of 4.2m, and has shell thickness of 0.2m. It is inclined at an angle of three degrees, and lined with temperature-resistant refractory brick of size 0.2m. They rotate at about 50 to 70 revolutions per hour. It has 1000tons production capacity in 24 hours.

From Fig. 1 the production volume and productivity of clinker is decreasing from time to time. In 1999E.C, the clinker production of the MCE was 604,458tons (Fig.1). In 2003E.C, it is 405,789tons. The ratio of output to input is 97.05% in 1999E.C and 89.05% in 2003E.C. Production difference within four years (1999 to 2003E.C) is 198,669 tons of clinker. Productivity is also reduced by 7.55%. This shows decrement in rotary kiln.


115

NCSTI-2012

0

200,000

400,000

600,000

800,000

1,000,000To

ns

Raw milledused forproducingclinker.

ClinkerProduced.

Raw milled used forproducing clinker.

977,689 996,453 879,463 817,652 725,364 709,864

Clinker Produced. 598,137 604,458 528,100 466,269 418,270 405,798

1998 1999 2000 2001 2002 2003

Fig. 1. Clinker produced of MCE from 1998-2003E.C.

Total excess spare parts used in 2003E.C for rotary kiln production machine are 130,636 numbers of bricks. The average cost of one brick for calcinations zone, transition zone, and sintering zone is 112.8birr. Also the average cost for one brick at inlet zone and outlet (cooling) zone is 10birr. The total financial losses due to brick spare parts for rotary kiln machine maintenance is 5,547,705.2birr for both production machines.

The amount that was invested for repairing/overtime rotary kiln production machine due to repeatedly breakdown is 979,853.70birr.

The total amount of financial loss in the year 2003 E.C due to repeatedly breakdown of rotary kiln production machine in case of fuel, brick and overtime is as follows:

A. Clinker production fuel consumption analysis The fuel consumption of producing one ton of

clinker is 98litres as per MCE standard. But from Table I in 1998E.C the average used fuel to produce one ton clinker is 99.5litres, this consume 1.5litre per ton extra. In 2003E.C, fuel consumption was increased up to 110.7litres, this consumes 12.7litre per ton extra. Fig. 1 shows clinker production volume decreases, but the energy consumption is increases as per Table I. This problem results in the low productivity of the MCE in the case of energy productivity. This also influences the capital productivity due to incremental of production cost.

28,344,990.3+5,547,705.2+979,853.70=34,872549.2birr

Taking the ideal condition, that is zero breakdowns, and if the preventive maintenance is performed in a way that doesn’t affect the production time, the MCE could have got an additional profit of as follows: 34,872,549.2 195,000,000100=17.9%

C. Rotary kiln downtime analysis Fig. 3 revealed that brick/refractory is the highest

defect with average of 63.7%, mechanical maintenance with average of 12.5%, and heating up with average of 11.3%. Fig. 4 shows the fishbone diagram for the major reasons for down time of rotary kiln. The root causes for these two reasons of downtimes can be grouped into machine, material, working environment, method, and operator.

TABLE IFUEL CONSUMPTION FOR BURNING FOR CLINKER

Years (E.C)

KilnProduction Time (hrs.)

Fuel Consumption

(litres)

Fuel Excess Used

1998 14,383 99.5 1.5

1999 14,514 100.4 2.4

2000 12,773 101.3 3.3

2001 11,429 106.9 8.9

2002 10,183 108.6 10.6

2003 10,464 110.7 12.7

0

50

100

150

200

250

Mechanical Electrical MCT Bricks Heating up Technology

Reason for downtime

Dow

ntim

e(hr

s)

March April May

B. Financial losses analysis due to breakdown of rotary kiln production machine

Fig. 3. Downtime of production line one kiln (2003 E.C).

The total amount loss annually due to excess fuel used for heating up of kiln to burn clinker due to repeatedly stoppage and also due to in proportion raw material input and other different related problem is 28,344,990.3birr. The reason why losses occur is due to frequent breakdown of rotary kiln machine.

D. Improvement suggestion plan for brick related problems

The area for improvement can be classified into operator, material, machine, work method and environment. Table II summarizes the improvement suggestion plan.


116

NCSTI-2012

.

TABLE II IMPROVEMENT ACTION PLAN SUGGESTION FOR BRICK

LINING

Type Action Plan Suggestion for Brick Lining

Brickbuilder (Kilnoperator)

Training must be given for maintainer.

Motivating worker to change the attitude/pay full attention for their duties.

Material Quality and proportionality of input material must be controlled starting from quarry. Because smaller variation on proportion and quality makes big difference.

Machine Overheating of machine must be controlled by controlling fed starvation, excess fuel, and kiln stoppages with the burner on, slowing the kiln down for long periods of time, defective burner pipe, and massive ash ring formation in the upper transition zone.

Work method and environment

The fan must work properly to keep the temperature at normal, for cooling of kiln, and also used material handling automatically rather than manual take brick to rich the kiln.

E. Comparison of world class OEE and one month calculated OEE value of the enterprise

The value of OEE is used for to know whether the machine is doing well or worse and tells where the

problem is mostly seen. The OEE percentage calculations for one month (May, 2003E.C) of line one rotary kiln production and world class manufacturing is shown on Table III.

TABLE III COMPARISON OF CALCULATED OEE AND WORLD CLASS

MANUFACTURING

OEE matrix World Class Manufacturing

Calculated Value ofMay, 2003(E.C)

Difference

Availability 90% 51% 39%Performance 95% 83% 12%

Quality 99.9% 95% 4.9%OEE 85% 40.2% 44.8%

OEE matrix’s is to help monitoring the machine to minimize losses and eliminate bottlenecks in process. From Table III mostly the three matrixes that is availability, performance, and quality needs improvement for improvement of equipment effectiveness. F. Equipment failure management and proposed maintenance work flow

Lack of scanner

Inspection Deformatio

MotivatioWorkingcondition

Overheating

Bricklining

AttitudeTraining

Lack of spare

Material handling

Brick TypesFlame characteristic

Safety

Burn-ability of raw material

Volatile

Machine condition

Operator

MaterialWorking method and environment

Fig. 4. Fishbone diagram for brick lining


117

NCSTI-2012

From existing system, lack of coordination between preventive and the corrective action of work flow is observed. And also the spare parts are not ordered at the work planning stage, thus the

downtime of the machine increases by waiting the spare part. Actually, this is not seen in the existing flowchart, but from their practice it is observed that the spare part order is at executing stage. Therefore, the workflow should be modified to alleviate the

above stated problems. Thus, the modified workflow, which considers the above problem, is proposed as shown in Fig. 5.

Every nature of production in cement industries is continuous, which calls for condition based maintenance system. It is most efficient interims of repair cost, downtime and availability of machine.

This model uses condition based maintenance type for these reasons. It is suggested that MCE should apply this proposed maintenance workflow to minimize the downtime due to repeatedly breakdown of machines as well as the maintenance cost and other related cost (Fig. 5).

IV. CONCLUSIONBy considering the burnability input raw material,

which is up to 1.5liter/ton additional fuel and enterprise will save 4,544,950.5 liters fuel by only minimizing the downtime rotary kiln and by improving the efficiency pre-heater. Enterprise saves, 4,544,950.5litres×5.5birr=24,997,230birr from the fuel saved and improve the capital productivity by 12.8%.

Currently in the MCE the waste gas dust content is 43.8g/m3 from calculation of dust content waste

gas from the enterprise, but the maximum standard gas dust content is 15g/m3 up to 28.8g/m3 waste gas dust content is released from the enterprise in 2003E.C. By reducing the fuel energy during pre-heating and clinker burning process by using the predictive maintenance before breakdown occurs, MCE can control the waste gas dust content to improve the productivity and protecting environment pollution and health of the workers.

The other major problem is brick lining problem in the rotary kiln, so due to brick lining repeatedly failure, MCE losses high production. To get the losses of production due to brick lining of rotary kiln


118

NCSTI-2012

is downtime of rotary kiln due to brick lining multiply by production capacity of the rotary kiln. Due to brick lining downtime of rotary kiln in 2003E.C is 2761hrs. and hourly production capacity of the machine is 41.7tons. By using preventive maintenance and following the strictly the condition of rotary kiln reduced the failures of the brick, decreasing the downtime of production machine, and increase the production by 16.55% by considering 1152hrs. for brick maintenance.

REFERENCES [1] H. Tulkens, “Efficiency of machinery analysis.

Some methodological issues and applications for cement industry,” Journal of Productivity Analysis, vol. 4, pp.183-210, 1993.

[2] K. Tsujimira, The Measurement of Productivity. Asian Productivity Organization, Japan, 1963.

[3] Z. Roman, Productivity and Economic Growth.Akademiai Kiaato, Budapest, 1982.

[4] E.R. Berndt, Energy use, technical progress and productivity growth: A survey of economic issues,” Journal of Productivity Analysis, vol. 2,pp. 67-83, 1990.

[5] T.J. Coelli, D.S.R. Prasad, G.E. Battese, AnIntroduction to Efficiency and Productivity Analysis. Kluwer Academic Publishers, 1998.

[6] S. Sattari and A. Avami, Assessment of energy saving opportunities of cement industries of Iran, 3rd, 2007.

[7] M.C. Eti, S.O.T. Ogaji, and S.D. Probert, “Development and implementation of preventive maintenance practices in Nigerian industries,” Applied Energy, vol. 83, pp. 1163-1179, 2006.

[8] A. Wilson, Asset maintenance management a guide to developing strategy and improving performance machinery downtime. 2nd ed., Conference Communication, 1999.


119

NCSTI-2012


Yonas Mitiku, R. Srinivasa Moorthy School of Mechanical and Industrial Engineering, Institute of Technology, Bahir Dar University, Bahir Dar, Ethiopia,

[email protected]

Abstract - The aim of this paper is to develop an algorithm for material minimization of straight-tooth bevel gears using an advanced and relatively recent heuristic search method called ‘Particle Swarm Optimization’ (PSO). The PSO Algorithm has many similarities to GA, in terms of initialization of the population, use of evaluation function, etc.; however, the way in which the PSO Algorithm traverses the search space is radically different and proves to have superior computational efficiency in many complex optimization problems.

Keywords: Particle Swarm Optimization, Genetic Algorithm, Population-based Search, Bevel Gear Design, Weight Reduction.

I. INTRODUCTION

Genetic Algorithm (GA) is increasingly being used in the design optimization problems as a potential tool for getting optimal and quick solution, particularly when the constraints and parameters involved are more and complex. The GA technique/tool has developed by leaps and bounds to cater feasible solutions to non-linear, multi-objective problems with several inherent complexities.

Optimization of design involves either maximizing quality or life or minimizing cost or weight. When the parameters involved are independent and the optimization is single-objective in nature with fewer constraints, then the problem becomes simple which can be attempted with the help of an algorithm which probes the entire search space in less time to achieve the optimization goal. But many of the optimization problems are multi-objective in nature with several influencing dependent parameters (non-linear) and conflicting constraints. This necessitated people to find a population – based algorithm which can smartly iterate and give quick convergence, though not accurate always.

Many improvements have been tried and tested to make GA a better efficient tool. This work is centred upon the application of one of the recent GA approaches called ‘Particle Swarm Optimization’ (PSO), for the material minimization of straight-toothed bevel gears.

Particle Swarm Optimization was first introduced by Dr. Russell C. Eberhart and Dr. James Kennedy in 1995. As described by Eberhart and Kennedy, the PSO algorithm is an adaptive algorithm based on a social-psychological metaphor [1].

Given the fact that bevel gears are highly used in gear boxes, machine tools and automobile

differential units and that the material minimization sequels weight reduction, cost-effectiveness, shortening of production cycles and increase in productivity, this work on design optimization of bevel gears using ‘Particle Swarm Optimization’ technique will prove to be a worthy attempt.

II. GA: A BRIEF OUTLINE

Genetic algorithms (GA’s) were invented by John Holland in the 1960’s and were developed by Holland and his students and colleagues at the University of Michigan in the 1960’s and the 1970’s. Evolution is, in effect, a method of searching solutions among an enormous number of possibilities. In biology the enormous set of possibilities is the set of possible genetic sequences, and the desired "solutions" are highly fit organisms—organisms well able to survive and reproduce in their environments. Evolution can also be seen as a method for designing innovative solutions to complex problems [2].

The salient features of GA are: 1. GA goes through solution space starting

from a group of points (initial population) and not from a single point.

2. GA uses information of a fitness function, not derivatives or other auxiliary knowledge.

3. GA use probabilistic transitions rules, not deterministic rules.

4. It is very likely that the expected GA solution will be a global solution.

Steps involved in Genetic Algorithm Optimization are:

1. Choosing a coding to represent problem parameters, a selection operator, a


120

NCSTI-2012

crossover operator and a mutation operator.

2. Choosing population size n, crossover probability pc, and mutation probability pm.

3. Initializing a random population of strings of size l. Choosing a maximum allowable generation number, tmax (Setting t = 0).

4. Evaluating each string in the population. 5. Terminated if t = tmax or other termination

criteria are satisfied. 6. Performing reproduction on the

population. 7. Performing crossover on pair of strings

with probability pc. 8. Performing mutation on strings with

probability pm. 9. Evaluating strings in the new population.

Setting t = t + 1 and going to Step 5 [3].

III. PARTICLE SWARM OPTIMIZATION

To overcome the drawbacks in traditional GA approach, several revisions and advanced tools and techniques have been suggested. One prominent version is Particle Swarm Optimization.

In PSO, a set of randomly generated solutions (initial swarm) propagates in the design space towards the optimal solution over a number of iterations (moves) based on large amount of information about the design space that is assimilated and shared by all members of the swarm [4].

A population of individuals (referred to as particles) adapts by returning stochastically toward previously successful regions. Particle Swarm has two primary operators: Velocity update and Position update. During each generation each particle is accelerated toward the particle’s previous best position and the global best position. At each iteration a new velocity value for each particle is calculated based on its current velocity, the distance from its previous best position and the distance from the global best position. The new velocity value is then used to calculate the next position of the particle in the search space. This process is then iterated a set number of times or until a minimum error is achieved [1].

Reynolds proposed a behavioral model in which each agent follows three rules:

1. Separation: Each agent tries to move away from its neighbours if they are too close.

2. Alignment: Each agent steers towards the average heading of its neighbours.

3. Cohesion: Each agent tries to go towards the average position of its neighbours.

Kennedy and Eberhart included a ‘roost’ in a simplified Reynolds-like simulation so that:

1. Each agent was attracted towards the location of the roost.

2. Each agent ‘remembered’ where it was closer to the roost.

3. Each agent shared information with its neighbours about its closest location to the roost [5].

Both Genetic Algorithms and Paticle Swarm

Optimizers share common elements, viz: Initialize a population in a similar manner. Use an evaluation function to determine

how fit (good) a potential solution is. Repeat the same set of processes for a pre-

determined amount of time (generational).

IV. OPTIMIZATION OF STRAIGHT TOOTH BEVEL GEARS USING PSO

Weight reduction of bevel gears is a thrust area of concern as it reflects positively in multiple issues like material and manufacturing cost, machining time, ease of handling and so. Bevel gear material minimization is based on the reduction of volume of cone frustums for both the pinion and the wheel, with multiple constraints related to module, contact strength and so. A. Objective Function

According to the volume formula of frustum of cone, volume calculation formula of straight tooth bevel gear pair can be expressed as [6]:

(1)

where, b is the face width, δ1, δ2 are the cone angles of pinion and gear respectively, m is the modulus, z1 is the number of pinion teeth, R is the cone pitch, z2 is the number of gear teeth.

B. Design Variable

The independent design parameters of volume of straight bevel gear drive include big end module m, number of pinion teeth z1 and face width coefficient . Hence the design variables are:

(2)


121

NCSTI-2012

C. Constraint Conditions

(i) Contact strength conditions:

The bending stress σH and contact stress σF of gears should not be more than the allowable values.

(3)

where, ZE is the elastic coefficient, ZH is the regional coefficient of pitch point, Zε is the coincidence coefficient, K is the load coefficient, T1 is input torque, u is the speed ratio and [σH] is the allowable contact stress. For straight gear, ZH = 2.5.

(ii) Tooth root bending strength conditions:

(4)

where, σF1 and σF2 are tooth root bending stresses, [σF1] and [σF2] are allowable bending stresses, YFSa1 and YFSa2 are tooth recombination coefficients and Yε is the contact ratio coefficient of small and large bevel gears, respectively.

(iii) The maximum peripheral speed conditions:

For straight bevel gear, the peripheral speed for the average diameter should meet:

(5)

where, vmax is the maximum peripheral speed.

(iv)Modulus constraints:

For power transmission gears, the minimum modulus should not be less than 1.5, i.e.

(6)

where, mmax is the maximum value of m.

(v) Face width coefficient constraints:

Usually, the range of face width coefficient is 0.25 to 0.3.

(vi) The condition to avoid root cutting for small bevel gear:

(7)

V. PSO ALGORITHM FOR BEVEL GEAR OPTIMIZATION

In PSO, each particle tries to modify its position using the following information: the current position, xid (t), the current velocity, Vid (t), the distance between the current position

and personal best (pbest) and the distance between the current position

and the global best (gbest). The velocity is updated by,

The position is updated using,

This Simple PSO approach is outwitted by a modified approach called Inertia PSO, in which in the current velocity computation, a inertia weight is added to the previous velocity of the particle so that the particle will continue moving in the same direction. The non-zero inertia weight is decreased over time. Hence,

The Inertia PSO Algorithm for bevel gear optimization is as follows: 1: procedure PSO 2: repeat3: for i = 1 to number of 50 do4: if F(xi) > F(pi) then /F() evaluates fitness/ 5: for d = 1 to dimensions do6: pid = xid /pid is the best state found so far/ 7: end for 8: end if 9: g = i /arbitrary/ 10: for j = indices of neighbours do 11: if F(pj) > F(pg) then12: g = j /g is the index of the best performer in the neighbourhood/ 13: end if 14: end for 15: for d = 1 to number of dimensions do 16.

/update velocity/ 17:

18: /update position/

19: end for 20: end for 21: until stopping criteria 22: end procedure


122

NCSTI-2012

where, t is the current time step, t − 1 is the previous time step. xid(t) is the current state (position) at site d of individual i. vid(t) is the current velocity at site d of individual i. ±vmax is the upper/lower bound placed on vid. pid is the individual’s i best state (position) found so far at site d. pgd is the neighbourhood best state found so far at site d. c1 social parameter 1, a positive constant, usually set to 2.0. c2 social parameter 2, a positive constant, usually set to 2.0.

is a positive random number drawn form a uniform distribution between 0.0 and 1.0.

is a positive random number drawn form a uniform distribution between 0.0 and 1.0.

is the inertia weight. The independent design variables are three –

module, number of teeth on small bevel gear and face width co-efficient as shown in (1). m and z1 are discrete variables and is a continuous variable. According to face width co-efficient constraint, it can be taken from 0.25 to 0.30 in steps of 0.01. Modules can be taken from the standard recommendations (1.5, 1.75, 2, etc.) upto mmax value chosen. The number of iterations can be taken as 50.

VI. CONCLUSION A generalized PSO algorithm was obtained for the non-linear constrained material minimization problem of straight-toothed bevel gears. The various parameters involved and the objective and constraint equations were discussed. The implementation of the algorithm using MATLAB,

Mathworks or any standard optimization tool will prove the efficacy of this advanced optimization technique. As it has proved in many other comparative tests, Inertia PSO is expected to perform better than the canonical GA approach, with a superior computing efficiency. The effectiveness can be compared only after the implementation of the obtained PSO algorithm.

REFERENCES [1] Mathew Settles, An Introduction to Particle Swarm

Optimization, Department of Computer Science, University of Idaho, Moscow, Idaho, November 7, 2005.

[2] Mitchell Melanie, An Introduction to Genetic Algorithm, A Bradford Book, The MIT Press, Fifth Printing, 1999.

[3] H. Ganesan, G. Mohankumar, K. Ganesan, K. Rameshkumar, “Optimization of Machining parameters in Turning process using Genetic Algorithm and Particle Swarm Optimization with Experimental verification”, International Journal of Engineering Science and Technology, vol.3, no.2, 1, pp. 1095. February, 2011

[4] Rania Hassan, Babak Cohanim, Olivier de Weck and Gerhard Ventor, “A Comparison of Particle Swarm Optimization and the Genetic Algorithm”, AIAA Publication, 2004. [5] Marco A. Montes de Oca, IRIDIA - CoDE, Universit´e Libre de Bruxelles (U.L.B.), “Particle Swarm Optimization: An Introduction”, May 7, 2007. [6] Xiaoqin Zhang, Qinhuangdao, Yu Rong, Jingjing Yu, Liling Zhang and Lina Cui, “Development of Optimization Design Software for Bevel Gear Based on Integer Serial Number Encoding Genetic Algorithm”, Journal of Software, vol. 6, no. 5, p.p. 2-3, May 2005.

Simulation and Optimization of Pump Case Casting

123

NCSTI-2012

Simulation and optimization of pump case casting

Abubeker Ahmed1, Mekonen Gebreslassie2

1 Department of Industrial Engineering, Mekelle University, Mekelle, Postal code: 231, Ethiopia, [email protected]

2 Mekelle University, Mekelle, Postal code: 231, Ethiopia, [email protected]

Abstract — Pump cases manufactured in Akaki basic metal industry exhibit defects like porosity, cold shut, blowholes and physical prototypes are done to try out the correct product. However, this methodology is expensive and most of the time partial. Therefore, this research deals with modeling, simulation, and optimization of pump case casting. The existing methodology is modeled and simulated to verify the defects and determine some of the mechanical properties of the casting. With the same methodology an optimized model is proposed and simulated.The main reason behind the formation of porosity is identified as the design of a rigging system and molten metal pouring rate. In the proposed model, a new rigging system is designed so that the formation of shrinkage porosity significantly reduces. It also results in better metallic yield, lower hardness, higher tensile and yield strength in all parts of the casting. In order to diagnose the problem and optimize the process design, casting modeling and simulation software named proCAST® is used together with CATIA V5 and GEOMESH. The result of a full scale experimental test shows a hardness number relatively lower than the virtual experiment due to a difference in heat transfer coefficient value between the actual and virtual mold. The metallographic test also validates the results of virtual experiment.

Key words- Porosity, Simulation, Optimization, Casting, Rigging system, Physical Prototype

I. INTRODUCTION

Casting is an economical way of producing components of required shape either in small or in larger lots. However, castings are less strong as compared to wrought components produced by processes such as forging [1]. All different casting methods do not result a similar property in the final product.Gravity castings (i.e., sand and die) have gained poor reputation for reliability and quality as compared to other types of casting methods; simply because their running systems have in general been badly designed [2]. The defects can be of porosity, blow-holes, shrinkage, mis-runs, hot tear etc. Apart from the limitations in the manufacturing processes, the caster’s knowledge and experience makes the defects more damaging. Modeling and Simulation of casting come to scene due to the need to minimize casting defects by understanding and controlling the flow of molten metal and thermal processes throughout the casting process [3]. However, modeling of casting processes is difficult because computational models

must consider variety of complex phenomena simultaneously, including complex fluid flows, heat transfer with solidification, nonlinear solid

mechanics, complex three dimensional mold geometries, and micro structural and defect evolution. Casting modeling and simulation provides the details about mold filling, solidification, and microstructure and stress formation phenomena. This gives important information about the non-uniform distribution of thermo-mechanical properties. Many foundries, around the world, use casting process modeling and simulation tools on a regular basis. Such tools provide valuable information for sound decision making on one side and are used for process documentation on the other. Here as well, trial and error is the key to success – however in case of an ‘error’ only a virtual product is lost, no raw material wasted, no tool has been cut and most important no production loss is accumulated.


124

NCSTI-2012

Casting optimization is a process of finding the best conditions (design and/or process) to reach a defined goal (minimize shrinkage, improve yield, etc...)[4].The decisions for optimizing a casting process must usually be done under time and resource constraints. Many actual prototyping are not permitted because of the expensiveness and time limitation. That is the main reason why experienced foundries utilize virtual modeling and simulation as a tool to undertake optimization. A. Problem Statement

Pump cases manufactured by the classic approach of trial and error exhibits porosity and many physical prototypes are done to try out the correct product. This method is costly, time taking and most of the physical and mechanical properties are not controlled. Hence this method needs to be improved/ substituted with a one which is better in quality and lesser in cost.

B. Objective of the Research

The general objective of the research is modeling, simulation and casting optimization of pump cases so as to validate and find a solution for the problems identified in the existing method. It is also to investigate how virtual modeling and simulation yields productivity and enhance quality. The Specific objectives include minimization of air-entrapments in the rigging system, minimization of porosity due to shrinkage, optimization of rigging systems design and avoidance of cold runs due to insufficient melt control in the main runner. C. Methodology

The methodologies used to study, verify and develop a solution for the problems found in the existing and proposed models are summarized as follows: The existing methodology is studied

computationally to understand the problems and develop an alternative solution. Thermal, fluid flow, stress and micro structural analysis are performed on the basis of their respective mathematical models.

Alternative solutions are studied computationally and the results are discussed by comparing with the existing methodology. The optimization is performed on the two proposed models based on the criteria of minimized defect and better productivity.

Finally, full scale physical prototyping is undertaken and tested for metallographic and hardness. Results of the tests are compared with that of the virtual model. After concluding remarks the study ends by recommending important points for future work.

In order to undertake the aforementioned tasks, state of the art modeling and simulation sofwares are utilized. Solid modeling and surface meshing are done on CATIA V5 and GEOMESH respectively. Then these semi processed data are exported to ProCAST® for analysis.

II. FUNDAMENTALS OF MODELING OF CASTING PROCESS

Casting process modeling involves the simulation of mold filling, solidification of the cast metal, microstructure formation, stress analysis on casting and mold, and etc. The fundamental principles are briefly explained as follows. Thermal Modeling In all transient casting heat-transfer analysis, there is heat transfer by conduction and convection from the molten metal to cause its solidification. [5] The governing finite element model is given as:

Using this basis, the heat capacity matrix [C] needs to be recalculated at each time-step to account for phase change and where non-linear thermal conductivity {k(T)} is present, the equation set needs to be solved iteratively within each step with an update of the thermal conductivity at each iteration[5].Fluid Flow modeling The flow of liquid metal may be assumed to be Newtonian and incompressible [6]. Leaving all the detail mathematical formulation aside, the final governing finite element model is given as.

Where M is the mass matrix, Ku is the velocity stiffness matrix, and Q. is the divergence matrix. The symbols fu and Δt denote the force vector and the time-step size, respectively. The nonlinear equations given above are solved by an iterative procedure (successive substitution). An unsymmetrical frontal solver is employed for simultaneous solution in each iteration, followed by under-relaxation [6].

III. RESULT AND DISCUSSION

A. The Existing Model

The pump case in production is converted to 3-D model using CATIA V5 software as shown in Fig. 1


125

NCSTI-2012

Fig. 1.The CAD model of the existing pump case with the rigging system Here, the existing pump case is modeled to study molten metal flow, thermo mechanical characteristics and the resulting defects due to the combination of different process parameters. Material definition of mold and cast

The casting material and mold used for the simulation of the existing pump case are grey cast iron; grade CG 2500, and Resin bonded sand having different composition fraction:

The convective heat transfer coefficient between the cast and the mold is taken as 500 W/m2k.The value is taken from the standards of the software database.

Flow rate for the existing method is calculated to be 4kg/s at 13500c.

The initial mold temperature is measured as 260c.Room ambient temperature.

Post processing of existing model analysis

a. Temperature distribution Temperature distribution of the pump case at the leg, suction periphery and main body are shown in the Fig. 2.

Fig. 2. Sample temperature distribution at three points

b. Shrinkage porosity The distribution of pores in the entire product for the existing model is shown in Fig.3. The sizes of pores are not shown in the figure. This results validates the existence of numerouspores which has been noticed at the shop floor.

Fig. 3. The entire distribution of porosity in the existing model

c. Velocity of molten metalThe velocity of molten metal at the sprue base and gate can be shown in Fig.4. These points are selected because they are the possible areas in which the molten metal gains high momentum. At the sprue base, the velocity of metal reaches a maximum value of 1.8 m/s at initial stages of the flow.But it doesn’t stay for long.Because as the tip of molten metal dirrectly collides to the sprue base/well,the reaction force makes to reduce its speed greatly. Afterwards, a uniform reduction in the pressure difference between the pump case cavity and the sprue makes the speed to constabtly reduce.

Fig.4. Velocity of the molten metal at sprue base and gate

At the gate the maximum velocity reaches 1.2 m/s. This value is approximately remains constant until the cavity fills. The reason behind is the relative

Main body

Suction periphery

Leg

Gate

Sprue base


126

NCSTI-2012

position of the ingate. It makes the pressure difference to be maintained throughout the filling time.

d. Mechanical properties The analysis of mechanical properties for the existing method indicates that higher values of tensile strength, yield strength and hardness are recorded in regions around edges and thinner sections. Because, these regions are susceptible for rapid cooling. The entire distribution of mechanical property can be shown starting from Fig.5 – Fig.7.

Fig.5. Tensile strengh of the pumpcase in the existing model

Fig.6. Yield strengh of the pumpcase in the existing model

Fig.7. Brinel Hardness of the pumpcase in the existing model

B. The Proposed Model

The proposed optimized model of the rigging system is mainly developed with the help of concepts suggested by J.campbell [7]. The total arrangement is generally selected because of the many advantages mentioned there. The 3D virtual model of the proposed pump case with its rigging systems is shown in Fig.8.

Fig.8. CAD model of the proposed casting with its rigging systems

Material definition of mold and cast

The casting material and mold used for the simulation of the proposed pump case are grey cast iron; grade CG 2500, and Resin bonded sand having different composition fraction:

The convective heat transfer coefficient between the cast and the mold is taken as 500 W/m2k.The value is taken from the standards of the software database.

Flow rate for the existing method is calculated to be 3.12kg/s at 13500c.

The initial mold temperature is measured as 260c.Room ambient temperature.

Post processing of proposed model analysis a. Temperature distributionAll points, at similar position to the existing method, have relatively gentle slope; because once the molten metal reaches at the points a continuous flow of metal over these points make to attain relatively similar temperature and equal exposure time. The flow pattern of molten metal at these points is downstream

b. Shrinkage porosity The distribution of pores in the entire product for the proposed model is shown in Fig.9. The sizes of pores are not shown here. This results clearly shows that shrinkage porosity significantly decreased in the proposed method.


127

NCSTI-2012

Fig. 9. The entire distribution of porosity in the proposed model

c. Velocity of molten metalAs can be seen in Fig.10, the velocity of molten metal continuously increases from 2m/s to 6.3 m/s at the sprue exit. Even if the value of the velocity is well beyond the suggested magnitude in the review, below the magnitude of this value the molten metal is prone to cold shut. The absence of a well and a direct connection of sprue with a runner make the molten metal to keep the integrity and avoid turbulence. This makes the molten metal to gain momentum and increases its speed.

Fig.10 . Velocity of the molten metal at sprue base and gate The reason behind the non-smooth pattern of velocity at the sprue base is a turbulence created by the collision of molten metal with a core. Turbulence due to collisions at less than 1m/s speed is tolerable. Around the gate region the maximum velocity is 0.8 m/s. This value is somehow maintained constant until 4.6 seconds. Afterwards it significantly reduced. When the moving molten metal inside the cavity starts to settle, the pressure difference decreases so that the velocity of the molten metal at the gate exit decreases.

d. Mechanical properties Similar to the existing method the mechanical properties have higher values of tensile strength, yield strength and hardness in regions around edges and thinner sections. Because, these regions are susceptible for rapid cooling. The entire distribution of mechanical property can be shown starting from Fig.11 – Fig.13.

Fig.11 Tensile strengh of the pumpcase in the proposed model

Sprue base

Gate Fig.12 Yield strength of the pump case in the proposed model

Fig.13 Brinell hardness of the pump case in the proposed model


128

NCSTI-2012

Summary of Comparison between the existing and proposed model

Parameters Existing Method

Proposed Method

Porosity Large and numerous

Small and Few

Metallic Yield

Additional 3 Kg to 3.5 Kg

3 Kg to 3.5 Kg less

Tensilestrength

315 Mpa to 354 Mpa

340 Mpa - 382 Mpa

Yieldstrength

271 Mpa to 338Mpa

224 Mpa - 336 Mpa

Hardness 218 HB to 242 HB

209 HB - 235 HB

The length and size of flakes in the two positions are different. Region nearer to the surface has thinner and shorter size flakes whereas region far from the surface has thicker and longer sizes. The difference is due to the change in rate of cooling. Relatively rapid rate of cooling creates thinner and shorter flakes and vice versa.

III.CONCLUSION The results of the study have shown that the rigging system used in the existing process is the main causes of shrinkage porosity, the proposed method yields 0.5 liter (3 to 3.5 kg) of molten metal per casting.

Table.1.Summary of results for the existing and proposed models

C. Experimental results The Brinell hardness test on the casting body gives an average result of 200 HB. This result has small difference with the value obtained from virtual experiment (209 HB to 235 HB).This is due to the difference in rate of cooling and accuracy of the instrument. Regarding the rate of cooling, it indicates that the heat transfer coefficient of the actual sand is less that 500 W/m2.K.Which results in slower rate of cooling and lower hardness.

Moreover, the whole body of casting does not exhibit uniform mechanical property in both models. The proposed method has relatively low hardness, higher tensile and yield strength in all parts of the casting. Except a relatively larger hardness in the main body of the pump case The virtual simulation can effectively predict defects, final mechanical properties of the casting, and increases productivity by eliminating many actual try-outs.

REFERENCES

The result of metallographic test at two different regions is shown in Fig. 14 and Fig.15. One sample is taken near the surface and the other deep in to the casting

[1] H.N Gupta, R. G. (2009). Manufacturing processes.New Dehli: New Age international.

[2] J.Campbell. (2000). Castings. Butherworth-Heinemann.

[3] International, A. (2008). Casting. USA: The materials information society.

[4] Casting optimization. (n.d.). Retrieved August 14, 2011, from esi-group.com: http://www.esi-group.com/products/casting/optimization/

[5] S.sulaiman, A. H. (2001). Modeling of thermal history of sand casting. Journal of materials processing technology , 245-250.

[6] K.Ravindran, R. (1998). Finite element modelling of solidification effects in mould filling. Finite Elements in analysis and design , 99-116.

Fig. 14 Distribution of flakes near the surface [7] J.Campbell. (2004). Castings Practice:The 10 rules

of castings. Oxford: Elsevier Butterworth-Heinemann.

Fig. 15 Distribution of flakes deep in to the cast


129

NCSTI-2012

Work-process-oriented approach of improving labor productivity in Ethiopian leather product manufacturing

Sisay Geremew Gebeyehu

Bergische University of Wuppertal, Department of Mechanical Engineering Gaussstr. 20, 42097 Wuppetral, Germany, E-mail: [email protected]

Abstract: This paper is aimed at introducing work-process–oriented approach of improving labor productivity in Ethiopian leather product manufacturing in an attempt to investigate how enterprise competitiveness can partly be achieved through engineering methods. While the concept of productivity (the ratio of output to input) is quite simple, the process of measuring and improving it is not an easy task particularly in a manufacturing setup. This is partly because the variety of inputs has always made productivity improvement challenging and leads to the application of diversified methods. The issue of labor productivity is usually viewed as a subject of human resource management and left for HR managers. However, there is a major stake for the engineering methods of productivity improvement. One possible way of addressing labor productivity improvement is therefore to investigate manufacturing work-processes with especial emphasis on the contents and contexts of particular working methods. Currently, the literature encourages the consideration of human resources as strategic factors, not only because they play important role in strategy implementation, but also because they are beginning to be reckoned as sources of sustainable competitive advantage. The research is relied on close investigation of 7 footwear manufacturing and 9 leather garment and leather goods making enterprises in Ethiopia. This study utilized a descriptive survey design to analyze the different factors that are potential symptoms of low labor productivity at workplace. Results challenge the conventional (managerial) methods of improving labor productivity and suggest the contribution of engineering methods as the bases for enhancing labor productivity. The investigation further indicates that without deploying the appropriate engineering methods, the application of soft managerial labor productivity improvement measures are unsuccessful.

Keywords: Labor productivity, work process analysis

I. Introduction

Generally productivity is viewed as the relationship between the output generated by a production or service system and the input provided to create this output.

inputoutput ty Productivi

Thus, productivity is the efficient use of resources-labor, capital, land, materials, energy, time, information- in the production of various goods and services [1]. One useful tool in productivity improvement is work process analysis. By identifying and analyzing the different tasks of

the manufacturing operation restraining productivity, it is possible to solve labor

productivity problems. In the usual way labor productivity enhancement programs tend to focus on overall enterprise wise issues by neglecting the immediate causes of the issue. Productivity measures reflect the operation of many factors such as changes in technology, equipment, capital investment per worker, utilization of plant capacity, layout and flow of materials and managerial skills [2]. However, Prokopenko[1] further explains that if productivity is defined for the individual worker as the relation of the volume of specific work done to the potential capacity of the worker (in numerical, cost or time terms), then for the enterprise or sector it can be expressed as the relation between value added and the cost of all input components. Work process analysis is usually used to describe manual and repetitive production jobs and is used by


130

NCSTI-2012

industrial engineers to determine standard rates of production, to set pay rates and to identify possible points of improvement in the production process. It involves the application of ergonomics and methods engineering. Globalization of the world economy has forced all nations including Ethiopia to be prepared to face a challenging business environment [3]. The driving sectors of the Ethiopian economy are also exposed to this competitive atmosphere and as a result, the leather sector gets more attention due to its importance in helping to enhance numerous economic activities. The Ethiopian leather sector is one of the most unexploited sectors with high potential for the country’s industrial development. However, its contribution for the national economy so far is negligible compared to its potential. The country has a major comparative advantage in the raw materials needed for the leather sector which makes it in principle very appropriate for leather product exporting. According to the report released by Business Development Service Ethiopia report [2008], it has the largest livestock production in Africa, and the 10th largest in the world. Ethiopia’s livestock population is currently estimated to be 50.9 million cattle, 26 million sheep and 22 million goats. Annually it produces 2.91 million hides, and over 10 million sheep and goat skins [4]-[5].[]Though Ethiopia has a large livestock population, its leather sector significantly lags behind many countries that are less endowed. Berhanu and Kibre [6] concluded that low level of capacity utilization which increases the unit costs of products, lack of sector specific skilled manpower, low quality raw material supplies, unreliable and periodical raw material supply are the major factors affecting the sector which add up on poor quality of the leather products. The highly labor intensive nature of the industry, combined with Ethiopia’s large resource base for the raw material needs of the industry and its reputation as a producer of good quality leather, recognized in international markets makes the leather sector a good candidate for a concerted effort to expand production and achieve competitiveness at the international level [6]. The low level operating capacity of the leather industries and their product quality problems is an indication of its weak potential for competitiveness in the international market as achieving competitiveness requires effective utilization of available resources, human and non-human. Inorder to address these problems, extensive study in the area improving the productivity of Ethiopian leather and leather products is very important. The objective of this paper is to discuss labor productivity improvement in the context of engineering methods. Specifically it intends to identify the role of work process analysis as a tool to improve labor productivity and to suggest a simplified model of employing

engineering methods. Thus, the focus is however, limited to the identification of factors contributing to low labor productivity at workplace in relation to actual work processes.

I. Methodology

This study utilized a descriptive survey design to analyze the different factors that are potential symptoms of low labor productivity at workplace and develop work process analysis approach as a technical tool to improve labor productivity. Representative sample of leather product manufacturing processes were taken to investigate and identify potential improvement points using engineering methods. A list of 16 leather product (leather shoes and leather garment and goods) manufacturing enterprises were taken in to consideration to assess the overall work process setup of the subsector. Sample enterprises were selected using stratified sampling technique due to budget and time constraints. However, care has been taken to ensure the representativeness of the enterprises in the sub-sector. Moreover, due to the lack of structured data showing employee productivity evaluation results, physical measures of productivity were not employed. This has indeed limitations as the measurement is not recorded continuously to generate certain data trends.

II. Analysis and Results

Though it is clear that any manufacturing process is a complex, adaptive and on-going social system, the research concentrates only on the job-related productivity analysis. Even in this particular category, the principal resources and the central factors of productivity are all the people and facilities in an organization that are participants in the value addition process of the manufacturing. Apart from the descriptive survey results showing a general trend of low labor productivity in the subsector, two of the most human intensive operations of leather product manufacturing processes (leather cutting and stitching/sewing) were closely investigated on how technical operators perform on the workplace. The leather product manufacturing in general follows either a batch production process where production lots may be produced at a certain regular intervals based on customer order or as a job-shop production where small lots of finished products with different varieties are produced for local market. As a result measuring individual employee productivity for enterprises that follow batch production is much complicated than the latter group of enterprises with


131

NCSTI-2012

job shop type of production. Thus, productivity in this sense is the degree to which the production line operators apply themselves to their work and the extent to which the application of their effort brings the desired results in output and quality. This is a function of the working method, personal skill, and the overall working environment of the manufacturing process.

In order to achieve the objective of the study, firstly, the flow of process to produce a single item was observed. From the observation, a total of 158 operators were selected in the sample enterprises. 61 of the operators were leather cutting machine operators whereas the rest 97 were stitching operators. Then, at each workstation operators were consulted for 6 predetermined factors which are assumed to be potential symptoms for reduced HR productivity at workplace. These factors were:

Individual performance levels Production line balancing problems Employed production technology Workplace conditions Work supervision systems and Other external functions like financial

incentives, HRM systems etc.

The survey also tried to concentrate on three fundamental effects that are the results of insufficient workforce productivity. These were formulated as the basic research questions. Q1.Which factors contribute most for productiondelays in the production line of leather product manufacturing?

The respondents identified production line balancing problems and individual performance levels as the major contributors of operator/machine idle times and hence increase in the production lead time. Table 1 and figure 1 shows the resulting percentage frequency table and sorted factor bar charts.

Fig. 1: Survey Q1. Responses sorted by ranking Q2. Which factors contribute most for materials

wastage in the production line of leather product manufacturing?

A similar survey suggests that material wastage and high scrap generation are mostly the results of poor working conditions and individual operator performance levels with response rate 23% and 22% respectively as shown in Table 2.

Table 2: Response to ‘which factor contributes most for material wastage at work place’Factor % of

responsesn=158

Individual performance level 22Production line balancing problems 17Production technology 17Work place conditions 23Work supervision systems 13Other external functions (Financial, HRM etc…) 11

Total 100

Q3. Which factors contribute most for product defects in the production line of leather product manufacturing?

Product defects are either damages on a product or faults made in the manufacturing process. Workplace conditions, the production technology employed and operator errors are found to be the major contributors for manufacturing faults. Table 3 shows the relative percentage of respondents of the survey.

Table 3: Response to ‘which factor contributes most for product defects at work place’

Factor % of responses

n=158Individual performance level 20Production line balancing problems 13Production technology 23Work place conditions 24Work supervision systems 18Other external functions (Financial, HRM etc…) 2

Total 100

Table 1: Response to ‘which factor contributes most for production delays at work place’

Factor % of responses

n=158Individual performance level 21

Production line balancing problems 37

Production technology 6

Work place conditions 19

Work supervision systems 15

Other external functions (Financial, HRM etc…) 2

Total 100


132

NCSTI-2012

Results of the survey suggest that employees’ productivity at workplace in the leather product manufacturing process is considerably influenced by the line balancing process, workplace conditions and individual operator performance levels.

A rank order list of factors obtained from descriptive data analysis is given in Table 4.

The survey was designed partly to determine what factors most influence the productivity of technical workforce in the Ethiopian leather product manufacturing enterprises. The most frequently noted factors limiting labor productivity were overall production line balancing and poor work place conditions. Respondents also identified individual performance levels and unfavorable work supervision systems as factors that need improvement in the leather product manufacturing setup. Therefore, this preliminary survey result implies that improving productivity requires that factors such as these be addressed and mitigated so that production develops into a predictable standard of performance. A simplified model for the work process analysis approach of productivity improvement (as shown in Figure 2.) is an important tool to apply engineering methods.

The above model depicts that by defining and critically measuring the actual work operations it is possible to discover and eliminate lost or ineffective times of the operation and compare operator performance against realistic expectations. The identification of non-standardization that exists in the workplace and non-value adding activities and waste again helps to analyze the working methods.

III. ConclusionTable 4: Rank order of factors affecting line productivity

Factor Rank

Production line balancing problems 1Work place conditions 2Individual performance level 3Work supervision systems 4Production technology 5Other external functions (Financial, HRM etc…) 6

As per the result of this preliminary survey, there is a need to focus on the engineering methods of labor productivity. In the surveyed leather product manufacturing enterprises productivity seems viewed as a more intensive use of such resources as labor and machines which should reliably indicate performance or efficiency if measured accurately. However it is important to separate productivity from intensity of labor, because while labor productivity reflects the beneficial results of labor, its intensity means excess effort and is no more than work speed up. As Leather product manufacturing is a process that involves remaking the same styles over and over and has got hundreds of repetitive operations for different styles, investigation such factors by breaking down hours of operations into their elements is very important. Moreover, issue of labor productivity should not be left totally as the job of human resource managers as there is a major role for the engineering methods to improve labor productivity. Management attention to these identified areas will likely lead to improvements in overall labor productivity in the leather product manufacturing work settings.

IV. References

1. Prokopenko, J. (1990). Productivitymanagement: a practical handbook. Oxford and IBH Publishing Co. Pvt. Ltd.

2. J.Rago, L. (1983). Production analysis and control. International textbook company

3. Loop, V. (2003). Industrialization, Value Chains and Linkages; The LeatherFootwear Sector in Addis Ababa, Ethiopia. Addis Ababa, Ethiopia: Addis Ababa University Press.

4. Ethiopian Investment Agency-EIA. (2008). Investment Opportunity Profille for the Manufacturing of Leather Garments and

Define and measure work operations

Design improved methods

Analyze operations & Operator

performance levels

Train, monitor and control improvements

Fig. 2. Simplified model of work process analysis


133

NCSTI-2012

Articlles in Ethiopia. Addis Ababa, Ethiopia: Ethiopian Investment Agency-EIA.

5. Central Statistical Authority of Ethiopia-CSA . (2010, February). Agricultural sample survey on livestock and livestock characteristics.Statistical Bulletin 2/468, pp. 1-94.

6. Berhanu , N., & Kibre , M. (2002, March). Declining Productivity and Competitiveness in the Ethiopian Leather Sector. Addis Ababa, Ethiopia: Unpublished.


134

NCSTI-2012


Woubishet Zewdu Taffese Materials Research and Testing Center (MRTC), Ethiopian Institutes of Architecture, Building Construction

and City Development (EiABC), P. O. Box 518, Addis Ababa, Ethiopia, [email protected]

Abstract — Construction and Demolition (C&D) waste has been dramatically increased in the last decade. Concrete waste accounts for the largest proportion and its utilization for new construction can significantly help to conserve natural resources, reduce waste disposal and haul which in turn produce both economic and environmental advantages. Therefore, the objective of this paper is to assess the suitability of recycled aggregate (RA) obtained from 100% concrete waste. Physical and mechanical tests of RA were conducted and compared to locally available natural aggregate (NA) and normal-weight aggregates (NWA). The test results revealed that RA had 25% and 20% lower unit weight in case of loose and compacted unit weight than NA, respectively. Ratio of loose unit weight to the compacted unit weight for RA is 0.88 and for NA is 0.90, which falls in the typical range of NA. RA’s bulk specific gravity is not in the typical range of NWA with 2.17, which is lower by 20% compared to the utilized bulk specific gravity of NA. In case of water absorption capacity of RA, 7.5% which is 4.2 times higher than NA. The Los Angeles abrasion value of RA is 32.88% which is 1.7 times weaker to resist abrasion compared to NA and do not fulfill the requirements for application of wearing surface.

Keywords- Recycled Aggregate, Sustainable Concrete, Waste Recycling

I. INTRODUCTION

Concrete is the most commonly used construction material in the earth. It is estimated that 25 billion tones of concrete are manufactured every year [1], which means 3.5 ton of concrete is consumed per person annually. Concrete is composed primarily of aggregate, cementitious material and water. Aggregate’s properties have a large influence on the properties of the concrete as it is the main ingredients in producing concrete which accounts for 60 to 75 % of the volume and 70 to 85 % of the weight of concrete [1]-[2]-[3]-[4]-[5]. It means that 15 to 19 billion tones of aggregate are consumed each year to produce concrete which is tremendous amount. In order to make the concrete industry sustainable, utilization of alternative aggregates is vital.

Construction and Demolition (C&D) waste has been dramatically increased in the last decade which represent approximately 35% of the total waste [7]. Among various types C&D waste, concrete accounts for the largest proportion which can be collected 75% from construction sites, 70% from demolition sites, 40% from general civil and 70% from renovation work [6]. The production of recycled aggregates (RA) from concrete and masonry wastes represents the most common case of recycling C&D wastes [7]. RA proved to be a good substitute for natural aggregate (NA) in the concrete production. It had been used more than five decades by using demolished concrete pavement for stabilizing the base course of road construction [7]-[8]-[9]-[10]. Its utilization for new construction can

significantly help to conserve natural resources, reduce waste disposal and haul which in turn produce both economic and environmental advantages.

RA should be hard, strong, dense, durable, clear and free from veins and adherent coatings; and free from injurious amounts of disintegrated pieces, alkali, organic matter and other deleterious substances [3].

II. GENERAL PROPERTIES OF AGGREGATES

Aggregates with satisfactory properties will always make good concrete. It is well known that aggregate strength is considerably higher than the normal range of concrete strength. It is also not possible to relate the potential strength development of concrete explicitly to mechanical properties of the aggregate [2]. Therefore, exact mechanical properties are not considered to be of much importance with the exception of very high-strength concretes. In this study, one of the mechanical tests, abrasion (Los Angeles test) has been carried out as it combines the processes of attrition and abrasion. This test results provides a good correlation with the actual wear of the aggregate in concrete as well as with compressive and flexural strengths of concrete. Instead of mechanical properties the most important aggregate properties are the particle-size distribution, shape, surface texture and absorption which considerably influence on the quality of fresh or hardened concrete. In this paper, properties of RA and NA is analyzed and compared to each other. In Addition, physical properties of RA is compared


135

NCSTI-2012

to typical ranges of physical properties of normal-weight aggregates (NWA).

III. PROPERTIES OF RECYCLED AGGREGATES

Recycled coarse aggregate obtained from a single source of 100% demolished concrete waste was examined. Its age expected to be around 55 years old. It was used as a machine foundation at Materials Research and Testing Centre (MRTC) of the Ethiopian Institute of Architecture, Building Construction and City Development (EiABC). The concrete was broken down into smaller part using a sledgehammer. Latter an ordinary hammer was used to break down further to make the aggregates in the acceptable size range. The physical and mechanical properties of RA are summarized in Table I and discussed in detail below.

A. Unit Weight and Specific Gravity

It is obvious that, the unit weight of aggregate often depends upon their packing, the particles shape and size, the grading and the moisture content. Since all the depending parameters of RA are varies with NA the unit weight was expected to be different. RA had 25% and 20% less unit weight for loose and compacted unit weight compared to NA, respectively as presented in Table I. The compacted unit weight of both RA and NA were not in the typical range of normal-weight aggregate (NWA) which are between 1280 and 1920 kg/m3 [11]. The ratio of the loose to compacted unit weight for RA was 0.88 and for NA was 0.93. Both are in the range between 0.87 and 0.96 where aggregates usually fall [2].

Bulk specific gravity of RA was 2.17, is lower by 20% compared to the bulk specific gravity of NA. It is not in the typical range of NWA as it is clearly seen in Table II. Indeed, exact value of bulk specific gravity is not the measure of the quality of the aggregate however it reflects a change in the porosity of the particle. A low bulk specific gravity may indicate high porosity and therefore poor durability and low strength [3].

B. Porosity and Water Absorption

Porosity is one of the most influential characteristics of aggregates which indicate its quality and it will significantly affect the density of the concrete. As aggregate represents three-quarters of volume of concrete it is obvious that the porosity and water absorption capacity of the aggregate contributes to the overall porosity and water absorption of the concrete. They influence the bond between aggregate and the cement paste, in turn, affect durability of concrete. Thus, determining porosity and water absorption of aggregate is often necessary. As it is presented in Table I, the water absorption capacity of RA was 7.5% which is 4.2 times higher than NA. Water absorption capacity of RA is extremely high that reflects the porosity of cement paste surrounding the RA. As it is clearly seen in Table II, the water absorption capacity of RA is even well above from the limit of typical NWA’s which is 4%.

Coarse Aggregate Property

RA NA

Fine Aggregate

Loose 1178 1567 1344 Unit weight (kg/m3)

Compacted 1344 1678 1400 Bulk 2.17 2.70 2.26 Bulk (SSD) 2.35 2.75 2.39 Specific Gravity Apparent 2.63 2.83 2.59

Moisture Content (%) 2.01 0.18 1.24 Clay/Silt Content (%) - - 5.34 Organic Impurities - - No – 2 Absorption Capacity (%) 7.5 1.78 5.6 L.A Abrasion (%) 32.88 19.11 - Nominal Maximum Size (mm) 37.5 37.5 - Void Ratio (%) 38 38 38 Fineness Modulus - - 2.2 Shape Angular Angular Well Round Angularity Number 5 5 - Texture Very rough Rough Smooth

Table I: Physical and Mechanical Characters of the Examined Recycled and


136

NCSTI-2012

C. Shape and Texture

It is well known that the particle shape and surface texture of aggregate considerably influence the properties of fresh and hardened concrete. Shape of RA and NA were angular with angularity number of 5 which is in the range between 0 and 10 where NWA usually falls. Whereas the surface texture of RA was rougher than NA. A rougher texture and a larger surface area of a more angular aggregate result in a greater bond between the particles and the cement matrix [2].

D. Particle-size distribution

Particle-size distribution is also one of the properties which affect the workability of fresh concrete. Particle-size distribution of RA and NA were determined by sieving according to a standard test method of ASTM C136. The particle size distribution is used to determine compliance of the particle size distribution with applicable specification requirements for aggregates. They are usually expressed as the percentage of material passing each sieve. The particle size distribution of the examined RA and NA were not fully in grading limits specified by ASTM C33 as presented in Table III. The nominal maximum size both NA and RA was 37.5 mm with the void ratio of 48% and 38%, respectively.

E. Resistance to Abrasion and Degradation

Aggregates should be also hard and tough enough to resist crushing, degradation and disintegration from any associated activities including production, placing and compaction of the concrete. The Los Angeles (L.A.) abrasion test was conducted to determine toughness and abrasion characteristics of RA in accordance with ASTM C131 and its value is compared to the average L.A value of NA. The abrasion value of RA was 32.88% whereas the average NA was 19.11%. It means that RA is 1.7 times weaker to resist abrasion compared to NA and is not in the accepted limit for wearing surface since it is above the maximum limit, 30%. But both RA and NA are suitable for other purposes as their LA values are well under 50% [3].

TABLE II RANGES IN PHYSICAL PROPERTIES FOR NORMAL-WEIGHT AGGREGATES USED IN CONCRETE [11]

Property Typical Ranges

RA NA

Absorption [%] 0.5 - 4 X √ Nominal Maximum Size [mm]

9.5 – 37.5 √ √

Bulk Specific Gravity 2.30 - 2.90 X √ Compacted Unit Weight [kg/m3]

1280 – 1920 √ √

Moisture Content [%] 0 – 2 X √

VI. CONCLUSION

Experimental investigation on physical and mechanical properties of RA was carried out. The physical test result confirmed that RA had a lower unit weight and bulk specific gravity with a considerable higher porosity and water absorption capacity compared to the NA. These properties were also compared to NWA. The test result showed that, RA is in the typical range of NWA in case of unit weight and loose to compacted unit weight ratio. However, the values are very close to the minimal threshold of the NWA. In case of bulk specific gravity, RA was just under the lower limit of NA. The shape of NA and RA was angular with the same angularity number but the surface texture of RA was rougher than NA. The mechanical property test confirmed that, RA is 1.7 times weaker to resist abrasion compared to NA. In general, some properties of RA have superior characteristics than NA and vice versa. Partial replacement of NA by RA can result even a better result in some hardened properties. Therefore, various mix proportion with NA and partially replaced by RA have to be prepared and examined thoroughly various properties of fresh and hardened concrete to conclude its effect on application.

REFERENCES

[1] CCAA: Use of Recycled Aggregates in Construction. (Cement Concrete & Aggregates Australia, 2008).

[2] A. M. Neville and J. J. Brooks: Concrete Technology (Pearson, Harlow 2010).

[3] S. K. Duggal: Building Materials (New Age International (P) Ltd., New Delhi 2008).

[4] Domone, P. L.J. "Concrete." In Construction Materials: Their nature and behaviour, by J. M. Illston and P. L.J. Domone, 89 - 220. London: Spon Press, 2001.

TABLE III PARTICLE-SIZE DISTRIBUTION OF RA AND NA WITH

ASTM GRADING LIMITS

Sieve [mm]

Cumulative retained

[RA]

Cumulative retained

[NA]

ASTM Allowable

Range 50.00 100 100 100

37.50 100 100 100

25.00 99 80 90 -100

19.00 77 51 40 - 85

12.50 45 21 10 - 40

9.50 32 6 0 - 15


137

NCSTI-2012

[5] S. Mindess, in: Concrete Constituent Materials,edited by G. N. Edward, Concrete Construction Engineering Handbook, chapter, 1, CRC Press, Florida (2008).

[6] V. W.Y. Tam and K. N. Le: Relationships among Demolished Concrete, Recycled Aggregate and Recycled Aggregate Concrete Testing, the 3rd ACF International Conference – ACF/VCA (2008).

[7] F. P. Torgal and S. Jalali: Eco-efficient Construction and Building Materials (Springer-Verlag, London 2011).

[8] S. H. Adnan, L.Y. Loon, I. A. Rahman, H. M. Saman and M. V. Soejoso: Compressive Strength of Recycled Aggregate Concrete with Various Percentage of Recycled Aggregate. National Seminar in Civil Engineering Research, Universiti Teknologi Malaysia (2007).

[9] R. I. Abdul, H. Hamdam, and A. M. A. Zaidi: Assessment of Recycled Aggregate Concrete, Modern Applied Science Vol. 3, (2009).

[10] G. F. Kheder, and S. A. Al-Windawi: Variation in Mechanical Properties of Natural and Recycled Aggregate Concrete as Related to the Strength of their Binding Mortar, Materials and Structures Vol. 38 (2005), p. 701 - 709.

[11] ACI Committee E-701: Aggregates for Concrete(ACI Education Bulletin E1-07, American Concrete Institute, Michigan 2007).


138

NCSTI-2012

Biogas Generation from Food Wastes From Students’ Cafeteria of Bahir Dar University

Libsu, S1, Chavan, R. B2 and Wonde, A3

1Bahir Dar University, Department of Chemistry, Bahir Dar 2Bahir Dar University, Institute of technology for textile, garment and fashion design (Iotex), Bahir Dar,

3Bahir Dar Universty, Department of Biology, Bahir Dar

“ABSTRACT” Bahir Dar University provides lodging and boarding services to a large number of students in its regular and continuing education programs. One such site hosting a substantial number of students throughout each year is the Main Campus, commonly known as the Peda Campus. Resident students in the campus are served meals thrice a day: breakfast, lunch and dinner. It is estimated that about 600-700 Kg of food is left over after the meals. The leftover food is collected by identified people and disposed off in various forms such as distribution to poor people, pig feeding etc. The waste food, consisting mainly of Injera (a staple food in Ethiopia) and bread, is carbohydrate-rich and can be converted into biogas through a process of anaerobic digestion. Biogas is not only a cheap and rich source of sustainable energy, but it also helps to prevent environmental pollution. In the present paper, a technology for the generation of biogas from a mixture of food waste from students’ cafeteria and cow dung is discussed. The floating dome anaerobic digestion technology is selected for biogas generation on account of its economics and easy installation compared to fixed dome digester technology. The food waste has been characterized in terms of total solid, volatile solid and carbon: nitrogen (C:N) ratio. Parameters that are crucial for biogas generation such as pH, solid content, and digestion time have been studied. Water-displacement method has been used for estimation of biogas generated. The flammability of the biogas generated is tested by using Bunsen burner. One of the major challenges faced during the standardization process pertains to the acidic nature of Injera, which hindered the generation of methanogenic organisms responsible for biogas generation. Attempts to neutralize the acid by alkali like sodium carbonate failed. However, optimum pH conditions for flammable biogas generation from the food wastes were achieved by using ash or natural limestone as a seeding material. The experimental results indicated that it is possible to get a flammable biogas which can substitute fire wood as a cooking fuel. The technology has been scaled up from batch wise laboratory scale digester of 20 L capacity to pilot scale digester of 150 L. The results of pilot scale digester indicated that it is possible to get a substantial quantity of combustible gas usable for water heating for the preparation of tea or coffee. It is planned to further scale up the digester to 1000 L and produce biogas in sufficient quantity to see the feasibility of its use as a cooking fuel.

Keywords – Food waste, injera, biogas, anaerobic digestion, floating dome digester, pH.

I INTRODUCTION Food waste is defined as uneaten portion of meals, leftover and trimmings from food preparation from household, restaurants and institute cafeterias [1]. The amount of food waste being generated from population can be reduced by offering smaller portions and applying better inventory. The other alternative solution is food donation to charity which was practiced in USA and Japan. In Ethiopia also part of the food waste is distributed among the poor people. Improper disposal of food waste causes odor and potential vermin (parasites) responsible for health hazards. Food waste is generally disposed off together with other municipal solid wastes (MSW) in landfill. With capacity of landfill gradually filling up and fewer landfills being commissioned, it is

critical to look for alternative disposal method. One possible method is composting. Composting is an aerobic process and it produces humus that can be used as fertilizers or soil conditioner. Incineration is another approach but it is not feasible because of the high moisture contents of food waste [2]. A Food Waste in Bahir Dar cityA considerable quantity of food is wasted from households, restaurants, eateries, educational institutes, jails, hospitals etc. The waste consisting of food, fruits, vegetables and other organic biodegradable components is known as organic fraction of Municipal Solid Waste (OFMSW). It is estimated that the private organization Dream Light Plc, which is responsible for the Municipal Solid Waste management on behalf of Bahir Dar City


139

NCSTI-2012

Administration, collects about 29 metric tons of waste food per day in Bahir Dar city. [3]Table I conveys the proportions of the Municipal Solid Wastes generated in Bahir Dar City by various sectors in the city. [4] Table I Municipal solid waste generated in Bahir

Dar City.

Category Waste generated Tons/day

Residential 54.49 Commercial HotelOther commercial

24.38

Institutional 17.08Street sweeping 4.05Total 100

B Types of waste generated in different categoriesTable II summarizes the proportions of various constituents that make up the Municipal Solid Wastes generated in Bahir Dar City.[4] Table II Components present in different waste categories

Component % Food 29.13 Paper 5.46Plastic 3.35Textile 1.54 Rubber 0.79Leather 0.12Yard waste 13.67Wood& charcoal 2.22Glass 1.55Metals 1.72Ash and soil 31.5Stones 4.79Animal remains 3.72Hazardous 0.36Electronic Waste 0.08 Total 100

It is seen from Table II that ash and soil constitute highest proportion (31.5%) of the municipal wastes, which can be attributed to the widespread use of firewood and cow dung as a fuel in majority of the domestic household. Food wastes constitute the second highest proportion (29.13%) of the municipal wastes. The food waste is usually generated from households, restaurants, eateries, educational institutes, jail, hospital etc. There is no authentic information on the amount of food wastes generated from student’s cafeterias of

the various campuses of Bahir Dar University. It is estimated that about 600-700 Kg of food is wasted students’ cafeteria located in the Main/Peda Campus of Bahir Dar University. The waste food, consisting mainly of Injera (a staple food in Ethiopia) and bread, is carbohydrate-rich. These food wastes can be subjected to anaerobic treatment which leads to biogas production that can be used for cooking, heating and electricity generation. This study looks into the potential production of biogas from food waste from a student’s cafeteria of Peda campus of Bahir Dar University. The food waste considered in the current study typically consists of injera, bread, pasta, other preparations consisting of peas, lentils, oils, etc.. The high moisture content of these food wastes is favorable for composting and anaerobic degradation

II MATERIALSA Food wasteFood waste consisting chiefly of injera, bread, pasta was collected from the student’s cafeteria of Peda campus, Bahir Dar University.B Floating dome digesterLaboratory scale floating dome digester (30 L digester and 20 L gas collector) was constructed using plastic cans purchased from Bahir Dar City market.C AlkaliSodium carbonate, lime stone, ash were used as alkali for pH control during digestion of food waste. D Water displacement systemWater displacement system was constructed using 1.5 L plastic bottle, glass tube and rubber tube.

III EXPERIMENTAL METHODSA Characterization of food wasteMixed food waste consisting chiefly of injera was characterized in terms of Total solid content, Volatile solid content and Carbon:Nitrogen ratio (C:N) using standard methods reported in literature. [5,6]B Measurement of biogas volume [7] Biogas volume was measured using water displacement method using standard procedure reported in literature. C Setting up the anaerobic digestion of food waste


140

NCSTI-2012

Anaerobic digestion of food waste was initiated using following quantities of a food waste, cow dung and alkali

Food waste 1 Kg Cow dung 1 Kg Alkali As required Water 20 litre

The cow dung acted as inoculums for providing methanogenic bacteria for anaerobic digestion. It also acted as a buffer to control the pH during digestion.The procedure for setting up the anaerobic digestion was as follows: Aqueous slurry consisting of 1 Kg cow dung and 5 L water was added to the digester and mixed well using a clean wooden stick. The food waste (1 Kg) was washed and converted into fine slurry by grinding in a kitchen mixer which was then added to cow dung slurry. The contents were mixed thoroughly by stirring with wooden stick. The final volume was made to 20 L with further addition of water. After having placed food wastes and cow dung slurry in the digester, gas collector in the inverted form was introduced into digester while keeping the gas outlet valve open. The gas collector was then pushed further into the digester till all the air from gas collector was expelled to get anaerobic digestion condition. The gas outlet valve was then closed and digestion was set till the biogas was generated as indicated by the rise of gas collector in the digester.Initial experiments were carried out without the addition of alkali. However, later, it was observed that the addition of alkali was essential to get generation of flammable biogas. D pH of digested slurryThe pH of digested slurry was determined using universal pH indicator having the pH range of 1-14. E Testing the flammability of biogasThe flammability of biogas was tested by burning the gas at the mouth of a Bunsen burner.

IV EXPERIMENTAL RESULTS AND DISCUSSION

A Initial experimentsThe initial experiments were carried out using food waste alone for anaerobic digestion. It was observed that no biogas was generated even after 15 days of digestion. While analyzing this observation, it was found that the initial slurry was acidic (pH 4-5)

which remained so throughout the period of digestion. The slurry of injera is typically acidic, which may be traced to the fermentation of teff powder which precedes injera preparation. Further acidity was developed during the digestion period maintaining the digester pH acidic.It is well documented in the literature [8] thatanaerobic digestion of organic matter takes place in three stages: a) hydrolysis of long chain hydrocarbon into smaller chain hydrocarbon; b) conversion of smaller chain hydrocarbon organic matter to acetic acid, fatty acid and hydrogen by acetogenic bacteria; c) conversion of acetic acid into methane and carbon dioxide (i.e. methanogenesis). This process is pH and temperature sensitive. The formation of acid at the second stage of anaerobic digestion lowers pH of the digester. The sequence of events believed to take place during anaerobic digestion is illustrated in Figure 1

Figure 1 Mechanism of anaerobic digestion of organic waste [8]

organic wastes (carbohydrate, protein, lipid)

Hydrolysis

sugar, amino acids, fatty acids

Acidogenesis

volatile fatty acids (VFA)

Acetogenesis

acetate, hydrogen, carbon dioxide

Methanogenesis

methane, carbon dioxide

Factors that influence the anaerobic process are pH, temperature and nutrients. Methane producing bacteria require a neutral to slightly alkaline environment (pH 6.8 to 8.5) in order to produce methane. Acid forming bacteria grow much faster than methane forming bacteria. If acid-producing bacteria grow too fast, they may produce more acid than the methane-forming bacteria can consume. This builds up excess acid in the system thereby inhibiting the activity of methane forming bacteria. Methane production may stop entirely.


141

NCSTI-2012

The major nutrients required by the bacteria are ammonium and phosphate. Household wastes are normally protein-rich and, as a result, there is not usually a shortage of ammonium ions in the decomposing waste. However, some additional phosphate may be required to ensure optimum process rates. Iron, manganese, copper, nickel, zinc, cobalt etc are among the micro-nutrients which methanogenic bacteria require in trace quantities. It is important to attain optimum conditions (pH 6.8-8.5, presence in sufficient amounts of micronutrients) for production of biogas by biodegradation process. [2]During the digestion of the food waste that consisted mainly of injera in the present work, the pH was found to be 4 which is far less than the ideal pH range. The acidity of the injera slurry is attributed to the initial free acid present in the fermented injera.B Determination of free acid in injera The presence of free acid in injera was determined by adding known volume of water to a known weight of injera. The fine slurry of injera was titrated with standard solution of sodium carbonate using phenolphthalein indicator. From this experiment it was observed that 14.31 g of sodium carbonate was required to neutralize the free acid present in 1 Kg injera.C Control of acidity during digestion Washing of injera before digestion The free acid in injera was removed substantially by washing it with water 4-5 times prior to digestion. After washing the pH of injera slurry increased from 4 to almost neutral. However, during digestion for 5 days, the pH of the slurry again decreased to about 4 due to formation of volatile fatty acids. As a result, no flammable biogas was obtained. D Control of pH with the addition of sodium carbonateA highly fermentable food waste like injera is hydrolyzed very easily during anaerobic digestion. If the first two stages of biogas generation process viz; hydrolysis to convert the high molecular weight substances to low molecular weight substances and acidogenesis/acetogensis with the liberation of volatile fatty acids and CO2 are rapid, the liberated acid builds up before the consumption by methanogenic bacteria and the acidity during digestion increases. The build of acidity inhibits the development of methanogenic bacteria and eventually the biogas generation fails in the third stage [9]Acidity during digestion can be controlled with the

addition of alkali like sodium carbonate, ash and

limestone. This process is known as alkali seeding [10,11, 12] Maintaining pH of the slurry in the optimum range was thought achievable by addition of appropriate amount of sodium carbonate to the slurry before the digestion is started. In a typical experiment, a pH value of 8 recorded just prior to start of digestion dropped to 5 from second day onwards. Although there was liberation of gas as revealed from the rising of gas collector, the gas did not burn. The non-flammability of generated gas indicated the dominance of CO2 in the gas mixture. This means that on addition of sodium carbonate the digestion process continued till stage 2 of biogas generation mechanism, leading to generation volatile fatty acid and CO2. However, the methanogenic bacteria did not get activated due to low pH 5 which is less than the ideal pH 6.8-7.5 required for activation of methanogenic bacteria. Thus the experiment to control the pH during digestion with the addition of sodium carbonate also failed. E Control of pH with the seeding of ash and limeSubsequently, attempt was made to control the pH with the seeding of ash and lime stone. Interestingly, in the presence of either ash or limestone, the gas generation started from second day onwards and the pH remained close to neutral. The gas generated was flammable as observed by burning of the gas with blue flame at the mouth of the Bunsen burner. These experiments were repeated successfully several times using varying amounts of ash and limestone. The success of the experiments involving ash or limestone in producing flammable biogas may be attributed, in addition to their alkalinity, to the presence of micro nutrients such as iron, manganese, copper, nickel, zinc, cobalt etc which methane-producing bacteria require in trace quantities. F Scaling up of biogas generationAfter confirming, on laboratory scale, the effectiveness of ash and lime in controlling the pH for biogas generation, the next obvious step was to scale up the experiments to confirm the observations. A floating dome digester of 150 L capacity was used. The conditions of laboratory experiments were repeated by feeding food waste and cow dung in equal proportion and controlling the pH with seeding of ash. On the third day substantial biogas was generated as indicated by the rising of gas collector. The generated biogas continued to burn for 1 hour at the mouth of Bunsen burner indicating the generation of sufficient quantity of biogas for boiling water for preparation of tea and coffeeG Work in progress


142

NCSTI-2012

In the present work, we have successfully accomplished generation of flammable biogas from food wastes and alkali seeding materials on both lab and pilot scale. Following aspects of the research are in progress 1. Quantification of biogas generation by measuring

the biogas volume by water displacement method and correlating the gas volume/Kg of food waste on the basis of volatile solids.

2. Standardizing the quantity of cow dung required along with food waste to get optimum biogas generation.

3. Further scaling to 1000 L capacity digester and testing the gas generated for cooking purpose.

4. Calculation of energy content of biogas and fire wood and determining the quantity of fire wood that can be saved during cooking at student’s cafeteria.

IV CONCLUSION 1. Control of pH in the range of 6.5-7.8 is essential

to get satisfactory generation of biogas from food waste.

2. It was possible to maintain the desired pH for biogas generation with the seeding of ash or limestone as alkali. Similar attempt made with sodium carbonate as alkali failed to produce combustible biogas.

3. The laboratory scale experiments successfully scaled up to 150 litre digester with the gas generation adequate to boil water for the preparation of tea and coffee, but not adequate for cooking.

4. The work is in progress for measurement of biogas volume, calculation of energy content of biogas and correlating to the quantity of fire wood that can be saved. It is aimed to further scale up the biogas generation using 1000 litre digester.

Acknowledgement The authors are thankful to the Office of the Vice President for Research and Community Service) for providing financial grant (BDU/RCS/SC/03/01) for the project without which the work would not have been possible. We are also grateful to the Bureau of Agriculture, Amhara National Regional State, for providing us with limestone free of charge. Thanks are also due to Mr. Fasika Kebede, technical assistant (Chemistry Dept. Bahir Dar University) for meticulously and whole heartedly performing the experiments.

References [1] http://en.wikipedia.org/wiki/Food_waste[2] Y.H Cheng,., S.X Sang,., H.Z Huang,., X.J Liu,. &

J.B Ouyang, “Variation of Coenzyme F420 activity and methane yield in landfill simulation of organic waste”, Journal of China University of Mining & Technology, Vol.17, No 3, pp 403-408., 2007

[3] “Assessment of SWM system of Bahir Dar town and gaps identified”, Forum for environment, June 2010, http://www.unep.or.jp/ietc/GPWM/data/T3/IS_7_4_ASSES.PDF

[4] “Solid waste characterization and quantification of Bahir Dar city for the development of ISWM plan”, Forum for environment, June 2010, www.doc-txt.com/Bahir-Dar-City.pdf

[5] J. I Eze. and K. E Agbo., “Studies on the microbial spectrum in anaerobic biomethannization of cow dung”, International Journal of the Physical Sciences Vol. 5, No.8, pp. 1331-1337, August, 2010

[6] AOAC (1995). Official methods of Analysis. Association of Analytical chemists 14th edn Alinton, Virginia 22209.

[7] P Parajuli, “Biogas measurement techniques and the associated errors”, M Sc Thesis, University of Jyväskylä, Finland, 2011.

[8] U.S. Environmental Protection Agency Region 9 Waste Programs, Organics: Anaerobic Digestion Science (2008). www.epa.gov/region9/ /waste.html [9] K.H. Chua, C. H. Yip, W. L. S. Nie, “A

Case Study on the Anaerobic treatment of food waste and gas formation”, International conference on building and construction technology, pp1-5, 2008

[10] J. De Haast, T. J. Britz and J. C. Novello “Effect of different neutralizing treatments on the efficiency of an anaerobic digester fed with deproteinated cheese whey” Journal of Dairy Research vol 53, pp 467-476, 1986,

[11] Chen Xiguang, T R. Romano, Ruihong Zhang, “Anaerobic digestion of food wastes for biogas production”, International Journal of Agriculture & Biological Engineering, December, Vol. 3 No.4 pp 1-13, 2010

[12] A. A. Adeyanju, “Effect of seeding wood ash on biogas production using pig waste and cassava peels”, Journal of Engineering and applied sciences, Vol.3 No.3, pp 242-245, 2008

Biogas Production using Anaerobic Co-digestion of Sanitary Wastewater and Kitchen Organic Solid Waste of Condominium House

143

NCSTI-2012

Biogas Production using Anaerobic Co-digestion of Sanitary Wastewater and Kitchen Organic Solid Waste of

Condominium House

Martha Minale1, Eng Teshome Worku2

1Department of Environmental Engineering, Addis Ababa Science and Technology University, Addis Ababa, P.O.Box 124221, Ethiopia. [email protected]

2Department of Chemical Engineering , Addis Ababa Institute of Technology, Addis Ababa, P.O.Box 385, Ethiopia. [email protected]

Abstract -- Addis Ababa is one of the fastest growing cities where high urbanization becomes a challenge. Consequently housing shortage is a big problem of the city. The municipality has launched a huge Condominium Housing Programme in response to the problem. However, sanitary wastewater and solid waste management’s are the critical problems to those houses. The wastes were collected and evaluated for its biogas production and fertilizer potential in order to solve the foreseen waste management problems. The physicochemical characteristics of the collected wastes were determined. A laboratory scale batch anaerobic co-digestion of both wastes with different mix ratio of 100:0, 75: 25, 50: 50, 25: 75 and 0: 100 by weight (sanitary wastewater: kitchen organic solid waste) were carried out at ambient temperature for 30 days. The amount of biogas and methane produced over the digestion period for those mixing ratios were compared. The highest biogas yields were obtained from 25:75 which is 65.6L and the lowest was from 100:0 as 9.5L. The percentage of methane gas obtained from the experiment was between 19.8-52.8%. From the study results, it is evidenced that the mixing ratio of 25:75 produced the maximum quantity of biogas and methane. With regard to the fertilizer potential of the digested sludge; composting and sun drying process were helpful for land application by completely inactivating the pathogen.Keywords- Condominium house, mixing ratio, biogas, compost, fertilizer

I. INTRODUCTION

Addis Ababa is one of the largest cities in Africa with 4 million people. The rapid population growth results in shortage of affordable houses. More than 80% of the populations are living in a shabby house, poor sanitation condition and less efficient service for waste management’s [1]. The city administration has decided to tackle the challenge of housing by supporting the construction of a massive, relatively low cost housing programme called condominium house. Currently in 119 sites, 77,430 houses have been built and transferred to the city beneficiaries. This program has brought a significant change in the image of the city. It also has technical and economical advantages due to the high population density with small area [2].

However, still there is a challenge with the programme such as providing adequate water, sanitation and solid waste management. In the case of solid waste, there are no recycling programs to exploit the organic fraction, instead the waste is gathered in a communal container around the house and hauled into the disposal site; results in

odorous environment and having negative impacts on human health. Additionally, the sewer lines of most condominium houses are connected to central WWTP; which receive large volume of wastewater beyond the designed capacity. One of the best options for treatment both wastes is the biological treatment of wastes, anaerobic digestion, which achieves both energy production and waste stabilization.

Co-treatment by anaerobic digestion of different types of wastes such as municipal solid waste, sanitary wastewater and other biowaste is a common practice for waste management [3]. The process provides improved nutrient balance from a variety of substrates which helps to maintain a stable digester performance that can steadily generate a high volume of biogas with high methane content. The objective of this study is to produce biogas using anaerobic co-digestion of sanitary wastewater and kitchen organic solid wastes generated from condominium house at different mixing ratios in laboratory scale experiments for proper urban development. It also investigated the potential use of digested sludge as a fertilizer.


144

NCSTI-2012

II. MATERIALS AND METHODS

A. Raw material and reactor preparation The sanitary wastewater and kitchen organic

solid wastes were collected from Gotera and Mickililand Condominium House, Addis Ababa. Lab-scale batch experiments were carried out at the laboratory of Chemical Engineering Department, Addis Ababa institute of technology. The reactors were two 5L cylindrical anaerobic digester W8 issue 3 armfield model. The temperature of each reactor is controlled by an electric heating mat wrapped around the external wall. The daily gas off-take from each reactor is taken to a volumetrically calibrated collector vessel operating by water displacement. A constant head, liquid seal device ensures that the gas pressure in the reactor is maintained at a constant value throughout the test run. B. Determination of sanitary wastewater and

kitchen organic waste characteristics Before digestion, the collected kitchen organic

solid wastes were shredded to an average particle size of 2mm (physical pre-treatment). Both types of collected raw wastes were analyzed for various parameters such as pH, moisture content(MC), total solids (TS), volatile solids (VS), biological oxygen demand (BOD5), chemical oxygen demand (COD), total nitrogen (TN), total phosphorus (TP), total potassium (TK), organic carbon, total metals and heavy metal (Mn, Na, Cu, Ca, Mg, Pb, Ni, Zn, Fe), total coliforms (TC) and fecal coliforms (FC). Most of the parameters were analyzed according to standard methods [4]. C. Co-digestion of sanitary wastewater and

kitchen organic solid waste at various mixing ratios

The sanitary wastewater and kitchen organic solid waste were mixed at different ratios of 100:0, 75:25, 50:50, 25:75 and 0:100 by weight, respectively. The kitchen organic solid wastes were mixed with water in order to maintain the total solid in the digester to 8-10 %, the desired value for wet anaerobic digestion. Inoculums sludge was also prepared from cow manure and was mixed with the ratio of 4:1 (feed: inoculums). The reactors were operated under ambient temperature, 25±2 ºC and the retention time was about 30 days.

D. Composting of digested sludge After the mix ratio based on maximum biogas

yield was determined, composting and sun drying process were performed on the digested sludge for three weeks to facilitate further use of the digested sludge as a fertilizer and for easy handling. The condition was sufficient to inactivate some of the pathogenic bacteria and helminth eggs [5]. The nutrient contents, heavy metal concentration and coliforms were also analyzed.

III. RESULT AND DISCUSSION

A. Characteristics of raw sanitary wastewater The characteristics of raw sanitary wastewater

were analyzed (Table 1). From the result, the sanitary wastewater contains considerable load of pollutants and pathogenic microorganism sources which may cause pollution of water and diseases to the community. However as compared to black water concentration, the physiochemical characteristics of sanitary wastewater show lower concentration value in TS, VS, BOD5, and COD. This might be in condominium houses sewerage system the grey water was mixed with toilet wastewater. Hence the grey water might dilute the toilet wastewater and ultimately affects its concentration. This result is also comparable with the finding of Duncan [6] reported as “grey water takes the largest percentage of domestic wastewater and hence contains relatively small concentration of organic pollutants compared to black water”. B. Metals and heavy metals concentration in

sanitary wastewater and comparison with standards

Additionally the concentrations of different metals and heavy metals were also analyzed so as to determine the suitability of sanitary wastewater for biogas production (Table 2). This analysis is used to assure whether a chronic toxicity exist or not for anaerobes in sanitary wastewater. As discussed above in condominium house sewerage system; the toilet wastewater was mixed with the grey water. From these perspectives, the wastewater generated from condominium house might limit the anaerobic digestion process due to the different metal and heavy metal ions. The sources of the metals and heavy metals might mainly from grey water since metallic ions and heavy metals were associated with grey water (wastewater generated from domestic activities


145

NCSTI-2012

such as laundry, bathing and kitchen sinks). Grace & Clare [7] reported a household wastewater might contain different metals and heavy metal contaminants and their major sources were from grey water. They also studied [7] some of the metals and heavy metals found in household wastewater were Mg, Na, Cu, Ni, Zn, Ca, Fe, Pb, etc which may come from a wide variety of chemicals such as detergents, soaps, shampoos, cosmetics, oil and grease.

The different metals and heavy metals might toxic to methanogenic microorganisms due to the high sensitivity for higher concentration. However Table 2 compares the result obtained from this study with maximum tolerable limits for anaerobic digestion. Therefore it is evidenced that the concentration of different metals and heavy metals in condominium house sanitary wastewater are within the tolerable limits for anaerobic digestion. C. Characteristics of raw kitchen organic solid

wasteThe physicochemical characteristics of raw

kitchen organic solid waste are shown in Table 3. The amount of VS and TOC were higher percentage while the amount of TN was relatively small percentage in kitchen organic solid waste. Moreover, the kitchen organic solid wastes are characterized by high percentage of MC (>80%). Luostarinen & Rintala[8] indicated that kitchen solid waste is rich in organic material and thus easily biodegradable. Table 3 also shows the result of the experiment were compatible with literature data. The higher value of TOC and VS indicates that the kitchen organic solid wastes is highly biodegradable and require other wastes containing high macronutrients. Therefore mixing the kitchen organic solid wastes with other organic feedstock is necessary in order to provide a nutrient balanced feedstock for AD. D. Characteristics of sanitary wastewater and

kitchen organic solid waste at different mixing ratios

The characteristics of sanitary wastewater and kitchen organic solid waste at different mixing ratios were analyzed (table 4). The pH ranged from 6.15-7.2. The increasing concentration of TS, VS, BOD5, COD as the percentage of kitchen organic solid waste increases, might be due to the highest organic matter content of kitchen organic solid wastes than sanitary wastewater. Claudia [9] reported the major factors for the increment of those parameters might be due to the higher solid

content of kitchen organic waste. Considering C:N artio, Michael[10] investigated as an important factor affecting the biological process in anaerobic digestion. It is often suggested that an optimum C:N ratio between 20:1 and 30:1for anaerobic digestion. Comparing the C:N ratio of each mixing ratio100:0 and 75:25 were below the optimum C: N value. This shows that the sanitary wastewater contains low value of organic matter and higher nitrogen content which may come from urine. The 50:50 and 25:75 mixing ratios were in agreement with the optimum C: N ratio. This is due to the smaller amount of organic matter in the sanitary wastewater may compensated by a high proportion of kitchen organic solid waste. E. Biogas production at different mixing ratios

The cumulative biogas produced during the experimental period for the different mixing ratio with time is shown in Fig 1. The mixing ratio of 75:25 (kitchen organic solid waste: sanitary wastewater) produces 65.6L of biogas and was the highest among the other mixing ratios used in this study. This may be due to the highly organic content of kitchen organic solid waste coupled with the supply of missing nutrients by sanitary wastewater make the C:N ratio within the desired range. This followed by the mixing ratio of 50:50, then 100:0, then 25:75, and lastly 0:100. Biogas productions for the mixing ratios were 52.7L, 50.0L, 23.7L and 9.5L, respectively.

Babel et al [11] conducted an experiment on anaerobic co-digestion of sewage and brewery sludge and found the maximum biogas production rate for mixing ratio of 25:75 (sewage sludge: brewery sludge). Amirhossein et al[12] investigated that the production of the cumulative biogas was high when the organic component in the sample was higher. The anaerobic digestion process for kitchen organic solid waste alone should given the highest amount of biogas due to the higher organic matter content but it produced less biogas as compared to mixture of sewage sludge and kitchen waste. This was due to the production of volatile fatty acid by the microorganism was more likely to accumulated rather than to release biogas. The lower production of biogas might be due to from the unsuitable C: N ratio [12].

Methane percentage was also analyzed for all mixing ratios and was shown in Fig 2. The highest methane percentage with an average of 52.8% was obtained for 25:75; while 43.2% for 50:50, 40.2% for 0:100, 28.2% for 75:25 and 19.8% for


146

NCSTI-2012

100:0(sanitary wastewater to kitchen organic solid waste). The activity of methanogenic bacteria is higher for the mixing ratio of 25:75 due to the higher organic content and suitable C:N ratio. Wendland et al[13] investigated that addition of kitchen refuse to black water (toilet water) improved the performance of the CSTR in terms of methane yield and COD removal efficiency.

Fig. 1: Cumulative biogas production with time at different mixing ratios (sanitary wastewater to kitchen organic solid waste)

Fig 2: Methane percentages with respect to different mix ratio

F. Removal efficiency of TS, VS, BOD5 and COD at different mixing ratio

The characteristics of the effluent were analyzed when the digestion was completed and the removal efficiency was calculated (Fig. 3). TS of the effluent were reduced to about by 34.2- 75.2%. Reductions in VS are also ranged from 56.2-82.0%. A higher total solid reduction is recorded by 0:100(sanitary wastewater: kitchen organic solid waste). Considering VS reduction, higher removal efficiency was achieved by 25:75. The

higher removal efficiency of VS than TS was a very good indication of high uptake rate of the organic fraction of total solids by methanogenic bacteria. From the percentage reduction of TS and VS, it can be concluded that co-digestion can reduce the amount and volume of kitchen organic solid waste which is disposed in dumpsites. It can also reduce the task of the municipality’s solid waste management sector. BOD5 and COD reduction ranged from 59.1-79% and 43.7-73.4%, respectively. The COD removal efficiencies over the duration of the experiment are comparable to those reported in the literature ranging from (55-75)% for co-digestion process [9]. The high removal efficiencies for COD are a good indication of the fact that the anaerobic co-digestion under proper operating conditions could be used for the treatment of sanitary wastewater and kitchen organic solid wastes before disposal.

Fig 3: Removal efficiency of TS, TVS, BOD5 and COD after anaerobic digestion for all mixing ratios

G. Characteristics of digested sludge compost and its suitability as a fertilizer

Among all mixing ratio, the digested sludge of 25:75, which produce maximum amount of biogas and methane was further treated using sun drying and composting process. The characteristics of the dried and composted digested sludge were determined (Table 5). The digested sludge compost shows near neutral pH values. The pH value meet the compost quality standards used for agriculture in Switzerland (pH < 8.2)[14]. The TN content is 2.7 %. It was also reported by Kuo et al[15] that in order for municipal solid waste compost to have fertilizing capabilities, the total nitrogen content must be in the range of 1-3 %. The available phosphorous and available potassium found are 47.16 and 67.52 mg/Kg, respectively. The same result was also obtained by Kuo et al[15] the range of available phosphorous in typical compost was mostly between 40 -110


147

NCSTI-2012

mg/Kg dry weights and available potassium content of compost should be 60-120 mg/Kg. Hence, the digested sludge compost used from this experiment had enough content of macronutrients that can be applied for crop cultivation. The TVS of digested sludge compost is 6% of TS. Heinonen-Tanski & van Wijk-Sibesma[16] suggested that the organic matter of the digested sludge compost is important to improve the soil structure; making more resistance to droughts and erosion during heavy rains.

A complete removal of TC and FC was also observed. This indicates that, the digested sludge compost can be used for safe agricultural purposes which can support urban agriculture. The result is compatible with that reported by Ludwing[17] a complete removal of microorganisms and parasite observed when the fermented slurry was dried using direct sunlight. Ellen [18] reported the limit value by USEPA for fecal coliform concentration is less than 1000 MPN/gram of solid. The

experimental results of digested sludge compost after sun drying for three weeks confirmed the complete removal of fecal coliforms.

The use of digested sludge compost as fertilizer has restricted applications based on heavy metal content since these digested sludge may contain varied amounts of heavy metals which may be toxic for human and animal consumption. Such limitations were published in Metcalf & Eddy[19]. From the result, none of the heavy metals measured are over the maximum established limit values. Issac [20] indicated that the compost made from digested sludge should meet consumer and market requirements. Some of the criteria that ensure the marketability are: the compost must be largely free of impurities, must not present any health hazards, the level of heavy metals and other toxic substances must comply with the standards and must have a visually attractive overall impression.

Table 2: Characteristics of raw sanitary wastewater Parameters Unit Concentration of SWW Toilet Water1

pH - 5.15 7.2-8.8 TS mg/l 7068 31,300-87,000TDS » 4794 6,000-22,000 VS » 4241 15,000-65,400COD » 15,097 36,600-175,000 BOD5 » 5586 14,200 - 52000 TN » 219.7 700 - 4050 TP » 202.7 67.2-98.4TK » 120.6 137.1-314.5 TC MPN/100ml 6.31 x 108 22×106

FC » 7.94 x 105 12 ×106

1Issac[20]

Table 3: Comparison of the characteristics of condominium house sanitary wastewater with maximum tolerable limit for anaerobic digestion

substances Unit Result from present study Maximum tolerable concentration for anaerobic digestion

Copper mg/l 0.28 1001

Manganese » 1.42 20.02

Nickel » 0.365 200-5001

Sodium » 264.5 3,500-5,5001

Calcium » 3.53 2,500-4,5001

Magnesium » 675.8 1,000-1,5001

Lead » 0.78 2.03

Iron » 11.3 20-1003

Zinc » 1.44 163.01

1Metcalf & Eddy[19]; 2Medhat & Usama[21]; 3Duncan & Nigel[22]


148

NCSTI-2012

Table 4: Characteristics of raw kitchen organic solid waste Parameters Unit Present study LiteraturepH - 5.3 5.51

TS % dry wt. 18.7 16.141

MC » 81.3 83.861

VS » 89.0 85-902

Organic carbon » 53.6 48-602

TN » 1.82 0.1-2.92

1Jayalakshmi et al [23]; 2Jeanger[24]

Table 5: Characteristics of sanitary wastewater and kitchen organic solid waste mixture at different ratios Sanitary wastewater to kitchen organic solid waste

Parameters Unit 100:0 75:25 50:50 25:75 0:100 pH - 6.15 6.9 7.1 7.3 7.2TS mg/l 7,068 14,395 25,328 48,789 56,084 VS » 4,241 9,357 20,265 42,324 49,915 COD » 15,099 42,128 59,871 95,344 135,863 BOD5 » 5,586 16,430 31,127 50,663 75,849 TOC » 2,356 4,950 11,258 23,513 27,730 TN » 220.4 448.6 513.2 849.4 894.7 TP » 203.1 364.7 381.8 419.5 578.2 TK » 120.2 179.3 212.5 235.7 276 .6 C:N ratio - 10.7 11.0 24.9 27.7 31.1

Table 7: Characteristics of digested sludge compost of 25:75(sanitary wastewater to kitchen organic solid waste) and comparison with standards

Parameter Unit Compost standard Result from the study pH - 7.5-8.5 7.9TS % dry weight - 94MC » - 6.0VS % TS - 16.0TN % dry weight 1-3% 2.7TP » 40-110 47.16 TK » 60-120 67.52 Total Cu mg/Kg 0.21 0.14Total Ni » 0.21 0.08Total Pb » 5.01 TraceTotal Zn » 2.01 0.56Total coliforms MPN/g - 9600 Fecal coliforms » <1000 897

1Metcalf & Eddy[19]

VI. CONCLUSIONS

Condominium residences sanitary wastewater contains excess valuable nutrients that can be used for the production of methane. Also the expected metallic ions and heavy metals are within tolerable limits and thus can be used for biogas production. On the other hand, kitchen organic solid waste are

loaded by high organic portion with the most valuable elements, carbon for anaerobes which lead the co-digestion of the two wastes had become higher yield of biogas.

The findings of this study show that, maximum production of biogas with maximum percentage of methane is obtained in mixing ratio of 25:75(sanitary wastewater: kitchen organic solid


149

NCSTI-2012

waste).Moreover, considerable average percentage removal of TS, VS, BOD5 and COD was found for this mixing ratio. From the percentage reduction of those parameters, a reduction in the volume of kitchen organic solid wastes may be achieved if AD is executed at the Condominium House.

A complete removal of pathogenic microorganisms in the digested sludge was observed after composting and drying process. In addition, the macronutrients content satisfied plant nutrient requirements. Moreover, heavy metals content was below the pollution control standard. Therefore composting and drying process could be the best options for reducing the potential risk by completely inactivating the pathogen of digested sludge as a fertilizer.

Generally, the co-digestion of sanitary wastewater and kitchen organic solid wastes generated from condominium houses in mixing ratios 25:75 enhanced the quality and quantities of methane yield. Thus the treatment option supports urban development by treating the organic waste at the source.

REFERENCES

[1] Yenoineshet, M. H., (2007). Integrated Housing Development Programs for Urban Poverty Alleviation and Sustainable Urbanization (The Case of Addis Ababa). International Conference on sustainable urban areas, 25-28 June. Rotterdam.

[2] AACHA., (2006). Addis Ababa City Housing Agency. Draft Report on Addis Ababa Condominium Program, Addis Ababa, Ethiopia.

[3] Voutsa, D.; Zachariadis, G.; Samara, C.; Kouimtzis, T., (1996). Evaluation of the toxic content of sludges produced during the biological treatment of municipal and industrial waste waters., J. Environ. Sci. Heal A, 31 (3), 657-671.

[4] APHA, (1999). Standard methods for examinations of water and wastewater, 20th Ed. American Public Health Association, Washington, DC.

[5] Binod, K. C., (2008). Dry Continuous Anaerobic Digestion of Municipal Solid Waste in Thermophilic Conditions, Phd Dissertation, Asian Institute of Technology, Thailand.

[6] Duncan, M., (2004). Domestic Wastewater Treatment in Developing Countries. Earthscan, Sterling, London.

[7] Grace, T. & Clare, D., (2006). Sources of critical Contaminants in domestic waste water, A literature review, CSIRO National Research Flagship Initiative, Australia.

[8] Luostarinen, S. & Rintala, J.,(2006). Anaerobic on-site treatment of kitchen waste in combination with black water in UASB-septic tanks at low temperatures. Bioresource Tech., 98 (2007), 1734–1740.

[9] Claudia, W., (2008). Anaerobic Digestion of Blackwater and Kitchen Refuse, Phd Dissertation, Hamburg Institute of Technology, Hamburg.

[10] Michael, C., (1979). A Chinese Biogas Manual, Popularizing Technology in the Countryside, Intermediate Technology Publication Ltd, China.

[11] Babel, S.; Sae-Tang, J.; Pecharaply, A., (2009).Anaerobic co-digestion of sewage and brewery sludge for biogas production and land application. Int. J. Environ. Sci. Tech., 6 (1), 131-140.

[12] Amirhossein, M.; Noor, E.; Sharom, M., (2004).Anaerobic Co-digestion of Kitchen Waste and Sewage Sludge for producing Biogas. Second International Conference on Environmental Management, Bangi,13-14September. Bangladeshi.

[13] Wendland, C.; Deegener, S.; Behrendt, J.; Toshev, P.; and Otterpohl, R., (2006). Anaerobic digestion of blackwater from vacuum toilets and kitchen refuse in a continuous stirred tank reactor (CSTR). Proceedings of the 7th Specialised Conference on Small Water and Wastewater Systems in Mexico, 7-10 March. Germany.

[14] Shiferaw, T., (2009). Comparative evaluation of the effect of municipal solid waste compost, farm yard manure, inorganic fertilizer and their combination on potato yield in Wolmera Woreda of West Shewa Zone, Oromia Region. Masters thesis, Addis Ababa University, Ethiopia.

[15] Kuo, S.; Ortiz-escobar, M.; Hue, N.; Hummel, R., (2007). Composting and compost Utilization for Agronomic and Container Crops. For the review book of Recent Developments in Environmental Biology ,Washington State University.

[16] Heinonen-Tanski, H. & van Wijk-Sibesma, C., (2004). Human Excreta for Plant Production. Bioresource Tech , 96, 403-411.

[17] Ludwing, S., (1988). Design and details of simple biogas plant, Federal Republic of Germany.

[18] Ellen, Z., (2004). Hygienic Implications of Small-Scale Composting in New York State. Final Report of the Cold Compost Project. Ithaca, New York.

[19] Metcalf & Eddy., (2003). Waste water engineering treatment and reuse, 4th ed. Tata McGraw-Hill publisher: New Delhi.

[20] Isaac, A. D., (2003). The potential for the use of upflow anaerobic sludge blanket reactor for the treatment of faecal sludge in Ghana.

[21] Medhat, M. A. & Usama, F. M., (2004). Anaerobic Digestion Technology for Industrial Wastewater Treatment. Eighth International Water Technology Conference, Alexandria, Egypt.

[22] Duncan, M. & Nigel, H., (2003). Handbook of water and wastewater microbiology. Elsevier, London.

[23] Jayalakshmi, S.; Sukumaran, V. and Kurian, J., (2007). Hydrogen Production from Kitchen Waste using Heat Treated Anaerobic Biogas Plant Slurry, International Conference on Sustainable Solid Waste Management, pp.356-362, Chennai, India.

[24] Jeanger, P. J., (2005). Optimizing Dry Anaerobic Digestion of Organic Fraction of Municipal Solid Waste. Masters thesis, Asian Institute of Technology, Thiland.


150

NCSTI-2012


Eshetu Getahu, Menilk gebeyehu, Fitfite, Nigus Gabbiye School of Chemical and Food Engineering Institute of Technology,Bahir Dar University,P.O.Box 26, Bahir Dar,

Ethiopia, [email protected]

Abstract: Petroleum based fuels are now widely known as non-renewable due to fossil fuel depletion and environmental degradation. Renewable, carbon neutral fuels are necessary for environmental and economic sustainability. Bio-fuel derived from oil crops such as biodiesel is a potential renewable and carbon neutral alternative to petroleum fuels. As crude oil price getting high, the need for developing renewable fuels has become acute. Alternative environmental friendly fuel is currently an important issue all over the world due to the efforts on reducing global warming which is mainly contributed by the combustion of petroleum or petrol diesel. With this regards, biodiesel is non-toxic, biodegradable, produced from renewable sources and contributes a minimal amount of net green house gases, such as CO2, SO2 and NO emissions to the atmosphere. To promote market competitiveness for biodiesel, it is necessary to develop cost-effective and technically sound processing schemes. Therefore, the overall goal of this work was to model and simulate biodiesel (Fatty Acid Methyl Ester “FAME”) production from oil and leather industrial wastes. To achieve this goal, several inter-connected simulation activities were undertaken. In the first place, the fundamental flow sheet was developed for the process. Then, the process synthesized and finally the performance of this flow sheet along with the key operating criteria, were identified by conducting computer-aided simulation using ASPEN Plus. Various scenarios were simulated to provide sufficient understanding and insights. Also, different thermodynamic databases were used for different sections of the process to account for the various characteristics of the streams throughout the process. Crude, degummed oil and leather industrial waste oils were represented by Triolein as the feedstock and also, mass and energy integration studies were performed to reduce the consumption of material and energy utilities, improve environmental impact, and enhance profitability. The separation column shows that biodiesel did not form azeotrope and easily separated using ordinary distillation columns and the purification distillation column shows that only four stages are enough to obtain standard quality biodiesel. Moreover, the simulation result shows that about 95% yield was obtained from the reactor, 99.9% of the entering methanol was recovered from the separation column and 99.4% purified biodiesel was obtained from biodiesel purification process. The model is flexible in that it can be modified to any feedstock composition, and changes in process chemistry and technology.

Keywords: Biodiesel, Modeling, Simulation, Aspen Plus, Distillation

1. Introduction

The consumption and demand for the petroleum products are increasing worldwide in the past decade due to industrialization and world population growth. This fact could be easily demonstrated by the increasing rate of world oil price from time to time [1]. In particular, the increase in crude oil price significantly affects

the economy of developing countries. Moreover, petroleum based fuels are widely known as non-renewable resources due to fossil fuel depletion and environmental degradation. As crude oil price getting high, the need for developing low cost renewable energy sources is becoming a focus of many investigators. Bio-fuel derived from agricultural biomass such as biodiesel is amongst a primary potential renewable and carbon neutral renewable energy alternative to


151

NCSTI-2012

petroleum fuels [1-2]. With this regards, biodiesel is non-toxic, biodegradable, produced from renewable sources and contributes a minimal amount of net green house gases, such as CO2, SO2 and NOx emissions to the atmosphere. Currently, a direct usage of oil seed from the agricultural sector has taken the greatest share for the production of biodiesel. For example, biodiesel can be produced from different kind of edible oil, such as soybean, canola, rapeseed, sunflower, palm, and coconut oils etc [2]. A major challenge for the commercialization of this technology using these virgin oil seeds however is the high cost of pure vegetable or seed oils, which constitutes between 70% and 85% of the overall biodiesel production cost. Thus, in order to reduce the cost of biodiesel production, alternative low cost raw materials that are available in large quantities must be considered. Among the alternate raw materials that have been investigated so far were non-edible oil crops such as Jatropha curcas [3]; animal fats such as lard [4] and beef tallow[5]; and waste cooking oil [6]. The advantage of biodiesel production from non edible oil source and industrial waste is associated to the value added products (increase economic), reduce environmental pollution load and solving social problems. For this particular study, the potential of wastes generated from oil and leather industries are investigated for the production of biodiesel.

Among the prominent industries, leather industry has an important role for countries economy in the light of its large scale of potential for employment, growth and exports. Ethiopian leather industry has taken its contribution for the world leather trade with its high export productivity. There are about 19 Leather Industries in Ethiopia [7]. The leather industry commonly uses hides and skins as raw materials, which are the by-products of meat and meat products industry. On the other hand, the leather industry is amongst highly pollutant generating industries and it produces huge amount of fat-originated solid and liquid wastes while processing hides and skins [8]. These wastes cause environmental problems and must be utilized. Most of the solid waste in the leather

production process is originated in the pre-tanning stages which are pre-fleshing, fleshing,shaving and trimming. Pre-fleshing process is carried out so as to remove flesh and to promote the penetration of the chemicals for better leather products. The fat content of the leather industry wastes is remarkable. However, these wastes are not utilized effectively and there is almost no application method to recover these wastes. One way to recover the leather industry wastes is using them as feedstock in biodiesel production due to their rich fat content [9] and modeling and optimizing the process to produce standard biodiesel effectively. Thus, the pollution caused from the leather industry wastes may be reduced significantly and more valuable products can be obtained by converting them to biodiesel for countries like Ethiopia where there is a huge investment on leather industry.

The plant oils and animal fats usually contain free fatty acids, phospholipids, sterols, water, odorants and other impurities. Because of these, the oil cannot be used as a fuel directly. To overcome these problems the oil requires a slight chemical modification, mainly transestrification, pyrolysis and emulsification. Among these, the transesterification is the key and the foremost important step to produce the cleaner and environmentally safe fuel from vegetable and waste oils [10]. Biodiesel is produced by transesterification of oils with short-chain alcohols or by the esterification of fatty acids. These processes includes base-catalyzed transesterification of the oil with methanol,- direct acid-catalyzed esterification of the oil with methanol and conversion of the oil to fatty acids, and then to alkyl esters with acid catalysis. The transesterification reaction consists of transforming triglycerides into fatty acid alkyl esters, in the presence of an alcohol, such as methanol or ethanol, and a catalyst, such as an alkali or acid, with glycerol as a by-product. It has been investigated that the process of transesterification is mainly affected by the mode of reaction condition, molar ratio of alcohol to oil, type and amount of catalysts, reaction time, FFA content and temperature [3-4]. The state of art of biodiesel production


152

NCSTI-2012demonstrated however that, the need for process optimization is an open aspect to be addressed. Moreover, the reaction kinetic of the process is not extensively investigated for a wide range of operating condition. Therefore the main objective of this work is to investigate in detail the effect of operating condition (oil to methanol ratio, catalytic ratio, and number of stages and feed location of the distillation column, temperature and reaction time on the overall performance of the process. Aspen Plus was used to model and simulate the optimization process.

Finally, the overall reaction is:

Figure.1 Chemistry of TAG (triglyceride) transesterification

2.2 Process simulation

2. Methodology In this study, the Radfrac and NRTL electrolyte(polar) models in Aspen plus, commercial program for process simulation, are applied for the methyl ester production with an alkaline catalyst. The process diagram for this process is shown in Fig.2. The major steps involved in simulation using ASPEN PLUS were:

2.1Modeling

Oils and fats are complex lipids derived respectively from vegetable and animal sources. Their compositions are primarily based on triacylglycerols (TAG), which molecules consist of a glycerol backbone attached by ester bonds to three long-chain carboxylic acids (fatty acids) [15]. Transesterification of TAG is a process of three consecutive and reversible acid- or basic-catalyzed reactions. Diacylglycerols (DAG) and monoacylglycerols (MAG) are intermediates. The stoichiometry of the overall reaction requires a molar ratio of 1:3 (TAG: alcohol) to give 3 mol of ester and 1 mol of glycerol. The reaction kinetics involves stepwise conversions of TAG to DAG to MAG to glycerol (GL) (Figure 1).

Developing the graphical simulation process flow sheet.

Specifying the components involved in the process.

Selection of thermodynamic model.

Specifying the operating condition (flow rat, temperature, pressure, composition etc.)

As the aim was to model and simulate the biodiesel production from comparatively

Figure.2 Flow sheet for the modeled production of bodies

MIXER-1

MIXER-2

PUMP-1

REACTOR1

SEP1

MEOH

NAOH MIX1

MIX2MIX3

MEOHREC

PUNP2

REACOUT2PUMP3

HX2

WASHCOL

EST2

AQU1

H2O

EST1

HX1

OIL-1REACOUT1

REACOUT

BDCOOL

WASHTOP

MEOHWATR

FAM E

OILREC

Figure .2 Flow sheet for the modeled production of biodiesel


153

NCSTI-2012

high free fatty acid feedstock, a moderate feedstock with14% free fatty acid (FFA) was considered. In process simulation Triolein (C57H104O6) was used to represent the triglyceride for biodiesel production. And the major free fatty acid found in vegetable oil and animal fat is oleic acid (C18H34O2), which was chosen to represent the free fatty acid (FFA) in this simulation. So, the expected biodiesel is methyl oleate (C19H36O2). To simplify the process flow sheet, one Stoichiometric reactor model has been used for both reactions (transesterification and esterification) rather than using two reactors in series. Stoichiometric reactor model is used in ASPEN PLUS simulation when the detail kinetic information of a chemical reaction is not available. Different studies have reported 95-97% conversion for base catalyzed transesterification [11, 12]. Acid catalyzed esterification as a pretreatment process for converting FFA to ester. This process can convert 98-100% of FFA to ester [13, 14]. The current study conversion is set to be 95% and quality of 99.4% after purification. In this case, Rad-Frac subroutine has been chosen, as it provides much more rigorous calculations compared to other subroutines. The selected operating temperature and pressure of the reactor was 65oC and 1atm respectively. To increase the inlet feed temperature in distillation column a pre-heater is used between the reactor and distillation column. The top product (SEP1)of distillation column (MEOHREC) represents the recovered excess methanol from the product. This excess methanol (99.9%) is recycled for reuse. Simulation of mass transfer equipment required the selection of some thermodynamic model and since the selected feedstock was polar compound, NRTL and UNIQUAC were the preferred models. The column bottom stream (EST1), mainly containing biodiesel and glycerol, is charged to a washing column (WASHCOl), where water is used to wash biodiesel, providing separation of methanol, soap, glycerol and catalyst from FAME. The top stream leaving the washing column (WASHTOP), mainly containing methyl esters and unconverted oil, is then feed to a vacuum distillation column (BDCOOL) in order to

separate FAME from water and methanol. FAME is obtained in the bottom stream of the column overhead condenser, with a mass flow rate of 875.465kg/h and quality of 99.4%.

3. Result and discussion

3.1 Yield of biodiesel

In the production of biodiesel the conversion is mainly affected by different parameters such as methanol to oil ratio, residence time, reaction temperature and pressure, catalyst ratio and catalyst type. The pretreated oil stream (OIL-1) with flow rate of 1050kg/h is sent to the transesterification reactor (Reacter-1), where a 5:1 to 10:1 molar ratio of methanol to oil is used with 1% (w/w) of sodium hydroxide to perform

Figure.3 effect of catalyst biodisel mole fraction

the reaction. A fresh methanol stream flow rate of 121.41 kg/h is mixed with anhydrous sodium hydroxide as catalyst and fed to the reactor. Transesterification takes place at 65oc and 1atm and reaches a 95% conversion of oil to FAME.

As it can be seen in the figure (3), the reaction is more vigorous at a low amount of catalyst mass. This could be due to the formation of OH ion at high catalyst mass that inhance the reaction of soap formation. Increasing the catalyst beyond the critical point (50Kg/hr), the side reaction such as soap formation would be favored by depressing the biodiesel reaction so that the amount of biodiesel was also decreased.


154

NCSTI-2012

3.2 Separation of biodiesel

Biodiesel is separated from glycerol, catalyst and residual methanol by washing using hot water in the distillation column and the result is presented in fig.3. In the ternary mixture, pure components are set in vertices. All the residue curves have the same vertices as origin and terminus which are the lowest and the highest boiling point respectively. We adopt here the convention that the direction of a trajectory points-out from lower to higher temperatures. Residue curves do never intersect each other. Residue curves form the same distillation region. This demonstrates that biodiesel separations do not form azoetrope and the separation is easily performed using ordinary distillation column.

Figure.4 Residue Curve for Methanol/Water/FAME

3.3 Effect of number of stage on biodiesel separation

As shown from fig.5, the temperature profile indicates that the heat supply only increase between stage number 2-3 and 5-6 of the separation (SEP1) column and 99.9% of methanol recovered for reused as raw material. The liquid composition of the separation column also indicates that methanol is more volatile and its composition dramatically decreased as the number of stage increase and since methanol is separated, the composition of biodiesel increased up to 54.3% as portrayed in fig.6. Most of the glycerol, catalyst and the residual

methanol separated from the biodiesel washing column that works adiabatically.

Figure.5. Temperature profile of the methanol separator stage

Figure.6 Liquid composition profile of methanol separator stage

3.4 Effect of number of stage on biodiesel purification

Methanol and water have high separation factor and the residuals are easily separated as the heat supply increased in the consecutive stage of the distillation column (fig.7.). Most of the non-volatile parts such as glycerol and catalyst were separated by water washing column. Washing the biodiesel is necessary to remove excess catalyst, glycerol, and a glycerides not fully reacted that would be detrimental factor for engine performance. The mixture of fatty acids methyl esters (FAME) obtained from the transesterification reaction must be purified in


155

NCSTI-2012

order to meet with established quality standards for biodiesel. To obtained high quality of biodiesel (99.4%), only four stages were enough in the distillation column as shown in fig.8.

Figure.7 Biodiesel purification separation factor of the constituents

Figure.8 Vapor composition of the biodiesel purification distillation column

3.5 Effect of temperature on viscosity

The viscosity or inner friction of diesel fuel generally increases with the density. It must be prevented from falling below the set minimum value to ensure that there is sufficient lubrication between the sliding parts of the fuel injection system. Viscosity is the most important property of biodiesel since it affects the operation of the fuel injection equipment, particularly at low temperatures when the increase in viscosity affects the fluidity of the fuel. Biodiesel has a viscosity close to that of diesel fuels [16]. The higher viscosity range of biodiesel helps to reduce barrel/plunger leakage and increase

injector efficiency in engines. This is especially important on engines. For some engines, it may be advantageous to specify a minimum viscosity because of power loss due to injection pump and injector leakage, and also because sulphur emissions. Maximum allowable viscosity, on the other hand, is limited by considerations involved in engine design and size, and the characteristics of the injection system [16]. The result obtained from the effect of temperature is displayed in fig.9. As the temperature increased, the viscosity dramatically decreased and the optimum kinematic viscosity of biodiesel in this study at 40oc is 4.5mm2/s (fig.9). The ASTM standard states that the viscosity of biodiesel must be between 1.9-6 mm2/s. It was also proved that the viscosity of biodiesel is very much less than the viscosity of the oil so that the engine performance is not affected by the biodiesel viscosity.

Figure.9 Viscosity of biodiesel with temperature.

3.6 Effect of temperature on density

Density is another essential parameter for the quality of biodiesel. As the density increases, the energy content increases per unit volume. Given an unchanging injected quantity of fuel, the energy supplied to the engine increases with the density, which increases engine performance. However, the exhaust emissions and, especially, the particles increase under a full load due to the richer mixture. On the other hand, the volumetric fuel consumption increases as density decrease [16]. The density of biodiesel


156

NCSTI-2012for European Standard EN ISO 3675 is between 860-900 kg/m3. The optimum density of biodiesel for this study at 40oc is 860 kg/m3 as shown in fig.10.

Figure.10 Effect of Temperature on Density

4. Conclusion

This work focused on the investigation of sustainable energy production from the low cost feed stocks which are abundantly obtained from leather and oil industrial wastes. The production flow sheet was synthesized. The performances of this flow sheet, along with the key operating criteria, were identified by conducting a computer-aided simulation using ASPEN Plus. In view of the simulation result; we found that the conversion of the stoichiometry reactor was 95% of the oil to biodiesel, recovery of methanol was 99.9% with reflux ratio of 0.5 since the boiling point of methanol is low compared to biodiesel and glycerol. And also for biodiesel purification column, only 4 stages were enough to obtain high quality of biodiesel (99.4%). It was also proved that the viscosity of biodiesel is very much less than the viscosity of the oil so that the engine performance is not affected by the biodiesel viscosity since it qualified the ASTM standards.

Reference

1. Ratledge, Boulton CA. Fats and oils. In: Moo Young M., Blanch HN, Dream S, Wang DIC, editors, Comparative Biotechnology in industry, agriculture and medicine, 1985, vol. 3. New York: Pergamon Press;, p. 983– 1003.

2. Lee I, Johnson LA, Hammond EG. Use of branched-chain esters to reduce the

crystallization temperature of Biodiesel. 1995, JAOCS; 72:1155–60.

3. Foidal, N.., Foidal, G., Sanchez, M., Mittelbach, M., Hackel, S.,; Jatropha Curcal L. as a sources for the production of biofuel in Nicaragua. 1996, Bioresor. Technol., 58, 77

4. Lee, K.-T., Foglia, T.A., Chang, K.-S.,: Production of alkyl esters as biodiesel fuel from fractionated lard and restaurant grease, .2002, J. Am. Oil Chem. Soc 79. 191-195

5. Nelson, R.G., Schrock, M.D., Energetic and economic feasibility associated with the production, processing, and conversion of beef tallow to a substitute diesel fuel; 2006, Biomass energy 30, 584591.

6. Kulkarni, M.G., Dalai, A.K., Waste cooking oil- an economic sources for biodiesel: a review, 2006, Ind. Eng. Chem. Res. 45, 2901-2913.

7. EthiopianTanners Association Members Brief Company Profile

8. Ravindran B, Sekaran G. Bacterial composting of animal fleshinggeneratedfrom tannery industries. Waste Manag 2010;30:2622 30.

9. Isler A, Sundu S, Tuter M, Karaosmanoglu F. Transesterification reaction of thefat originated from solid waste of the leather industry. Waste Manag2010;30:2631–5.

10. Meher L. C., Vidya Sagar D. and Naik S. N.,"Technical aspects of biodiesel production by transesterification-areview"

11. Y Zhang DDMMK M A Dube, Biodiesel production from waste cooking oil: 1. Process design and technological as sessment, Bioresource Technology, 2003. 89:pp. 1–16

12. A HWest NE D Posarac, Assessment of four biodiesel production processes using HYSYS.Plant, Bioresource Technology, 2008. 99:pp. 6587–6601

13. Official website of The Northeast Regional Biomass Program,(NRBP): http://www.nrbp.org, [last accessed on January 3, 2010]

14. AS Ramadhas CM S Jayaraj, Biodiesel production from high FFA rubber seed oil, Fuel, 2005. 84:pp. 335–340

15. Formo, M. W. (1954). Ester reactions of fatty materials. Journal of the American Oil Chemists’Society,Vol.31,No.11, (November 1954), pp. 548-559, ISSN 0003-021X


157

NCSTI-2012


Meron Mulatu Mengistu

Chemical Engineering Department, Bahir Dar University , Bahir Dar, P.O.Box 2133, Ethiopia, [email protected]

Abstract — The energy consumption rate of non-OECD countries rises about 2.3 percent per year as compared to the energy consumption rate of OECD countries which is 0.6 percent. If developing countries use energy efficient technology and integrate renewable energy systems in the new building their carbon dioxide emission rate reduces by 25 to 44 percent. However, even now, renewable energy integrated buildings are hardly considered while constructing them. This thesis work focuses on the study of solar cooling system options for residential house in Bahir Dar city. To meet the demand of housing in the city, different type of apartments and villa houses are under construction. Case study was made focusing on two types of residential houses (condominium apartment and Impact Real-estate Villa house) to determine the cooling load and to select cooling system. Simulation results of IDA ICE software show that the average operative temperatures and cooling loads for condominium apartment and Real-estate Vila are 31.8oC and 30.7oC, 5.53 kW and 5.73 kW respectively. Most of the residences are not satisfied at this operating temperature. There are different types of solar cooling systems. Solar Sorption cooling systems are commonly used which can also be classified in to absorption, adsorption and desiccant cooling systems. Solar adsorption cooling systems are easy to manufacture locally as compared to solar absorption cooling systems. They do not have moving parts. Some of the working medium pairs used in adsorption cooling system are: Activated Carbon/Ammonia, Silica gel/ water, Zeolite/water. Adsorption chillier with Silica gel/ water as a working pair was selected since it can operate at regeneration/desorption temperature as low as 45oCcoming from flat plate collectors. At 75oC regeneration temperature, the system delivers 9oC chilled water temperature. From cooling load simulation result direct solar irradiation is the highest source of cooling load for both houses. This gives an opportunity for passive solar cooling technology.

Keywords- Adsorption, Cooling Load, Condominium, IDA ICE, Impact real-estate

I. INTRODUCTION Access to affordable energy service is fundamental to

human activities, development, and economic growth. Development goals in the areas of water, health, agriculture, and biodiversity often cannot be met without energy inputs and the policies adopt in these sectors similarly impact the availability and reliability of energy services. This is the reason behind the large increment in energy consumption in all over the world and consequently the change in climate [1].

The current energy consumption rate of OECD countries is 0.6 percent in contrast with non-OECD countries which is about 2.3 percent per year. This is due to rapid growth and increase in energy consumption in the non–OECD countries, and availability of energy efficient technologies and incentives in OECD countries. [2] Like many developing countries, Ethiopia’s energy consumption is increasing due to development and new construction. Biomass contributes, the largest portion of energy consumption, which is traditional and inefficient. Domestic sector covers 89 percent of energy consumption, which is biomass and mostly in the rural area, used for cooking and lighting. [3]

Ethiopia is close enough to the equator and gets plenty of solar power whenever the sun is shining. That is why the name “thirteen months of sunshine” is given. Even in the summer season, there is sunlight for a short period of time. Though there is a huge potential for solar energy utilization as a most promising renewable energy resource, solar energy is not harnessed to generate power in a desirable way. Bahir Dar city has a solar insolation of a maximum in April and a minimum insolation in July. [4] This paper focuses on the study of solar cooling options considering the two housing system, low cost housing system and villas constructed by the government and real estate companies respectively.

II. ENERGY IN BUILT ENVIRONMENT Due to increment of living standard and comfort,

energy demand in built environment is increasing. 25-30 percent of entire energy allied carbon dioxide emission of developed countries comes from buildings. Also in developing countries, building carbon dioxide emission increased from 11 percent to 19 percent from year 1973 to 1990. There is a reduction of carbon intensity of energy services as a result of improvements in efficiency and better technology but the increase of energy


158

NCSTI-2012consumption for more services has plagued these intensity reductions. Carbon dioxide reduction in buildings include both energy-efficiency and non-energy efficiency technologies. Fuel changes to non-carbon intensive and also renewable energies are included in non energy-efficiency technology. Energy efficient technology in buildings includes enhancements to the building shell, better management of energy consumption, and improving the efficiency of different end-uses. For example, energy-efficient windows compact florescent and improved efficiency of biomass stoves. In OECD countries there is a potential of 6-16 percent of reduction in carbon dioxide emission while non-OECD and developing countries have a larger potential of 25-44 percent when comparing energy efficient scenarios to business as usual trends. This higher reduction potential is due to construction of new buildings that gives the opportunity to build in a more efficient way, while in developed countries, energy consumption reduction is done usually by retrofitting on existing buildings. Growth and implementation of new technologies that make energy production and consumption more efficient in non-OECD countries would bring a significant reduction in carbon dioxide emission. [5]

The majority of the population in Ethiopia use traditional and inefficient biomass (only 5 to 10 percent of efficiency) for cooking, lighting and in cold season for heating. This results in deforestation and indoor air pollution. Most residential houses are not constructed to take in account the peoples comfort at the design stage. Thermal comfort is done by using external shading systems that are natural like planting trees or artificial shading systems, such as infiltration of air through doors and windows, adjusting their wearing style etc. [6]

Bahir Dar city is found at altitude of 1800 meter and the heating/cooling load is not as huge as cities in the north such as Stockholm. The hot season’s duration is longer than the cold season, and working on technologies for cooling can be much feasible than heating technologies. Bahir Dar is one of the fast growing cities in the country. To meet the demand of housing in the city, different type of apartments and villa houses are under construction. However, renewable energy integrated building are hardly considered in the construction of theses housings. This can lead to an energy intensive housing system in the near future. Solar technologies can be installed even after the construction of buildings. In the design of solar built environment, several scenarios have to be considered. The resource availability is the first issue to be addressed. A solar energy resource decreases when going from equator to north or south. Ethiopia, which is found near the equator, has a potential to utilize huge available solar energy in buildings. Availability of solar technologies with affordable price is the other main issue to employ solar built buildings in low income countries like Ethiopia. The type of solar technology depends also the required heat/ cooling load in a building.

III. METHODOLOGY So as to propose an affordable solar cooling

technology for Bahir Dar city, it is crucial to study different solar cooling technology, potential of solar irradiation and cooling load demand of the houses. Depending on people income level and the type of building house, two case studies will be considered. Low income level people are most likely to live in low cost housing (Condominium) while people with high income level live in villas (such as real-estate), which are mostly found in the suburb of the city. For these reasons solar cooling for condominium apartments and Impact Real-estate are the two case studies considered in this study. A study of house hold energy demand and internal heat gain of the two cases mentioned will be done by interviews to gather information on household energy demand and material of construction for both cases by preparing questionnaires for household responsible and for construction managers, using IDA ICE software to determine the cooling loads for the two house cases from the data’s of the interviews. Depending on the simulated load, appropriate solar cooling technology will be selected.

IV. SOLAR SORPTION REFRIGERATION The process of attracting and sustain a gaseous or

liquids is described as sorption. In sorption refrigeration the solar thermal energy is directly converted to cooling effect by physical or chemical attraction between a pair of substances. The pair of substances are sorbate and sorbent. Sorbate is a substance having lower boiling temperature and plays the role of the refrigerant. The sorbent have the ability to attract and hold other gases or liquids. [7]

Sorption refrigeration system can be classified as closed and open sorption system. Closed sorption system includes absorption and adsorption refrigeration. Desiccant refrigeration is an open sorption system. [8] Absorption is a type of sorption process in which the sorbent absorbs a refrigerant molecule internally and changes its property (physical or chemical) in the process. Adsorption involves a solid sorbent and does not involve phase change during the process; rather the sorbent only increases by weight due to the adsorbed sorbate. The major difference between absorption and adsorption is the nature of sorbent and the duration of the process which is longer for adsorption. [9] Desiccants are sorbents having a special attraction to water. Here the sorbent or desiccant, absorbs moisture from humid air without changing the physical characteristic of the desiccant. [8]

A. Absorption

Absorption is the most common types of solar refrigeration system. It is a reversible process. For the same capacity it has smaller physical dimensions than adsorption due to the high heat transfer coefficient and


159

NCSTI-2012fluid property of absorbent. The phenomena in absorption refrigeration include fluid phases having strong affinity. [10]-[7]

Fig. 1. Schematic diagram of closed solar sorption refrigeration [7]

As shown in Fig. 1 the main components of absorption chillers are generator, absorber, condenser, evaporator, pump and expansion valves. The generator is a component in which desorption (regeneration) takes place using heat and the absorber is a component where sorption takes places. Cooling is formed in the evaporator where the refrigerant is vaporised by removing a cooling load of Qe. The vaporised refrigerant goes to absorber and absorbed by the sorbent. Here the dilute solution of sorbent sorbate is formed and pumped to generator. In order to increase the efficiency of the process, cooling is done. The heat Qg produced by the solar collector regenerates the sorbent that absorbs the refrigerant (sorbate) fluid in the absorber. The concentrated vaporised refrigerant goes to condenser where heat (Qc) is rejected, whereas the sorbent goes to absorber to absorb the coming refrigerant from the evaporator and rejects a sorption heat Qa to the ambient. After the refrigerant condenses it will go to evaporator for another cycle. [9]

A fundamental requirement of refrigerant/absorbent mixture is the margins of miscibility that must be at the range of operation temperature of the cycle. In addition they must be non-toxic, chemically stable and non explosive. There are many types of working fluids, but water/NH3 and LiBr/water are most commonly used. In the first case NH3 is the refrigerant and in the second one water vapour is the refrigerant. The water /NH3 system needs a rectifying column to assure no water vapour enters in the evaporator, which will result in freezing in the evaporator. In addition the system needs a high

generators temperature. Generally NH3/water system is usually used for refrigeration in industrial applications, while LiBr/water systems are common in air conditioning applications [9].

A single effect water/NH3 solution with a heat source regeneration temperature of 80 to 120oC can give a COP of 0.3 to 0.7. A LiBr/water absorption chiller usually works with a heat source temperature of above 88oC and results a COP of 0.6. Even though these systems can operate with flat plate collectors usually to obtain a higher COP they are equipped with an evacuated type collector especially for LiBr/water working fluid. For a double effect LiBr/water absorption chillers a regeneration temperature of 150oC is required which is obtained by concentrated type collector. [11]

Most commercial available absorption chillers have a capacity of 100 kW and a small capacity less than 50 kW is very rare. [12] But recently small capacity even less than 10 kW are becoming available in market. SolarNext a German company produces a 10 kW single effect NH3/water absorption chillier and it requires a supply hot water temperature of 68 to 75oC. The produced chilled water temperature and COP are 19 to 6oC and 0.64 respectively. Another small scale market available absorption chillier is developed by Rotartica Company in Spain. It is a single effect LiBr/water having a cooling capacity of 4.5 kW. For a regeneration hot water supply of 90oC its COP will be 0.7. By varying the supply temperature different COP and chilled water can be obtained. [11]

B. Adsorption Chillers

Adsorption process results from an interaction between a solid (adsorbent) and fluid (refrigerants). Depending on the type of adsorbent and refrigerant reaction, the reaction can be categorised as physical and chemical adsorption. Physical adsorbents are highly porous and have high surface to volume ratio that can selectively catch and hold refrigerants through the type Van der Waals force. [9] Common working physical adsorbent refrigerant pairs are activated carbon-methanol or ammonia and silica gel-water. Strong chemical bond between the refrigerant and the adsorbent is characteristic of chemical adsorption. This strong bond makes the process complex as it needs more energy in order to regenerate and reverse the process than the physical adsorption process. The most common chemical adsorbent used in solar cooling system is calcium chloride (CaCl2) with ammonia and water as a refrigerant. [7] In solar powered adsorption activated carbon, silica gel and zeolite are common types of adsorbents while water, methanol, ethanol and ammonia are common refrigerants used. [9] Silica gel-water is the best combination adsorption refrigeration due to its low regenerating temperature. It can operate with a supply hot water temperature of 45 to 90oC. This temperature can be achieved by flat plate solar collector. In addition it will enable the chillier to work more than eight hour in a day. At lower regeneration temperature a COP of 0.3 can be


160

NCSTI-2012

obtained. [11] Zeolite-water pair needs a regenerating temperature of above 200oC and activated carbon-ammonia pair needs around 150oC. And these temperatures cannot be obtained by flat plate or evacuated type collector. [13]

Fig. 2. Working principle of adsorption chillers [13]

Components of adsorption chillers are similar with absorption chillers. The adsorption process can take place with a single adsorbent bed called fixed bed or multiple adsorbent beds. If fixed bed is used the process will operate without any moving parts which results in silent and highly reliable. However the process will be intermittent, as a result COP of the system will decreases; therefore multiple adsorbent beds are required for continuous process and to increase COP of the system. [14] For example in double adsorbent bed Fig. 2, the refrigerant in the evaporator create a cooling effect by vaporisation. Then it goes to bed 2 and adsorbed by the sorbent bed while the refrigerant in bed 1 is regenerated by using hot water. In this case bed one act as an adsorbent bed while bed 2 is a regenerator. For the second cycle bed 2 will act as a generator while bed 1 is an absorber, the adsorbent bed is changed between regeneration stage and adsorption stage so as to form a pseudo cyclic system. [13]

Unlike absorption process crystallization and corrosion are not a problem in adsorption process. It is much flexible in regeneration temperature and can be best applicable for part load process. [14] The principal limitations of the adsorption system are weak heat and mass transfer character of the adsorbent beds. The adsorbents like zeolites, activated carbon and silica gel have low thermal and poor porosity characteristics, as a result its component are bulky collector/generator/absorber and expensive. And thus, its excessive heating capacity leading to rather low thermal COP. [9] Optimum adsorbent bed structure lies between a high porosity required for fast vapour diffusion and the high density required for good thermal conductivity. To improve the mass and heat transfer of the adsorbent bed so as to decrease the size and cost of the system, addition of packing density of adsorbent, using selective coating material, using heat transfer fins, employment of consolidated adsorbent and selecting a suitable working environment are the main technologies. [14]

Most market available adsorption systems working fluids are water /silica gel with capacity that ranges between 50 kW and 500 kW. [13] For residential application with small cooling capacity adsorption chillers are very limited in the market. Sortech, a German company developed a silica gel/water adsorption chillier with a capacity less than 10 kW. At a regeneration temperature of 75 to 67oC, it produces 5.5 kW capacities with 18 to 15oC chilled water production. [11]

C. Solar Desiccant Cooling

Open sorption system is usually called desiccant cooling where the desiccant or sorbent is used to dehumidify air. The conventional air conditioning systems usually don’t control the humidity but only the temperature of conditioned space. If they do, they control it indirectly through temperature. But, desiccant cooling systems directly achieve the dehumidifying process through the use of desiccant materials [15] and this make a desiccant cooling system a complete HVAC system. Desiccant materials have the ability to absorb water. Silica gel, activated alumina, zeolite, LiCl and LiBr are examples of desiccant or absorbing sorbents. [7]

Basically there are two types of desiccant systems; liquid and solid depending on the phase of sorbent used. In liquid desiccant cooling technology both the liquid and the air flows between a dehumidifier and a regenerator. The working fluid for liquid desiccant is LiCl/water which works at atmospheric pressure, which will reduce the cost of sealing when compared to absorption chillers. It can be used with a low temperature of 40oC. However this technology has a problem of corrosion which is formed by inorganic salt contained by the container. To solve this problem a polymer type of equipments is used. In addition due to its liquid working fluid its systems are big and also it is not available in the market but researches and developments are ongoing. [11]

A typical solid desiccant cooling system is shown in Fig. 3. Unlike the liquid, it is more compact. The return air from the conditioned space passes through an evaporative cooler and becomes cold and humidified (5 →6). Then it passes through a sensible heat exchange and becomes warmer.

Fig. 3. Schematic diagram of solar solid desiccant cooling [16]


161

NCSTI-2012This warm and humid air passes through the solar

heating coils and becomes hot and humid air (7 →8), which heats and regenerate the desiccant wheel while passing through it and exhaust to the ambient (8 →9). Fresh air then passes through the regenerated desiccant wheel where water in the air is absorbed by and becomes dry and hot (1 →2). This dehumidified hot air enters into the sensible heat exchanger and cools down (2 →3) by preheating the cold air (6 →7). After this it can be directly supplied to the conditioned space or cools down by the evaporative cooler if necessary (3 →4). Evaporative cooler is used to adjust the humidity and temperature if it is necessary. The sensible heat in the air will evaporate the water in the evaporative cooler and results in a lower temperature and higher humidity content of supply air to the conditioned room. [7] This type of desiccant system is applicable for temperate zones where the dehumidification process is not as high as compared to Mediterranean countries. [12]

LiCl and silica gel are widely used in solid desiccant wheel with regeneration temperature of 60 to 120oC and 80 to 150oC respectively. Compared to the other types of desiccant LiCl has a high moisture removal capacity. Researches show that using a composite type desiccant like silica gel Haloids it is possible to increase a moisture removal of 20 to 30 percent compared to silica gel alone. [11]

Desiccant cooling is a complete HVAC system and performs efficiently in humid climate than all sorption systems. It is easy for maintenance and reliable but when compared to other sorption systems it is big in size and complex for residential installation. It is very expensive and problems related to installation of the complex system components while connecting to the solar heating system, results in problem of successful integration with buildings. [15]

Commercial availability of this technology is very limited especially in small size. In China 11 kW cooling power desiccant wheel is installed for demonstration. Experiment shows that when the outdoor condition is 35oC and 23.2 g/Kg of relative humidity, it supplies air at temperature of 25oC and 17.7 g/kg. [11]

V. ESTIMATION OF SOLAR RADIATION IN BAHIR DAR CITY

Bahir Dar city is found at latitude and longitude of 11.15 E and 37.77N, which is in north western part of Ethiopia at 578 km from Addis Ababa the capital city of Ethiopia. The city is place on the northern shore of Lake Tana (the largest lake in Ethiopia and the source of the Blue Nile River (Tis Abay)). It is the capital city of the Amhara region.

Majority of the year season is summer. June, July and August months are winter seasons having high rainfall and relative humidity (up to 81 percent) in relative to other months as shown in Fig. 4. As a result the ambient temperature drops in these months. The rest of the year is summer season (relatively hot and dry), in which there is

low relative humidity (minimum of 46 percent). March, April and May are the hottest months of all.

Fig. 4. Monthly average temperature and relative humidity

In order to select solar cooling system for the city, the amount of solar irradiation must be known. National Metrological Agency Service (NMSA) is responsible for supplying metrological data of weather in the country. But regularly recorded data of solar irradiation is not available for all regions, only for Addis Ababa city. For the city available metrological data found are sunshine hour, temperatures (monthly minimum and maximum), relative humidity and wind speed. Therefore global solar irradiation is estimated using an empirical formula called Linear (1) and Quadratic Angestrom (2) method. These empirical formulas that are well known correlate sunshine hour with global solar radiation. Global solar radiation is the total solar radiation that is measured on a horizontal surface, and it can also be measured using pyranometer instrument. [17]

= (1)

(2)

Where; H= monthly average of the daily global radiation on a horizontal surface, W/m2

Ho= average value of daily extraterrestrial solar radiation on a horizontal surface for each month, W/m2

n= monthly average of daily bright sunshine hours

Nd= average of the maximum possible daily hours of sunshine

a, b, ao, a1, a2 are dimensionless empirical coefficients


162

NCSTI-2012The coefficients a, b, a0, a1, and a2 are 0.385, 0.348,

0.0317, 1.54 and -0.936 for Bahir Dar city respectively and have different value for different cities in Ethiopia. [17] Average values using (1) and (2) and the given parameters which are latitude, daily sunshine hours and coefficients are shown in table 1.

H Linear,

W/m2

H Quadratic, W/m2

HNASA W/m2

Average 252.25 250.12 249.81

Table 1: Average solar irradiation calculated results

Fig. 5. Comparison of solar radiation results

Solar radiation results from linear regression formula align with solar radiation from NASA Fig. 5. The result from quadratic regression formula deviates by increasing in March, April and June months and decreasing in July and August months. Therefore linear regression can be taken as a best solar radiation estimation method for Bahir Dar city. From Fig. 5 the city has minimum irradiation in July and August and a maximum in March, April and May. It has an average monthly daily solar irradiation of 250 W/m2. Compared with an average value of Ethiopia which is 231.48W/m2 it gets a higher solar irradiation [18].

VI. IDA ICE SOFTWARE AND BUILDING DESCRIPTION

Cooling load is the amount of heat that must be removed from a building in order to meet the design conditions of temperature and humidity for comfortable zone [19]. Cooling load estimation of the buildings is simulated using IDA ICE software. It can model a building with one or more rooms with existing heating and ventilation systems. [20] A room to be conditioned is called zone. It is a space to be conditioned and controlled by with one thermostat. The thermostat is used to sense the temperature of the room and should not be kept in very high or low temperature regions. This is done in order to take the average room temperature and as a result to keep it in a desirable and comfortable

condition. A zone can be one or more rooms together that have uniform heat gain. [19] Therefore one house in both cases (condominium and Impact real-estate) is taken as one zone and after simulation of the cooling load of one zone then, it is multiplied by the total number (of houses) in the building.

Impact Real-Estate

Condominium

Heat Gain (W) Heat Gain (W)

OCCUPANT 808.12 461.78

LIGHTING

Bulb 168 48

EQUIPMENTS

Television Set 21.67 21.67

Radio/Tape 4.67 1.49

Refrigeration 40 40

Stove 68.75 -

Micro-Oven 125 -

Laundry Machine

562.5 -

Desktop Computer

14.37 -

TOTAL INTERNAL HEAT GAIN

1813.08 572.94

Table 2: Internal Heat Gain of Impact Real-Estate and Condominium

Inputs for load simulation are climate data, property of material of construction, AutoCAD drawing of floor plan, orientation, and internal heat gain that include occupant and different appliance. Internal heat gain for the two house cases is shown in table 2. Climate data for the software input is generated from METEONORM software by using the latitude, longitude, elevation and time zone as input parameter. It results in a one year hourly base data of air temperature (oC), relative humidity (%), wind speed (m/s) and direction, normal and diffused solar irradiation (W/m2) [21].

Building envelop separate the indoor condition and the outdoor air. Heat transfer or energy exchange between the indoor and outdoor environment is dependent on the type and property of building envelope. To collect type and thickness of material of construction of these envelopes an interview is made with a construction manager of Impact real-estate. Table3 shows U_ values for different building envelope that is dependent on thickness. Material of construction for condominium


163

NCSTI-2012apartment and impact real-estate are more of similar except finishing parts such as wall paints and tile finishing floor that are only for Impact real-estate.

Thickness [mm] U_ value

External wall 312.5 0.6629

Internal wall 146 0.6187

Window 1 pane glazing 5.8

Floor 125 1.025

Roof 708 0.172

Table3: Characteristics of Building Envelop [22]

A. Condominium Building Description

In Ethiopia there is a scarcity of living houses, for this reason the Ethiopian Federal Democratic government builds a low cost houses called condominium in many cities of the country. In Bahir Dar city there are two sites, “Kidanmehret” and “Abay Mado” site. Depending on the number of rooms, the houses in one building are classified in to three: having two bed room, one bed room and studio. “Two bed rooms” from the name itself it has two bed rooms, a salon, kitchen and bath room. “One bed room” has the same number of rooms of utility rooms as “two bed room” but with only one bed room. The “studio” has only one living and bath room.

Fig. 6. 3-D Diagram of condominium from IDA ICE software

For this project the Abay Mado site is selected as a model case. It is oriented towards south. In this site there are three buildings, each with ground plus four. In one floor there are five numbers of zones (houses) with a total of twenty five zones in one building. From the three types “one bed room” house is taken as a model zone for the simulation, Fig. 6. The model zone selected has a floor area of 27 m2 and a height of 2.6 m. It has a total of four windows and one entrance door. The average number of occupant is four.

B. Impact Real-Estate Building Description

Impact real-estate is a private limited company which builds luxury homes, appendix 14.6. It is the first and at this times the only company in Bahir Dar. For this project

a zone having 500 m2 surface areas is selected as a model zone with floor plan area of 162 m2and 2.6 m height. As shown in Fig. 7 it has eight numbers of rooms (three bed rooms, two bath rooms, one guest bath room, one modern Kitchen, and one salon) and ten windows. There are a total of 25 such zones and are oriented to East wards. Until this project is done the houses are not giving service for the customer and they are under construction. The number of occupant is assumed to be seven.

Fig. 7. 3-D Diagram of impact real-Estate from IDA ICE

VII. RESULTS OF IDA ICE The simulation result includes heat removed (cooling

load) in watt, dry bulb and operating temperature in degree centigrade, and percentage of dissatisfaction (PPD) for the single and total zone (table 4). Cooling design results are simulated at the time of maximum cooling load, which is the maximum heat that must be removed from the conditioned room in order to satisfy the occupants. It is the capacity of cooling equipment device. Heat gain is the rate at which energy is generated within or transfer to a space. But only when the indoor air receives the energy by convection does this energy become cooling load. The radiant energy gain from different sources does not directly heat the indoor air but first it is absorbed by interior envelops and masses.

From simulation result cooling equipment capacity should have a maximum of 5.53 kW for condominium (for 25 zones) and 5.73 kW (for 2 zones) for impact real estate (table 4). Cooling load of impact real-estate is higher than condominium, mainly due to its high internal heat gain (table 2) and larger surface area of windows. However the design main temperatures (table 5) of condominium are higher than impact real-estate, which result in high PPD (Predicted Percentage of Dissatisfied) value for condominium (98.69 percent) than impact real-estate (93.96 percent). This is mainly due to heat gain per unit area of condominium is higher than Impact real-estate, even though there are relatively high internal heat gain sources in Impact real-estate. For the same heat gain if large and small buildings are compared, keeping the other factors similar, heat gain per unit area of small building will be higher than the larger building. This results in a higher operative temperature for the smaller building.


164

NCSTI-2012

Zone multiplier

cooling load, W

Mean air Temperature oC

Operating temperature oC

PPD, %

1 221 32.1 31.82 98.69

Condominium

25 5525

1 2863 31.21 30.7 93.96

Impact real-estate

2 5726

Table 4: Cooling design results of IDA ICE

A. Main Temperatures

Main temperatures in IDA ICE software are indoor operative and mean air temperature .The average value for both houses can be seen in table 5, monthly values can be seen in Fig. 8. Condominium gets its minimum main temperature in June, July and August and maximum in March and April. Impact Real-estate attains its minimum main temperature in January and August, and maximum in March, April and May. Average main temperature of condominium (28.7oC) is higher than impact real-estate (27oC). This is mainly due to heat gain per unit area of condominium is higher than Impact real-estate.

Table 5: Average Mean air and operative temperature of Condominium and impact real-estate

When the average main temperature results are compared to the comfort condition [23], almost in all months both houses are not in a comfortable condition so that cooling system is required. Only in November, December, January months the operative temperatures of Impact real-state houses are 26, 25.2 and 24.8oC respectively and shows that occupants are in a comfortable zone and no cooling or heating system is required.

Fig. 8. Operative temperature of condominium and impact real-estate

B. Total Heat Balance

Average values of heat balance can be seen in table 6 and Fig. 9. It shows that heat gain through envelop, window and thermal bridges reduce building heat gain and become heat loss whereas direct solar, equipment, occupant and lighting add cooling load. The average total heat gain of condominium is 174 W while for impact real-estate is 266.6 W per zone. These result a total heat gain for the whole building approximately equal to 4350 W (25 zones) and 533 W (2 zones) respectively. For both houses direct solar irradiation is the maximum heat gain with an average value of 434 W for condominium and 1471 W for impact real-estate. Heat loss through walls and floors shares the maximum heat loss from both houses.

Condominium

Impact real-estate

thermal bridges

-37 -137.2

walls and floors

-445.6 -1897.3

direct solar 433.8 1471.5

equipment 76.8 837

windows openings

-191 -689.3

lighting, 46.3 162.1

occupants 292.2 538.2

air flows -1.5 -18.4

Total heat balance,

174 266.6

Table 6: total heat balance of condominium and impact real-estate, in W.

Condominium Impact real-estate

Mean Air Temperature, oC

Operative Temperature, oC

Mean Air Temperature, oC

Operative

Temperature, oC

28.7 28.5 27.2 27


165

NCSTI-2012

Fig. 9. Total heat balance of Condominium and Impact real-estate in W.

Condominium attains minimum heat gain In July and August (121 W and 117 W respectively) while Impact real-estate minimum heat gain is in January and August (193 W and 195 W respectively) (table 7). Both houses attain maximum heat gain in April 397.7 W for impact real-estate and 216.7 W for condominium. The reason is the maximum heat gain sources is solar irradiation that increases in summer seasons and decreases in winter.

month Impact real-

estate Condominium,

January 192.9 176.7

February 244.6 192.2

March 336.9 213.3

April 397.7 216.7

May 389.4 187

June 282.6 145.2

July 218.5 121.5

August 194.8 116.8

September 241 154.4

October 267.4 185.5

November 225.1 187.8

December 209.3 191.9

mean 266.6 174

Table 7: monthly average total heat balance in W.

VIII. CONCLUSION To select a cooling system, investigation of the

potential of resource availability for the area must be done. Since the country is located in tropical zone, Bahir

Dar gets quite enough solar irradiation, about 250W/m2. It has minimum irradiation in July and August (local winter seasons) and a maximum in March, April and May (local summer seasons).

Houses of Impact Real-estate Villa are bigger in size than the condominium houses. As a result, the total heat gain per zone of the Impact Real-state villa house is around 267 W where as this value for condominium house is 174 W. For both houses direct solar irradiation shares the highest source of heat gain while building envelope shares the highest heat loss. At an operating temperature of 30.7oC and 31.8oC, IDA ICE simulation gave design cooling loads of 5.53 kW and 5.73 kW for 25 zones of condominium and for 2 zones of Impact Real-state Villa respectively.

A single effect water/NH3 vapour absorption chiller with a regeneration temperature of 80 to 120oC has a COP value in the range of 0.3 to 0.7. A LiBr/water absorption chiller usually works with a heat source temperature beyond 88oC and has COP value of about 0.6. These systems operate using evacuated tube solar collector to have better performance. Though its COP is low, about 0.3, Silica gel/water adsorption chiller can operate at regeneration temperature as low as 45oC. This temperature can easily be achieved using a simple flat plate solar collector and also it will enable the chillier to work more than eight hour in a day. Zeolite/water pair needs a regenerating temperature of above 200oC and activated carbon/ammonia pair needs around 150oC. These temperatures cannot be achieved from flat plate or evacuated type collectors. Thus, silica gel/water adsorption chiller having a capacity of 7 kW is proposed both houses.

IX. RECOMMENDATION AND FUTURE WORK

Solar irradiation through windows and openings is the highest heat gain in the system. Hence, it is recommended if roof overhang or other direct sunlight blocking system is integrated in the building.

Orientation of condominium building is towards south which is disadvantage for hot climate like Bahir Dar. Therefore, it will be better if most of the windows are oriented toward the north.

The ambient temperature is less than the indoor operative temperature so that passive technology such as solar chimney that ventilates the room can be considered and further investigation should be taken.

Future work can be done on design of solar cooling system that considers simulation of the whole solar cooling system with the building.

In addition to solar potential resource Ethiopia also has a biomass potential so it is recommended further investigation on this resource.

REFERENCES [1] C. Flavin and M.Hull Aeck, “Energy for

Development”. Worldwatch Institute, available at http://www.worldwatch.org/system/files/ren21-1.pdf


166

NCSTI-2012[2] International Energy Outlook 2009 (IEO2009),

available at http://www.eia.doe.gov/oiaf/ieo/pdf/0484%282009%29.pdf

[3] Renewable Energy Development (DR, 2006),”Renewable Energy in Emerging and Developing Countries” available at http://www.energyrecipes.org/reports/genericData/Africa/061129%20RECIPES%20country%20info%20Ethiopia.pdf

[4] GASIMA (2007), “Bahir Dar Sunrise, sunset, dawn and dusk times, table”, available at http://www.gaisma.com/en/location/bahir-dar.html

[5] S. Wiel, N. Martin, M. Levine, L. Price, J. Sathaye, (1998), “The role of building energy efficiency in managing atmospheric Carbon Dioxide”, Environmental Science and Policy, vol. 1; 1998

[6] House Hold Energy Network (HEDON, 2007), "Ethiopia country profile”, available on http://www.hedon.info/Ethiopia , viewed on April 2010.

[7] D.S. Kim, C.A. Infante Ferreira, (2008), “Solar refrigeration options-state –of-the-art review”, International journal of refrigeration, vol.3 I, August 2007

[8] G.A. Florides, S.A. Tassou, S.A. Kalogirou, L.C. Worbel, (2002), “Review of solar and low energy cooling technologies for buildings”, Renewable & sustainable energy reviews, vol.6, June 2002

[9] Y. Fan, L. Luo, B. Souyri, (2007), “Review of Solar sorption refrigeration technologies: development and applications”, Renewable & sustainable energy reviews, vol.11, January 2006

[10] P. Srikhirin, S. Aphornratana, S. Chungpaibulpatana, (2001), “A review of absorption refrigeration technologies”, Renewable & sustainable energy reviews, vol.5, February 2001

[11] R.Z. Wang, T.S. Ge, C.J. Chen, Q.Ma, Z.Q. Xiong, (2009), “Solar Sorption Cooling system for residential applications: options and guidelines”, Refrigeration, vol. 32, February 2009

[12] Hans–Martin Henning, (2007), “Solar Assisted Air Conditioning of Buildings –an overview”, Applied Thermal Engineering, July 2007

[13] ESTIF (2006), “Solar Assisted Cooling-State of the Art”, available at http://www.estif.org/fileadmin/estif/content/policies/downloads/D23-solar-assisted-cooling.pdf viewed on May 2010

[14] B. Choudhury, P.K. Chatterjee, J.P. Sarkar, (2010), “Review paper on solar-powered air-conditioning through adsorption route”, Renewable & sustainable energy reviews, March 2010

[15] G. Panaras, E. Mathioulakis, V. Belessiotis, N. Kyriakis, (2010), “Theoretical and expermintal investigation of the performance of desiccant air-conditioning system”, Renewable energy, vol.35, November 2009

[16] K.F. Fong, T.T. Chow, C.K. Lee, Z. Lin, L.S. Chan, (2010), “Comparative study of different solar cooling systems for buildings in subtropical city”, solar energy, vol.84, December 2009

[17] F. Drake, Y. Mulugetta, (1996), “Assessment of solar and wind energy resources in Ethiopia. I. Solar Energy”, Solar Energy, May 1996

[18] G. Bekele, B. Palm, (2009), “Feasibility study for a standalone solar-wind-based hybrid energy system for application in Ethiopia”, Applied Energy, Vol.87, July 2009

[19] F. C. McQuiston, J. D. Parker, J. D. Spliter (2005), “Heating, Ventilating, and Air conditioning-

Analysis and Design”, John Wiley and Sons, Oklahoma, USA

[20] Vitalijus Pavlovas, (2004), “Demand controlled ventilation A case study for existing Swedish multifamily buildings”, Energy and Buildings, Vol.36, 2004

[21] METEOTEST, J. Remund, S. kunz, C. Schilter, S. Moller, (2010) “METEONORM Version 6.0- handbook part 1: software”, available at http://www.meteonorm.com/media/pdf/mn6_installation_en.pdf

[22] EQUA Simulation AB (2009) “IDA Indoor Climate and Energy 4.0 Manual”, available at http://www.equa.se/deliv/ICE4eng.pdf?lic=ICE40X:ED140.380.361.1.17

[23] G. Hauser, C. Kempkes, B.W. Olesen, “Computer simulation of Hydraulic Heating/cooling system with embedded pipes”, available at http://www.cibse.org/pdfs/Embedded%20Hydronic%20Pipe%20Sys.pdf


167

NCSTI-2012


Temesgen E.Nigussie1 2, Assefa M. Melesse3

1Ethiopian Institute of Water Resources, Addis Ababa University, Ethiopia,[email protected] 2School of Civil and Water Resources Engineering, Institute of Technology, Bahir Dar University, Ethiopia

3 Department of Earth and Environment, Florida International University, USA, [email protected]

Abstract— Accurate estimation of evapotranspiration (ET) is essential in water resources management and hydrological practices. The FAO-56 Penman–Monteith (FAO-56 PM) is a sole global standard method, but it requires numerous weather data for the estimation of reference ET. A new simple temperature method is developed which uses only maximum temperature data to estimate ET. Six class I weather stations data were collected from the National Meteorological Agency of Ethiopia. These stations were distributed over the study area to represent different climatic setting. This method was compared with the global standard PM method and the well-known Hargreaves (HAR) temperature method. The coefficient of determination (R2) of the new method was as high as 0.74 and 0.88, when compared with PM ETo and HAR methods, respectively. The annual average coefficients of determinations over the six stations when compared with PM and HAR methods were 0.68 and 0.84, respectively. The method is able to estimate daily ETo with an average Root Mean Square Error (RMSE) and an average Absolute Mean Error of 0.58mm and 0.46mm, respectively from the global standard PM ETo method. The method is also able to estimate annual ETo with an average error of less than 3.3% from the PM ETo estimations. The method was also tested in dry and wet seasons in the study area and found performing well in both seasons; with an average R2, RMSE, and AME of 0.63, 0.47mm, and 0.37mm in dry season and 0.59, 0.4mm, and 0.31mm in wet seasons, respectively when compared to the PM method. The average R2 of the new method with the HAR method was 0.82 and 0.81 in dry and wet seasons, respectively. The method could be used for the estimation of daily ETo where there is insufficient data. Locally calibrated coefficients of this method provide better results and is most recommended for annual estimation of ETo in the study of hydrology.

Keywords- Evapotranspiration, air temperature, Penman–Monteith method, Hargreaves method, new simple method, Ethiopia

I. INTRODUCTION

Evapotranspiration (ET) is one of the main components of the hydrologic cycle. It is the major component next to precipitation and its accurate estimation is essential in agricultural and hydrological practices. Food and Agriculture Organization of the United Nations (FAO) published a manual in 1998 for estimating crop water requirements [1]. The FAO-56 Penman–Monteith (FAO-56 PM) is a physically based approach which requires measurements of air temperature, relative humidity, solar radiation, and wind speed. The FAO-56 Penman–Monteith (FAO-56 PM) method requires numerous weather data: maximum and minimum air temperature, maximum and minimum relative humidity, solar radiation (or sunshine hours), and wind

speed at height of 2m. The main limitation for widely using the globally accepted FAO-56 PM method is the numerous required data that are not available at many weather stations. The quality of these numerous data is also another problem. Solar radiation data are always lacking reliability. The reliability of the average 24-hr wind speed at 2m height is also questionable. Relative humidity measurement by electronic sensors is commonly full oferrors. But maximum temperature is largely and easily available in many regions of the world. In regions where there is insufficient data, a simple temperature method which requires only temperature data is important for the estimation of reference ET.


168

NCSTI-2012

There are different ET estimation methods developed in different time and space. Some methods work well in an area where they have been developed. When tested in other climatic conditions it performs less. The PM ETo is the only globally accepted method which performs well in many climatic conditions in the world. Many of the methods available are either data intensive and/or the data they require are not easily available.

The objective of this study is not to develop a new method which replaces FAO-56 PM, but to develop a new simple temperature method that uses only maximum temperature for areas where other climatic parameters are not available at the desired scale. The FAO-56 PM will be used as a standard method in the development of this new temperature method. This new simple temperature method could be used in areas where there is insufficient weather data. The specific objectives are to (1) develop an air temperature based method for estimating ET, (2) evaluate the performance of the new approach by comparing it to the most commonly used complex ET method (Penman–Monteith) and simple but widely used Hargreaves approach [2], and (3) evaluate the seasonal variation of simulated ET using dry and wet season temperature data.

II. STUDY AREA AND DATA

Ethiopia is found in the horn of Africa; it lies between about 3° to 15° N and 33° to 48° E. There is high altitude difference in the country which ranges from an elevation of about 116m below sea level, in the Dallol depression of the Afar region, to an elevation of about 4620m above sea level at Ras-Dashen, in the Semien Mountains in the northern part of the country. Due to this high altitude difference, there is high spatial variability of temperature. Whereas, seasonal variability of temperature in relatively minimum due to its geographic location. Ethiopia is locally subdivided in to five climatic regimes: moist, dry sub humid, semi-arid, arid, and hyper arid regimes. The Ethiopian National Meteorological Agency (ENMA) defines three seasons in Ethiopia:

rainy season (June to September), dry season (October to January) and short rainy season (February toMay). Camberlin [3] reported that the Indian monsoon activity is a major cause for summer rainfall variability in the East African highlands. The rainfall season over the study area starts in June and ends in October. Majority of the study area receives rain during this summer season (locally called “Kiremit”). The rest of the time it remains relatively dry and hot. There is high variability of rainfall distributions over time and space in the study area. Meteorological data were collected from the National Meteorological Agency of Ethiopia. Five to thirteen years of data were collected from six “class I” stations distributed over Ethiopia. The data collected from the stations include: daily maximum and minimum temperature, daily maximum and minimum relative humidity, sunshine hours, and wind speed at 2m. The stations are chosen because, they represent majority of the climatic in the upper Blue Nile. These stations lie within 10.3oN and 12.3oN latitudes (Fig. 1 and Table 1). Station names, their respective locations and data periods are shown in Table 1. Short period missed data were filled by simple averaging; whereas relatively longer periods of missed data were discarded from further analysis in this study.The distributions of the meteorological stations over the study area are shown in Fig. 1.

III. THEORY AND METHODS

Food and Agriculture Organization of the United Nations (FAO) in 1998 published a manual for estimating crop water requirements [1]. The FAO-56 Penman–Monteith (PM) is a physically based approach which requires measurements of air temperature, relative humidity, solar radiation (or sunshine hours), and wind speed. Reference evapotranspiration (ETo) is the potential ET from a reference surface of a hypothetical green grass of uniform height, 0.12m, well watered actively growing and a constant albedo of 0.23 with fixed surface resistance of 70s m-1 [1]. After the aerodynamic resistance, ra = 208/u2 and the


169

NCSTI-2012

surface resistance rs = 70s m-1 are estimated; for such a reference crop, the PM equation can be rewritten as:

2

2

34.01273

900408.0

u

eeuT

GRET

asn

o

(1) where, ETo=reference evapotranspiration (mm day-1) Rn=the net radiation at the crop surface

(MJ m-2 day-1), G = soil heat flux density (MJ m-2 day-1), assumed zero on daily basis,T= mean daily air temperature at 2 m height (°C), u2=wind speed at 2m height (m s-1), es = saturation vapour pressure (kPa), ea = actual vapor pressure (kPa), es = ea

Table I Station locations and the respective period of data Location

StationsLatitude Longitude

Elevation (m)

Time PeriodNo. of years

Ayehu 10.65° 36.78° 1774 2004-2008 5 Bahir Dar 11.36° 37.24° 1805 1998-2008 11Dangila 11.25° 36.83° 2000 1996-2008 13DebreMarkos 10·33° 37·72° 2430 2004-2008 5Gondar 12.3° 37.25° 1966 2004-2008 5Motta 11.06° 37.88° 2440 2004-2008 5

Fig.1. Location map of the study area


170

NCSTI-2012

saturation vapour pressure deficit (kPa), ∆ = slope of vapour pressure curve (kPa0C-1), γ = psychrometric constant (kPa0C-1).The FAO-56 PM combination equation (Eq. 1) was proposed as the sole standard method for estimating reference evapotranspiration, and for evaluating other equations [4, 5, 6]. It has been proved that this equation could overcome shortcomings of the previous methods and provided more consistent values in different regions of the world [7, 8, 9, 10, 11, 12, 13, 14]. In this study, the FAO-56 PM method was used as a standard method for the development of a new simple temperature method that use only maximum temperature for the estimation of reference ET. This study was based on the sensitivity analysis result over the study area. Temesgen [15] has showed that solar radiation and air temperature were almost equally sensitive to PM ETo. If temperature is then equally sensitive as solar radiation to PM ETo; and since, simple radiation methods like Abtew [16] and Modified Makkink [17] were performing well in the estimation of ET over the area, then there could be a simple temperature method that could perform as good as radiation methods. It is this sensitivity analysis result leads to the development of a new simple empirical temperature method which uses only maximum temperature data for the estimation reference ET. The new simple empirical temperature method, which we named it “Temesgen’s simple temperature method”, was developed as:

k

TETn

omax

(2) where, ETo is reference evapotranspiration (mm day-1), n = 2.5 which can be calibrated for local conditions, k = coefficient which can be calibrated for local conditions ranging from about 600 for lower mean annual maximum temperature areas to 1300 for higher mean annual maximum temperature areas. The coefficient, k could be approximated as k = 48*Tmm – 330 for combined wet and dry conditions, k = 73*Tmm – 1015 for dry seasons, and k = 38*Tmm – 63 for wet seasons, where

Tmm(oC) is the mean annual/seasonal maximum temperature. The coefficients of this new method were first calibrated using the global standard PM ETo method as a true value of ET. With these calibrated coefficients, the new method was compared with the PM ETo and a promising result was found. This new method was also compared with the known temperature method. The comparison of the new method with the Hargreaves (HAR) method was only to see the coefficient of determination (R2). This is because the Hargreaves method was also required local calibration of the coefficient over the study area. The Hargreaves and Samani [2] equation is well known temperature based method for the estimation of daily reference ET. This method requires daily maximum and minimum air temperature, and extraterrestrial solar radiation data. The extraterrestrial solar radiation is computed from the information of latitudes of the study site and Julian day of the year. The Hargreaves and Samani [2] equation is defined as:

(3) where, ETo is daily reference evapotranspiration (mm day-1), Tmax = daily maximum temperature (oC), Tmin = daily minimum temperature (oC), Tm = daily mean temperature (oC), and Ra = is the daily extraterrestrial solar radiation (mm day-1).

IV. RESULTS AND DISCUSSION A. Combined dry and wet season data

The simple new temperature model result was compared with the global standard PM ETo method and the well-known Hargreaves temperature method at all six stations. In semi-arid region of Ayehu, where the mean annual maximum temperature is higher 28.1oC, the coefficient k requires to be as large as 1200 to provide better results, whereas in areas where


171

NCSTI-2012

the mean annual maximum temperature is relatively lower about 23oC the coefficient k was found as low as 720 for the method to provide better results.

The comparison is made with coefficient of

determination (R2), Root Mean Square Error (RMSE), Absolute Mean Error (AME), and annual percentage error of the new method with the PM ETo. The coefficient of determination (R2) of the new method, when compared with PM ETo was as high as 0.74 at Debre Markos, and Gondar stations, and as low as 0.61 at Bahir Dar station. The annual average coefficients of determinations over the six stations were 0.68 and 0.84 when compared with PM and HAR methods, respectively. Even though, Hargreaves method over estimates daily ETo when compared with PM ETo, which indicates local calibration of the Hargreaves coefficients is required. The coefficients of determinations

(R2) of this new temperature method were also compared with the Hargreaves temperature method. The coefficient of determination (R2) of the new method with the Hargreaves temperature method was as high as 0.88 at Ayehu and Debre Markos stations, and as low as, 0.76 at Gondar station. Generally, this new simple method was able to estimate daily ETo with an average Root Mean Square Error (RMSE) and an average Absolute Mean Error of 0.58mm and 0.46mm, respectively from the global standard PM ETo method. The method was able to estimate annual ETo with an annual average error of less than 3.3% from the PM ETo estimations. The summary of comparison results of the new method with PM ETo method and the Hargreaves method is presented in Table 2. The comparison scatter plots of the new method with the PM ETo and the Hargreaves methods are also shown in Figs. 2 and 3, respectively.

Table II Combined wet and dry season comparison results new method with PM and HAR methods. R2 PM Stations Period

(years) PM HAR RMSE

(mm) AME (mm)

Annual avg. error (%)

Ayehu 2004-2008 0.68 0.88 0.64 0.52 3.6

Bahir Dar 1998-2008 0.61 0.81 0.56 0.44 5.7

Dangla 1996-2008 0.67 0.87 0.53 0.43 2

DebreMarkos 2004-2008 0.71 0.88 0.58 0.45 3.85

Gondar 2004-2008 0.74 0.76 0.54 0.44 1

Motta 2004-2008 0.67 0.83 0.6 0.48 3.2

Average 0.68 0.84 0.58 0.46 3.23

B. Wet and dry season variation The performance of this simple new

temperature model result was also tested in the dry and wet seasons in the study area. It was compared with the global standard PM ETo method and the well-known Hargreaves temperature method in the dry and wet seasons at all stations considered. The comparison was based on: coefficient of determination (R2), Root Mean Square Error

(RMSE), and Absolute Mean Error (AME) of the new method with the PM ETo, and only R2 for comparison of the new method with the HAR method.

Dry seasons In dry seasons, the new method comparison with the PM method; R2was as high as 0.65 at Dangila, Gondar, and Motta stations and as


172

NCSTI-2012

low as 0.6 at Bahir Dar, with an average R2 value of 0.63 over the six stations. The RMSE and AME of the new method from the standard PM method also range from minimum value of 0.42 and 0.32 at Ayehu station, to maximum value of 0.54mm and 0.42mm at Motta station, respectively.

The R2 comparison of the new method with HAR was as high as 0.84 at Dangila and as low as 0.75 at Gondar stations. The average R2 of new method with HAR was 0.82 over the stations. The detail results are shown in Table 3. The scatter plots of the dry season’s comparison of the new method with the PM and HAR methods are shown Figs. 4 and 5, respectively.

Table III Dry season comparison results new method with PM and HAR methods

Period R2 PMStations

(years) PM HAR RMSE (mm)

AME (mm)

Ayehu 2004-2008 0.61 0.83 0.42 0.32 Bahir Dar 1998-2008 0.6 0.81 0.5 0.39

Dangla 1996-2008 0.65 0.84 0.44 0.34 DebreMarkos 2004-2008 0.61 0.83 0.46 0.36 Gondar 2004-2008 0.65 0.75 0.46 0.36 Motta 2004-2008 0.65 0.83 0.54 0.42 Average 0.63 0.82 0.47 0.37


173

NCSTI-2012

Fig. 2. Combined dry and wet season comparison of the new method with PM method


174

NCSTI-2012

Fig. 3. Combined dry and wet season comparison of the new method with the Hargreaves method

Fig. 4. Dry season comparison of the new method with the PM method

Fig. 5 Dry season comparison of the new method with the Hargreaves method Wet seasons


175

NCSTI-2012

In wet seasons, the new method comparison with the PM method; R2 was as high as 0.64 at Ayehu station and as low as 0.56 at Debre Markos, with an average R2 value of 0.59 over the stations. The RMSE and AME of the new method from the standard PM method also range from minimum value of 0.33 and 0.27, at Ayehu station to maximum value of 0.47mm and 0.34mm, respectively at Motta station. The R2 comparison of the new method with HAR was as high as 0.87 at Bahir Dar and Gondar and as low as 0.67 at Motta station. The average R2 of new method with HAR was 0.81 over the six stations.

During wet seasons, the new method is also

performing well in terms of both RMSE and AME. The detail results are shown in Table 4. The scatter plots of the wet season’s comparison of the new method with the PM and HAR methods are shown Figs. 6 and 7, respectively.

Generally, the method performs well in the

study area in both dry and wet seasons too with an average R2, RMSE, and AME of0.63, 0.47, and 0.37 in dry seasons and 0.59, 0.40mm, and 0.31mm in wet seasons, respectively when compared to the PM

method. The average R2 of the new method with the HAR method was 0.82 and 0.81 in dry and wet seasons, respectively.

V. CONCLUSIONS AND RECOMMENDATIONS

This simple method which requires only daily maximum temperature is able to estimate daily ETo with average R2, RMSE and AME of 0.68, 0.58mm and 0.46 mm, respectively and an average annual percentage error of less than 3.3 % from the global standard PM ETo method.

The method could be used for the estimation

daily ETo where data is insufficient. Calibration of the coefficients is recommended. If there is no sufficient data for calibration, the coefficient k can be approximated as k = 48*Tmm – 330 for combined wet and dry seasons, k = 73*Tmm – 1015 for dry season, and k = 38*Tmm – 63 for wet season, where Tmm(oC) is the mean annual/seasonal maximum temperature. This method is most recommended for the estimation of annual ETo in the study of hydrology.

Table IV Wet season comparison results new method with PM and HAR methods R2 PMStations Period (years)

PM HAR RMSE(mm) AME (mm)

Ayehu 2004-2008 0.64 0.83 0.33 0.27

B-Dar 1998-2008 0.57 0.87 0.38 0.3

Dangla 1996-2008 0.57 0.79 0.38 0.3

D-Markos 2004-2008 0.56 0.84 0.41 0.33

Gondar 2004-2008 0.6 0.87 0.41 0.33

Motta 2004-2008 0.59 0.67 0.47 0.34

Average 0.59 0.81 0.40 0.31


176

NCSTI-2012

Fig. 6. Wet season comparison of the new method with the PM

Fig. 7.Wet season comparison of the new method with the Hargreaves method


177

NCSTI-2012

Acknowledgment

We thank the School of Civil and Water Resources Engineering, Institute of Technologyof Bahir Dar University for supporting this work, and we also acknowledge the Ethiopian Meteorological Agency for providing data.

REFERENCES [1] Allen RG, Pereira LS. 1998,“Crop

Evapotranspiration-Guidelines for Computing Crop Water Requirement”, Rome, FAO Papers No.56, Roma

[2] Hargreaves GH, Samni ZA. 1985, “Reference crop evapotranspiration from temperature.” Transactions of the American Society of Agricultural Engineers

[3] Camberlin P. 1997, “Rainfall anomalies in the Source Region of the Nile and their connection with the Indian Summer Monsoon”J. Climate1380–1392.

[4] Allen RG, Smith M, Perrier A, Pereira LS. 1994a, “An update for the calculation of reference evapotranspiration”,ICID Bull43(2): 35–92.

[5] Allen RG, Smith M, Perrier A, Pereira LS. 1994b, “An update for the definition of reference evapotranspiration”, ICID Bull 43(2):1–34.

[6] Slavisa T. 2005, “Temperature-Based Approaches for Estimating Reference Evapotranspiration”,J. Irrig. Drain. Eng. Vol. 131, No. 4, July/August 2005, pp. 316-323.

[7] Bois B, Pieri P, Van LeeuwenC, Wald L, Huard F, Gaudillere JP, Saur E. 2008,” Using a remotely sensed solar radiation data for reference evapotranspiration estimation at a daily timestep”,Agricultural and Forest Meteorology 148(4): 619-630

[8] Dingman SL. 2002, “Physical hydrology.Prentice Hall, Upper Saddle River”, 646 pp.

[9] Liu Y, Pereira LS, Teixira JL. 1997, “Update definition and computation of reference evapotranspiration comparison with former method”Journal of hydraulic

engineering 6: 27-33.(In Chinese with English abstract)

[10] Rana, G. and Katerji, N., 2000,“Measurement and estimation of actual evapotranspiration in the fieldunder Mediterranean climate: a review”,European Journal of Agronomy 13(2-3): 125-153.

[11] Ventura F,Spano D, Duce P. 1999, “An evaluation of common evapotranspiration equations”Irrig. Sci. 18.163-170.

[12] Du YD, Liu ZX, Zhang YF. 2000,“Evaluation of two reference crop evapotranspiration calculation methods”,Journal of Henan Agricultural University35(1):57-31. (In Chinese with English abstract)

[13] Nandagiri L, Kovoor GM. 2005, “Sensitivity of the food andagriculture organization Penman-Monteith evapotranspirationestimates to alternative procedures for estimation of parameters, ”J. Irrig. Drain. Eng131(3):238-248

[14] Temesgen B, Eching S, Davidoff B. 2005, “Comparison of some reference evapotranspiration equations for California”,J. Irrig. Drain. Eng. 131(1): 73-84

[15] Temesgen E.N. 2009, “Estimation of evaporation from satellite remote sensing and meteorologicaldata over the Fogera flood plain, Ethiopia”, Enschede, MSc thesis ITC, p. 89, 2009.

[16] Abtew W. 1996, “Evapotranspiration measurements and modeling for three wetland systems in South Florida”. In: Water Resources Bulletin, 32(1996)3: 465-473.

[17] De Bruin HAR..1981,“The determination of (reference crop) evapotranspiration from routine weatherdata”,Proc. and Inform.Comm.Hydr. Research TNO, The Hague 28: 25-37.

Analyzing the Combined Impact of Climate and Land Use Changes on Sediment Yield and ...

178

NCSTI-2012

Analyzing the Combined Impact of Climate and Land Use Changes on Sediment Yield and Stream Flow in the Upper

Gilgel Abbay Catchment, Ethiopia

Kirubel Mekonnen a , Gebreyesus Brhane Tesfahunegna, Kassa Tadeleb

a College of Agriculture, Aksum University-Shire Campus, P. O. Box 314, Shire, Ethiopia b Arbaminch University, P.O.Box 21,Arbaminch, Ethiopia

Abstract- Climate and land use are among the main factors that directly influencing the hydrological processes in a catchment. It is thus crucial to know the combined effects of land use and climate change for sound land use planning and water resource management. Despite this fact, little is documented in assessing the combined impact of different scenarios of climate and land use changes on sediment yield and stream flow. The Soil and Water Assessment Tool (SWAT) model was used to examine the combined effects of the possible land use and climate change scenarios on sediment yield and stream flow in the Upper Gilgel Abbay Catchment. For this purpose, three plausible land use scenarios such as trend, conservation, and plan based land use scenarios were developed and integrated with the projected meteorological data from climate local model simulation in the foreseeable future of 2011-2025. The combined effect of land use and climate change scenarios were evaluated and quantified by comparing SWAT outputs to baseline simulation (1996-2010). The simulation results indicated that the combined effect of trend based land use and climate change scenario significantly increased the average annual sediment yield by 29% and stream flow by 5.2% as compared with the existing catchment condition. However, the integrated effects of conservation based land use and climate change scenario will considerably decreases sediment yield and stream flow by 13.7% and 2.4%, respectively. The implementation of both plan, and conservation based land use scenarios in 2011-2025 periods will likely reduces the increased sediment yield (1.6%) and stream flow (4.6%) attributable to the projected climate change impacts and can be thus used as an adaptive strategy in tackling the problem of sediment yield and stream flow caused by climate change. In all the scenarios simulated, the combined effects of land use and climate change would considerably alter the hydrological behavior of Upper Gilgel Abbay Catchment condition than the separate effects of land use and climate change. Therefore, in the planning of adaptation strategies for climate and land use change impacts on catchment hydrology, it is necessary to consider the combined effects of climate and land use change rather than to simulate their impacts separately.

Keywords: Land use change, Climate change, sediment yield, Stream flow, SWAT model, Adaptation strategies

I. Introduction Climate and land use change are ubiquitous

drivers of global environmental change such as land degradation [1]. Land degradation is a series threat in the Ethiopian highlands which is reflected in the form of soil erosion and soil fertility decline [2]. Water erosion is a major cause of land degradation that affects the physical and chemical properties of soils and resulting in on-site nutrient loss and off-site sedimentation of water resources (Phillips, 1989). The off-site effects of erosion such as reservoir sedimentation and water resources pollution are usually more costly and severe than the on-site effects on land resources [3]. Therefore, proper management practices that tackle on-site effects of soil erosion should be designed to reduce the risks and negative impacts of downstream water resources due to water erosion. Tackling the on-site effects of soil erosion requires an understanding the rates and processes of soil erosion.

Soil erosion rates may be expected to change in response to changes in climate for a variety of reasons, the most direct of which is the change in the erosive power of rainfall [4-8]. Also, the report by [9] states that soil erosion responds highly to the change of rainfall intensity and energy rather than

rainfall amount alone. In line to this, simulations of future rainfall patterns show no clear trend of increase or decrease of rainfall amount in the Upper Gilgel Abbay catchment, Ethiopia, which is a bias towards more intense rainfall events in the next century [10-12]. The trend towards precipitation occurring in more extreme events must be adequately simulated for soil erosion assessment, because most soil loss is caused by infrequent severe storms [13]. The potential for such projected changes to increase the risk of soil erosion and related environmental consequences is clear, but the actual damage is not known and needs to be thus assessed in this study.

Moreover, the change of land cover /land use pattern has also led to accelerated soil erosion which has raised a serious potential threat on agricultural production. The land cover in the Upper Gilgel Abbay catchment between the years 1973-2001 changed significantly in terms of spatial extent. For example, in the year’s between1973-1986, forest covers decreases by 1.38% while it was 0.92 % between the years1986-2001. This was mainly due to the expansion of agriculture land at the rate of 1.53% and 1.07% corresponding to the


179

NCSTI-2012

same period. The reduction forest cover has caused an increase of peak flow (1.3%), soil erosion (2.2%) and the base flow (0.7%) in the dry season [14]. In addition, substantial studies elsewhere showed that the changes in land-cover have affected the surface and groundwater hydrology and altering the hydrological cycle [15,16].

Land cover change is regarded as the single most important variable of global change affecting ecological systems that associated with climate [17].Natural and anthropogenic factors as well as other emerging environmental (climate), social and political factors can have significant effects on future land use/land cover. Patterns of land use/land cover change, and land management are shaped by the interaction of economic, environmental, social, political, and technological forces on local to global scales. These effects can vary as functions of seasonality and changes in climate [18]. The change of climate variables directly affects the amount and intensity of precipitation and potential evapo-transpiration [19]. This indirectly affects plant water use efficiency through altering plant growth rate and species composition which result in the change of land cover/land use [20]. It might be thus appropriate to study climate and land use/land cover and their combined effects on catchment hydrology in order to formulate sound land use planning and water resource management.

The study was conducted in the Upper Gilgel Abbay Catchment in Ethiopia (Fig.1). The study catchment is bounded by latitudes 10o56’to 11o5’ N and longitudes 36o44’to 37o23’ E. The area of the study catchment is 1637 km2. The study catchment receives high amount of rainfall between June and September, while the small rainy season is usually between March and May. The average annual rainfall is 1800 mm and the mean annual temperature falls in the range of 16o

C to 20oC. The

spatial variation of rainfall amount in the area indicated that a decreasing trend from South to North part of the catchment. The dominant soil type is Haplic Luvisoil and covers about 56% of the catchment area. The study catchment is characterized by landforms which are ranged from flat plains, undulating plains and rolling land to steep mountains. The description of the topography is adopted the slope capability classification made by [29] and the slope ranges and its area coverage of each land forms were estimated from digital elevation model as presented in Table 1.

Many studies revealed that hydrologic models provide a framework for examining the complex effects of both climate and land use changes on soil erosion and stream flow [21,22], among which distributed hydrological models have important applications because they relate model parameters directly to physically observable land surface characteristics[23]. Among the distributed hydrological models, previous studies suggested that Soil and Water Assessment Tool (SWAT) model is appropriate to examine the impact of land use and climate change on soil erosion and stream flow [24-26]. These researchers pointed that the effect of climate is more dominant on stream flow than soil erosion whereas land-cover change may have a moderate impact on stream flow, and it strongly influences soil erosion as land cover pattern determines to a large extent the sediment delivery ratio of a drainage basin. Despite the above facts, however, little is known on the combined effects of climate and land use change using physical models across different physiographic regions in general and in the Upper Gilgel Abbay Catchment in

particular, with regard to their impacts on soil erosion. The importance of such research gap has also been pointed in recent reviews [27-28].

The aim of the present study is to analyze the combined effect of climate and land use changes on sediment yield and stream flow at spatial and temporal scale in the foreseeable future (2011-2025), using the SWAT model in the Upper Gilgel Abbay catchment of Ethiopia. Such information is an important step towards a comprehensive probability-based estimation of environmental change impact assessment.

II. Materials and Methods A. Catchment Description

Land forms Slope(%)

Areacoverage (ha)

Area(%)

Flat land 0-2 17208.8 10.5

Undulating land 2-8 89286.1 54.5 Rolling land 8-15 24527.8 14.9 Hill to rolling 15-35 24920.8 15.3 Steep mountain >35 7763.9 4.8 Total 163707.4 100

Table 1. Land forms and their area coverage in the Upper Gilgel Abbay Catchment, Ethiopia.


180

NCSTI-2012

Figure 1.Map of the study area (kirubel et al., 2012)B. Description of SWAT Model

The SWAT model is a catchment-scale, physically based distributed hydrological model developed to predict the impact of land management practices on hydrologic and water quality response of complex watersheds with heterogeneous soils and land use conditions [30]. The model partitions a watershed into sub watersheds and organizes input information for each sub watershed into the following categories: climate, hydrologic response units (HRUs), ponds/wetlands, groundwater, and the main reach draining to each sub watershed.

SWAT simulates surface runoff for each HRU based on USDA NRCS Curve Number (CN) method (USDA-SCS, 1972) or Green and Ampt infiltration method (Green and Ampt, 1911). Green and Ampt infiltration method is required half -hour climate data to estimate surface runoff, however, these data were unavailable in the study catchment. For this reason, the CN method was used. SWAT calculates the surface erosion caused by rainfall and runoff within each HRU using the Modified Universal Soil Loss Equation (MUSLE). MUSLE is a modified version of the Universal Soil Loss Equation (USLE) developed by [31]. The SWAT model calculates potential evapo-transpiration by three methods: (1) the Penman-Monteith method; (2) the Priestly-Taylor method; and (3) the Hargreaves method (Neitsch et al., 2005). Among the three methods, Penman-Monteith (Monteith, 1965) was used for this study because this method

has the aerodynamic and energy balance components which gives reliable results for tropical Africa countries like Ethiopia [23] .

SWAT model InputsData required for SWAT Model are classified in

to two groups: spatial input data and hydro-meteorological data. The details of these data were given below.

a) Spatial Input data Digital elevation model (DEM), land use and soil

map are the three spatial inputs data required by the SWAT model. These data were collected from the Ethiopia Ministry of Water and Energy (MoWE). A 90-m resolution DEM was used in extracting stream network and delineating catchment and sub-catchments.

The land use/ land cover map derived in the year of 2002 was used for the SWAT model and considered as the base reference for the study catchment. This is mainly due to the fact that the extent of land use/ cover changes in the Upper Gilgel Abbay catchment were insignificant since 2002. The reclassification of the land use map was done to represent the land use according to the model specific land cover types. Accordingly, the land use and land cover in the basin can be categorized into cultivated land (AGRL, 74 %), shrub and bush land (RNGB, 11.3%), grassland (PAST, 9.6%) and forested lands (FRSD, 5.1%). Soil data is another spatial input data required by the SWAT model. Soil compositions and properties used to define the basin soils in the SWAT model


181

NCSTI-2012

were collected from Abbay Basin Master Plan document and categorized based on FAO soil classification. The major soil types are Halpic Luvisoils (LVh, 56.3%), Haplic Alisol (ALh, 40.5%), Eutric Vertisol (VRe,1.8%), Eutric Regesol (RGe,0.8%) and Haplic Nitisols (NTh, 0.6% ) .

b) Hydro-Meteorological Data

The meteorological inputs during 1996-2010 were collected from the Ethiopia National Meteorology Agency, which included daily precipitation, maximum and minimum temperatures, net radiation, near surface wind, and relative humidity of the air and direction for seven weather stations in or around the Upper Gilgel Abbay catchment. Moreover, the dynamically downscaled future climate data from Climate Local Models (CLM) simulation were obtained from the International Water Management Institute (IWMI) for the period of 15 years (2011 to 2025). The CLM is designed to downscale climate information from coarse-resolution of Global Circulation Model (GCM) to local or site level. The Intergovernmental Panel on Climate Change (IPCC) scenario of A1B (ECHAM5) was used. This climate scenario is based on IPCC emission scenario (SRES A1B). The A1B emission scenario assumes an integrated world where income and way of life converge between regions, with rapid economic growth, a quick spread of new and efficient technologies and a balanced emphasis on all available resources.

C. Model Calibration, Validation and Performance Evaluation

The more we know about the catchment to be modeled and the better the model is able to ‘represent’ this knowledge, the more reliable the model results will be obtained [32]. Taking this in to account, monthly model calibration was employed after identifying the most sensitivity parameter for flow and sediment for a period of 1996-2002 in which 1996 was used as a warm-up period. The warm-up period allows the model to get the hydrologic cycle fully operational. The manual calibration method outlined in the SWAT Version 2005 user’s manual [33] was used to minimize percent difference of observed and simulated model results and maximize Nash Sutcliffe efficiency. Stream flow and suspended sediment concentration measurements were used for comparisons against the modeled stream flow and sediment load during model calibration and validation. Daily stream flow and suspended sediment concentration (SSC) data were obtained from MoWE. However, the suspended sediment concentration data observed in the study catchment were not adequate to calibrate and validate the SWAT model. Thus, we used

sediment-rating curve to estimate SSCs to calculate sediment loads.

Using independent data sets, model validations were undertaken for flow and sediment load for the period of 2003-2005. The SWAT model results for calibration and validation periods were evaluated and tested using different techniques which include: (1) the Nash–Sutcliffe efficiency (NSE), (2) coefficient of determination (COD) and (3) Average percent difference of the simulated and observed values (APD). The NSE values goes to 1 as the fits improves. A value between 0.6 and 0.8 indicates that the model performs reasonably. Values between 0.8 and 1indicates the model performs very well (Nash and stucliffe 1970). The value of the COD describes the proportion of variance in the observed data explained by the model. Values can range from 0 to 1. A value of 1 indicates that the model explains all of the variance observed in the measured data. A value of 0 indicates that the model explains none of the observed variance. Generally, COD values above 0.5 are considered adequate [34]. The percentage difference between mean modeled values and mean observed values was also calculated to show overall model bias. [35]states that stream flow APD should be within 15 percent and for sediments APD should be within 50 percent.

D. Integrated Land Use and Climate Scenario Development in 2011-2025 Period

A changed or an altered land use condition was not chosen arbitrarily for this study. Instead, it was based on a reasonable assumption about the trend of driving force that influencing land use. Three plausible land use scenarios such as trend, conservation, and plan based land use scenarios were developed for the foreseeable future of 2011-2025 and then integrated with projected meteorological data from climate local model simulation. The descriptions of the scenarios are given as follows.

Scenario1: The first land use scenario (scenario 1) refers to the baseline condition of the catchment that simulated using the SWAT model.

Scenario 2: Scenario 2 (trend based land use scenario) was developed based on past trends of land use/land cover change (1973, 1986, and 2001) of the study area and assumed recent trends will continue in 2011-2025 period . Time series satellite image analysis results showed that agricultural land has increased at an alarming rate for the past 30 years .Since agriculture is the major source of income in the study area, the conversion of the current natural ecosystems into agriculture is unavoidable in fact, a logical choice during the projected period of 2011 -2025. The main questions to be addressed using scenario 2 are that


182

NCSTI-2012

at which location it will occur, the size of total cropland and the type of crops to be cultivated. Thus, for this scenario, land covers such as Forest, Pasture and Shrub which are located on the slope below 8% and having a high biomass production of greater than 8050 kg/ha ,less sever to soil erosion (less than 10 t/ha) and shorter distance from the water sources (less than 0.5 miles) are likely to be changed in to agricultural production

Scenario 3: This scenario is a plan based land use scenarios, which more focuses on the existing comprehensive and sustainable land use plan that are implemented. Sustainable land use plan criteria set by Tana Beles Integrated Water Resources Development, TBIWRD, (2009) were used to modify the base reference land cover of the HRUs. For instance, agriculture should take place in areas that meet the following environmental characteristics: No degraded soils (biomass production greater than 8050 kg/ha ), slopes < 8 % for rain fed agriculture and slopes should be < 6% for irrigated agriculture and outside protected land areas. Besides, certain ecologically sensitive land cover classes, e.g., forests with C-factor = 0.001, woody savannah C-factor = 0.002 or land cover classes with a competitive economic value such as pastures for livestock with C-factor = 0.0016 with a lope over 2% were conserved and exempted from any future agricultural activity. However, cultivated

lands on the slope of above 35% are returned to forest.

Scenario 4: This scenario (conservation based land use scenario) places high priority for catchment management and ecological restoration in order to reduce sediment inflow to the planned Gilgel Abbay reservoir.. For this scenario, HRUs with slope above 35%, and a biomass production of less than 8050 kg ha-1 were changed in to natural forest. Besides this, best management practices such as contour bund, parallel terrace and filter strip were modeled in the high erosion prone sub-basins( greater than 25 t/ha) and their parameter values were represented by the SWAT Model.

E. Evaluating the Effect of Land use and Climate Change

The approach to evaluate the effect land use and climate change on soil erosion and stream flow was similar to the one used by [26]. The approach of one factor at a time was used (i.e., changing one factor at a time while holding others constant). Meteorological data of the two time-slices of 1996–2011 and 2011–2025 were selected, and each time-slice included one land use scenario. The base reference land use represent 1996–2011 periods and the respective various future land use scenario represent 2011–2025 time-slices. The influences of the land use and climate change were quantified on monthly and annual time step by comparing SWAT outputs to baseline run

Table 2. Combination of scenarios to evaluate the effect of land use and climate change

Scenarios Description of scenarios S1 Base reference land use map and1996-2011 climate ( baseline run) S2 Base reference land use and 2011-2025 climate S3 Trend based land use and1996-2011 climate S4 Trend based land use and 2011-2025 climate S5 Plan based land use and1996-2011 climate S6 Plan based land use and 2011-2025 climate S7 Conservation based land use and 1996-2011 climate S8 Conservation based land use and 2011-2025 climate

III. Results and Discussion

A. SWAT Model Sensitivity Analysis, Calibration and Validation

Among 27 flow parameters, only eight sensitive parameters are identified for the model to avoid model over parameterization. Table 3 shows sensitive parameters for flow and their calibrated values for the gauged Upper Gilgel Abbay catchment. The curve number (CN), base flow alpha factor and threshold water depth in the shallow aquifer are the three most sensitive flow parameter which affects modeled surface and sub-

surface flows of the catchment. Hence, the surface flow component of average annual total water yield is balanced by adjusting the runoff CN for forested, shrub, pastured and cropped land use (Table 3) and an effort was made to keep the CN close to standard table values. The most sensitive parameters for prediction of sediment yield are linear parameter which include channel sediment routing (Spcon), channel cover factor (Ch_Cov), exponent parameter for calculating sediment re-entrained in channel sediment routing (Spexp) and USLE equation support practice factor (USLE_P). These sediment parameters are presented with their calibrated values in Table 4.


183

NCSTI-2012

After calibrating SWAT model for flow and sediment load, a simulation was executed. The results of the model performance measures in the calibration and validation period are presented in Table 5. The monthly calibration and validation of the SWAT model for flow and sediment load have shown that the model can predict the flow and

sediment load as indicated by model performance evaluation measures. The agreement between the measurement and simulation values is generally very good which was verified by higher NSE, R2 and APD and an acceptable results are thus obtained according to the model evaluation guidelines reported in [34,35] .

Table 3. Sensitive and calibrated flow parameters for the gauged Upper Gilgel Abbay

Rank Description Flow parameter Lower and upper bound

Initial &default value

Optimized value

1 Initial SCS Curve Number II (Cn2) -25 -25 Default +10%

2 Base flow alpha factor (Alpha_Bf) 0 – 1 0.048 0.048 3 Threshold water depth in the shallow

aquifer required for return flow (Gwqmn) 0 -5000 0 32

4 Threshold water depth in the shallow aquifer revap to occur (Revapmn)

0 -500 0 32

5 Soil available water (Sol_Awc) -25-25 From literature -20%

6 Soil depth (Sol_Z) 0-3000 From literature Not adjusted 7 Soil evaporation compensation (Esco) 0-1 0.95 0.75 8 Ground water revap coefficient

(Gw_Revap) 0.02-0.2 0.02 0.2

Rank Description sediment parameter

Lower and upper bound

Initial and default value

Optimized value

1 Spcon 0.0001 - 0.01 0.0001 0.001 2 Ch_Cov 0-1 0 0.4

3 Spexp 1-2 1 1.3 4 USLE_P 0-1 1 0.8 5 USLE_C ± 25 Default *Based on land cover of HRU

Table 4. Sensitive and calibrated sediment parameters for the gauged Upper Gilgel Abbay

*USLE_C factors were adjusted based on the land covers of Hydrological response unit Table 5. Monthly model evaluation statistics for flow and sediment load in Upper Gilgel Abbay

Calibration(1997-2002)

Validation (2003-2005)

Performance measure

Stream flow Sediment Stream flow Sediment Nash Sutcliff efficiency(NSE) 0.92 0.86 0.87 0.81 Average Percentage difference(APD) -0.54% 3.39% -4.2% -10.07% Coefficient of Determination (R2) 0.94 0.88 0.90 0.84

B. Land Cover Classes and Land Use Changes under the Different Land Use Scenarios

The possible area coverage of land covers and land use change that will be occurred in trend, plan, and conservation based land use are presented in Table 6. Agricultural land is the dominant land cover in all the land use scenarios conditions. The area coverage of agricultural land is increased from

74% to 82% in the trend based land use scenario which is most probably associated with the decreasing of shrub and pastureland cover classes that have high evapotranspiration and high biomass production (> 8050 kg ha-1). This change of land covers are mainly expected in undulating plain (2-8%) and rolling land forms (15-30). However, the area coverage of cultivated land is reduced by 8% in the plan and by 5.5% in the conservation based


184

NCSTI-2012

land use scenario, even though the demand for cultivated land will increase. This is mainly due to the fact that both plan and conservation based land use assumed that clearing of natural vegetation for agricultural purpose is not a solution in satisfying the demand of the population during 2011-2025 periods. Rather both scenarios have given due attention of the improvement of agricultural production per unit area using various inputs

(quality of seed, management practice, irrigation etc).

The land use changes per catchment slope were assessed for the three plausible land use scenarios and presented in Table 7. During the projected period of 2011–2025, the major likely conversion of land use in trend, plan and conservation based land use scenarios will be shrub land to agriculture (4.7%), agricultural land to pasture (4.8%) and agricultural lands to forest (5.5%), respectively

Table 6. Land cover classes and their area coverage under different land use scenarios

Area coverage of land covers under different land use scenarios

Base reference Plan Trend Conservation

Land cover classes

Km2 % Km2 % Km2 % Km2 % Agricultural land(AGRL) 1211.4 74.0 1080.5 66.0 1342.4 82.0 1121.4 68.5 Shrub land (RNGB) 185.0 11.3 201.4 12.3 108.0 6.6 185.0 11.3

Pasture(PAST) 157.2 9.6 235.7 14.4 103.1 6.3 134.2 8.2

Forest (FRSD) 83.5 5.1 119.5 7.3 83.5 5.1 196.4 12.0

.

Table 7. Summary of land use change types and percentage area of change per catchment slope under different land use scenario

Changed area per catchment slope (%) in km2

Land use scenarios

Land use Change types

2-8 8-15 15-35 >35 Total

Percentage of area change in the basin

RNGB to AGRL 14 63 - - 77 4.7 Trend based PAST to AGRL 21 33 - - 54 3.3 AGRL to PAST - 7 74 - 81 4.8 AGRL to FRSD - 9 8 20 37 2.2

Plan based

AGRL to RNGB - 14 3 - 17 1 AGRL to FRSD 6.5 29.7 33.8 20 90 5.5 Conservation

based PAST to FRSD 23 - - - 23 1.4

C. Land Use Change Impacts on Sediment Yield and Stream flow

The land use change impact assessment focuses more on the internal dynamics of the hydrological system. The effects of land-use changes were estimated under the same climate conditions. The differences between the newly developed land use scenarios and the baseline condition showed the effect of land use change on the catchment’s

sediment yield and stream flow as the results are summarized in Table 8.


185

NCSTI-2012

Table 8. Simulated average annual catchment sediment yield, stream flow and their percentage change relative to the baseline condition for the changed land use scenarios in the period of 2011-2025

Flow Simulation Sediment yield simulation Land use Scenarios

M3 s-1 BCM Change % (Million ton) Change % Baseline (1996-2010) 52.44 1.65 - 2.48

Conservation 51.59 1.63 -1.62 1.98 -20.2

Plan 51.89 1.64 -1.05 2.2 -11.29

Trend 53.35 1.68 1.73 2.9 17

BCM = Billion Cubic Meter.

The average annual stream flow and suspended sediment yield at the outlet of the Upper Gilgel Abbay catchment were estimated as 1.65 BCM and 2.48 million ton y-1, respectively, for the current/baseline condition (Table 8). The suspended sediment load result is quite comparable to 2.82 x 106 ton y-1 that estimated by [38] which includes bed load as well. In general, the model simulation with different land use scenarios provided very interesting results. For example, the implementation of conservation based land use in 2011-2025 periods will reduce the average annual flow and sediment load by 1.62% and 20.2 %, respectively. The decrease of stream flow in the conservation based land use scenario may be attributed to the changed agricultural land to forest land (table 7) which leads the increase of the rate of water loss in the study catchment because of large ET of forest land. In line with this, the decrease of sediment yield in this scenario is mainly associated with the area coverage of larger canopy of forest and the best management practices modeled that reduces the detachment and transport of soil particles.

In the plan based land use scenario, the flow and sediment yield is reduced by 1.05 % and 11.3%, respectively. The decrease of flow and sediment yield in the catchment is related to the area coverage of the changed forest and grass land covers. However, the change of land cover/land use in trend based land use scenario would affect the hydrological behavior of the Upper Gilgel Abbay

Catchment at large and it increases stream flow and sediment yield by 1.73% and 17%, respectively, as compared to the baseline condition (Table 8). This will affect the economic life of the planned Gilgel Abbay reservoir unless implementing appropriate land use planning and soil and water conservation measures.

The effect of the different land use scenarios on sediment yield and stream flow are shown in Fig. 2. From the graphs, it is evident that all land use scenarios may reduce sediment yield in the whole period of simulation except the trend based land use scenario. The plan and conservation based land use scenario can increase catchment discharge in the dry periods i.e from November up to May whereas the trend based land use scenario increases the peak flow and sediment yield by 4.4% and 3.5 %, respectively . The results revealed that alteration and reformation of the land use would be an efficient process in reducing sediment yield in the Upper Gilgel Abbay Catchment. Sediment yield is likely to be more affected by changes land use/cover land-cover than stream flow, and the change of land cover in all land use scenarios may have a moderate impact on annual stream flow (table 8). However it strongly influences seasonal stream flow (fig. 2)

Figure 2. Average monthly basin discharge and sediment load anomalies (relative to the baseline condition) under different land use scenarios 2011-2025


186

NCSTI-2012

D. Projected Climate Change Scenario Results

The projected climate change results indicated that there is an overall increasing trend in annual temperature and significant variation of monthly and seasonal precipitation from the base period level (Fig. 3 and 4). In the main rainy season, the mean monthly rainfall indicates a decreasing trend in the beginning of the rainy season (May and June) and an increasing trend towards the end of the rainy season (September and October). Compared with the base line climate, annual mean precipitation in 2011-2025 increased about 6.7% and annual mean

temperature increased about 10C in the Upper Gilgel Abbay catchment. Such climate change would possibly influence the hydrological behavior of the catchment. For example, an increasing of precipitation and a decreasing of temperature during August and September would increase stream flow significantly owing to an increase of sediment yield. Besides, a temperature increase in the month of May would possibly increase evapotranspiration and subsequently decrease stream flow(fig.3).

Figure 3. Monthly rain fall and temperature of the Upper Gilgel Abbay Catchment during 1996-2010 and 2011-2025 simulation periods.

Figure 4. Trend of average annual precipitation and temperature of the Upper Gilgel Abbay catchment (1996-

2025).


187

NCSTI-2012

E. Projected Climate Change Impacts on Sediment Yield and Stream Flow

Table 9 shows the average annual stream flow and sediment yield simulated by the SWAT model under the projected climate change scenarios for the catchment. This impact assessment mostly concentrates on the changed meteorological forcing. The projected climate change scenario cause an increasing of annual flow and sediment yield by 4.6% and 1.6% respectively. Stream flow at gauging station is expected to be 1.73 BCM in the foreseeable future, 2011-2025. Hence, the availability of water in the study catchment will not be affected by climate change impacts in the 2011-2025 time periods.

Figure 5 and 6 shows that the basin discharge and sediment yield increases in the month of July,

August, and September under the projected climate change impacts. Compared with the baseline run, the predicted sediment yield and flow in the month of August is the highest and increase by 1.21 million ton and 0.75 BCM, respectively. The substantial sediment yield and runoff (stream flow) increased for this month was attributable to the increased precipitation. In general, there was a close similarity between changes in total precipitation and changes in sediment yield and stream flow (Fig. 3, 5 and 6). An increase in precipitation amount was often accompanied by an increase in stream flow and sediment yield. Besides, rainfall and air temperature changes had a more predictable impact on stream flow than soil erosion.

Table 9 Simulated –average annual sediment yield, stream flow and their percentage change relative to the baseline run for the changed climate scenario (1996-2011)

Flow simulation Sediment yield Simulation Scenarios Precipitation

mm M3 S-1 BCM Change % Million ton Change %

Baseline run 1802 52.44 1.65 - 2.48 -

Base reference land use

map with 2011-2025

climate

1922 54.87

1.73 4.6

2.52 1.6

Figure 5. Average monthly basin discharge and Anomalies (relative to the baseline run) under climate change (2011-2025)


188

NCSTI-2012

Figure 6. Monthly sediment load and anomalies (relative to the baseline run) under climate change (2011-2025)

F. Combined Effects of Land Use and Climate Change on Soil Erosion and Stream Flow

Table 10 shows the annual mean sediment yield and stream flow simulated by SWAT under different land use and climate scenarios. The difference between the different scenarios such as S1 and S4, S1 and S6, and S1 and S8 indicated the combined effects of land use and climate change (Table 10). Trend based land use with climate change scenario (S4) can increase the flow by 5.2% while land use and climate change alone increases the flow by 1.73% and 4.6%, respectively. The

effect of climate change has contributed about 88.46% of the total change of stream flow from the combined effects whereas land use change accounted for 33% of it.. Besides, S4 increases sediment yield by 29% of which about 59% of total change comes from land use change whereas about 8.3 % of total change of sediment yield was accounted for climate change. Compared with other scenarios, S4 (trend with climate change scenario) would considerably affect the hydrological behavior of the Upper Gilgel Abbay catchment during 2011‐2025.

Flow simulation Sediment yield simulationScenarios Description of M3 s-1 BCM Change % Million ton Change %

S1 Current /baseline run 52.44 1.65 - 2.48 -

S2Base reference land use with climate change

54.87 1.73 4.6 2.52 1.6

S3Trend without climate change

53.35 1.68 1.73 2.9 17.2

S4 Trend with climate change 55.2 1.74 5.2 3.2 29

S5 Plan without climate change 51.89 1.64 -1.05 2.2 -11.29

S6 Plan with climate change 54.71 1.73 4.33 2.31 -6.8

S7Conservation without climate change

51.59 1.63 -1.62 1.98 -20.2

S8Conservation with climate change

53.6 1.69 2.14 2.14 -13.71

Table 10. Annual average soil erosion and stream flow in the catchment under different land use and climate


189

NCSTI-2012

Plan based land use with climate change (S6) will increases the flow by 4.33% and reduced sediment yield by 6.8% from the baseline run whereas S2 increases stream flow and sediment yield by 4.6% and 1.6%, respectively and S5 increases the flow by 1% and reduce sediment load by 10.3% (Table 10).Comparing S2 and S5 scenarios with S6 scenario, the projected climate change impacts on stream flow and sediment yield has reduced by 0.3% and 8.5%, respectively. Similarly, comparing S2 and S6 with S8 scenarios, the projected climate change impacts on stream flow and sediment yield has significantly reduced by 2.3% and 15%,

respectively. This implied that the implementation of plan and conservation

based land use scenarios in 2011-2025 period will offset the projected climate change impacts and can be thus used as an adaptive strategy from sediment yield and stream flow perspectives. On the other hand, the findings of this study demonstrate that the flow was more sensitive to climate changes than land use changes; however, soil erosion is likely more affected by land use changes rather than climate changes. These results are consistent with previous studies [23,9,25,26] .

IV. Conclusion SWAT model performance in the Upper Gilgel

Abbay Catchment in Ethiopia is acceptable in predicting flow and sediment load. The Nash Sutcliff efficiency (NSE), Coefficient of determination (R2) and average percentage difference for monthly flow was 0.92, 0.94 and -0.54% for calibration period and 0.87, 0.9 and -4.2 % for validation period, respectively. Similarly, for monthly sediment load about 0.86, 0.88 and 3.39% for calibration period and 0.81, 0.84 and -10.07% for validation period, respectively, were assessed. Besides, results of this study indicated that the SWAT model proved to be a power full tool to implement the on-going land use plannof the study catchment. The simulation results indicated that the combined effect of land use and climate change can considerably alter the hydrological behavior of the upper Gilgel Abbay catchment during the 2011-2025 period. This effect is highly magnified in trend based land use and climate change scenario condition. On the other hand, simulation of plan and conservation based land use with climate change scenarios revealed that the increased soil

erosion and stream flow caused by the projected Climate change will be reduced due to plan and conservation based the land use changes. This may indicates that plan and conservation based land use can be used as an adaptation measures for the projected climate change impacts. In general, in all scenario simulation, the flow is more sensitive to climate changes than land use changes. However, soil erosion is likely more affected by land use changes rather than climate changes, even though the combined effect of land use and climate change would affects both soil erosion and stream flow at large. Therefore, in the planning of adaptation strategies for climate and land use change impacts on catchment hydrology, it is necessary to consider the combined effects of climate and land use change rather than to simulate their impacts separately.

Acknowledgment The authors gratefully acknowledge the financial

support by the Horn Africa Regional Environmental center and Net work, and the field work assistance provided by Aksum and Arbaminch Universities in Ethiopia.

Reference 1. Klik,A.and Eitzinger,J..2010. Impact of climate change on soil erosion

and the efficiency of soil conservation practices in Austria. Journal of

Agricultural Science (2010), 148, 529–541. f Cambridge University

Press 2010 . 2. Setegn, Srinivasan,R. Melesse,M. Spatial delineation of soil erosion

vulnerability in the LakeTana Basin, Ethiopia Hydrol. Process. 23,

3738–3750 (2009) 3. Gebreyesus B. and Kirubel M. 2009. Estimating Soil Loss Using

Universal Soil Loss Equation (USLE) for Soil Conservation planning at

Medego Watershed, Northern Ethiopia. Journal of American Science

2009: 5(1), 58-69 4. Favis-Mortlock, D.T., Savabi, M.R., 1996. Shifts in rates and spatial

distributions of soil erosion and depositionunder climate change. In:

Anderson, M.G., Brooks, S.M. (Eds.), Advances in Hillslope Processes.

Wiley,Chichester, pp. 529– 560. 5. Williams, J.R., Nearing, M.A., Nicks, A., Skidmore, E., Valentine, C.,

King, K., Savabi, R., 1996. Using soil erosion models for global change

studies. Journal of Soil and Water Conservation 51 (5), 381– 385. 6. Favis-Mortlock, D.T., Guerra, A.J.T., 1999. The implications of general

circulation model estimates of rainfall for future erosion: a case study

from Brazil. Catena 37, 329– 354

7. Nearing, M.A., 2001. Potential changes in rainfall erosivity in the U.S.

with climate change during the 21st century. Journal of Soil and Water

Conservation 56 (3), 229–232 8. Pruski, F.F., Nearing, M.A., 2002a. Climate-induced changes in erosion

during the 21st century for eight U.S. locations. Water Resources

Research 1298. 9. M.A. Nearinga,*, V. Jettenb, C. Baffaut.2005. Modeling response of

soil erosion and runoff tochanges in precipitation and coverCatena 61

(2005) 131–154 10. Kim U, Kaluarachchi JJ, Smakhtin VU (2008). Climate change impacts

on hydrology and water resources of the Upper Blue Nile River Basin,

Ethiopia. Colombo, Sri Lanka: International Water Management

Institute. 27p (IWMI Research Report 126)

11. Abdo, K. S. 2008. Assessment of climate change impacts on the

hydrology of Gilgel Abay catchment in Lake Tana Basin, Ethiopia.

MSc Thesis in ITC 12. Mulugeta H. 2009 .Evaluation of Climate Change Impact on Upper

Blue Nile Basin Reservoir, (Case Study on Gilgel Abay Reservoir,

Ethiopia). MSc Thesis in Arbaminch University.


190

NCSTI-2012

13. Edwards, W.M., Owens, L.B. 1991. Large storm effects on total soil

erosion. Soil and Water Conservation 46, 75– 78 14. Emeru, W. 2009. Hydrological response to land covers change in

gilgel abbay catchment Ethiopia. MSC thesis in ITC Enschede 15. Sun, G., M. Riedel, R. Jackson, R. Kolka, D. Amatya, and J. Shepard.

2004. Chapter 3: Influences of management of Southern forests on

water quantity and quality. In Southern Forest Sciences: Past, Current,

and Future, 394. Gen. Tech. Rep/ SRS-75. H. M. Rauscher and K.

Johnsen, eds. Ashville, N.C.: U.S. Forest Service, Southern Research

Station 16. Skaggs, R. W., D. M. Amatya, G. M. Chescheir, C. D. Blanton, and J.

W. Gilliam. 2006. Effects of drainage and management practices on

hydrology of pine plantation. In Proc. Intl. Conf. on Hydrology and

Management of Forested Wetlands, 3‐14. St.Joseph, Mich.: ASABE 17. Sivrikaya, F., Cakir, G., Kadiogullari, A. I, Keles, S., BasKent, E.Z.,

and Terzioglu, S. 2007. Evaluating land use/land cover changes and

fragmentation in the Camili forest planning unit of Northeastern Turkey

from 1972 To 2005. Land degradation & development, 18: 383–396.

18. Huxman, T.E. et al, 2005. Ecohydrological implications of woody plant

encroachment. Ecology 86, 308–319 19. Calder, I.R., 1998. Water-resource and land use issues. SWIM Paper 3.

Colombo: IIMI. Huxman, T.E. et al, 2005. Ecohydrological

implications of woody plant encroachment. Ecology 86, 308–319. 20. Kim, U.; Kaluarachchi, J. J.; Smakhtin, V. U.

2008. Climate change impacts on hydrology and water resources of the Upper Blue Nile River Basin, Ethiopia. Colombo, Sri Lanka: International Water Management Institute. 27p (IWMI Research Report 126

21. Xu, C. 2000. Climate change and hydrological models: A review of

existing gaps and recent research developments. Water Resource

management 13, 369‐382 22. Arnold, J. G., and N. Fohrer. 2005. SWAT2000: Current capacities and

research opportunities in applied watershed modeling. Journal of

Hydrological Process 1 563‐572 9,23. Legesse, D., C. Vallet‐Coulomb, and F. Gasse. 2003. Hydrological

response of a catchment to climate and land use change in tropical

Africa: Case study south central Ethiopia. Journal of Hydrology 275 ,

67‐85 24. Nearing MA, Pruski FF, O‟Neal MR. 2004. Expected climate change

impacts on soil erosion rates: A review. Journal of Soil and Water

Conservation 59, 43–50 25. Hua Guo, Qi Hu, Tong Jiang .2008. Annual and seasonal stream flow

responses to climate and land-cover changes in the Poyang Lake basin,

China .Journal of Hydrology 355, 106– 122 26. Zhi Li, Wen-zhao Liu, Xun-chang Zhang, Fen-li Zheng. 2009. Impacts

of land use change and climate variability on hydrology in an

agricultural catchment on the Loess Plateau of China. Journal of

hydrology 377, 35–42 27. Poesen J, Nachtergaele J, Verstraeten G, Valentin C. 2003. Gully

erosion and environmental change: importance and research needs.

Catena 50, 91–133. 28. Boardman J. 2006. Soil erosion science: reflections on the limitations

of current approaches. Catena 68, 73–86. 29. Checkun T. 2002.Vetiver in the Rehabilitation of the Degraded ZegZeg

Watershed in Ethiopia, Addis Ababa, Ethiopia. Report, 131-136. 30. Arnold JG, Srinivason R, Muttiah R, Williams JR. 1998. Large area

hydrologic modeling and assessment part I: model development.

Journal of the American water Resources Association 34 , 73–89 31. Wischmeier WH, Smith DD. 1978 .Predicting rainfall erosion losses –

a guide for conservation planning. U.S. Department of Agriculture,

Agriculture Handbook 537, 20-152. 32. Bronsert A .2004. Rainfall-runoff modeling for assessing impacts of

climate and land-use change. Journal of Hydrological Process 18, 567-

570 33. Neitsch SL, Arnold JG, Kiniry JR, Srinivasan R, Williams JR. 2005.

Soil and Water Assessment Tool, Theoretical Documentation: Version

2005 .USDA Agricultural Research Service and Texas A&M Blackland

Research Center: Temple

34. Moriasi DN, Arnold JG, Van Liew MW, Binger RL, Harmel RD, Veith

T. 2007. Model evaluation guidelines for systematic quantification of

accuracy in watershed simulations. Transactions of the ASABE 50(3):

885–900. Multi-variable catchment models. Journal of Hydrology 324,

10–23. 35. Santhi, C., Arnold, J.G. Williams, J.R. Dugas, W.A., Srinivasan, R. and

Hauck, L.M. 2001. Validation of the SWAT Model on a Large River

Basin With Point and Non-Point Sources. Journal of American Water

Resources Assoc. 37, 1169-1188

Application of SWAT Model for assessment of best management practices (BMPs) on Soil...

191

NCSTI-2012

Application of SWAT Model for Assessment of Best Management Practices (Bmps) on Soil Erosion/Sedimentation in Gilgel Gibe

Basin Watershed-EthiopiaT.A.Demissie1, F. Saathoff2, A.Gebissa2, Y.Sileshi3

1 Department of Civil Engineering, Jimma Institute of Tehnology, Jimma University, Jimma, Ethiopia,[email protected]

2Institue for Environmental Engineering, University of Rostock, Rostock, Germany 3Civil engineering Department, Addis Ababa Institute of Technology, Addis Ababa University, Addis Ababa, Ethiopia

Abstract--Soil erosion/Sedimentation is an immense problem threatening the live storage capacity of dam reservoirs in Ethiopia. This in turn reduces the power generation capacities of hydropower reservoirs. Therefore, studies which give insight into soil erosion/sedimentation mechanisms and mitigation methods is important. The hydropower plants of Omo-Gibe basin namely Gilgel Gibe 1, and Gilgel Gibe 2 are cascade hydropower schemes with installed capacities of 180MW and 420MW respectively. The poor land management practices coupled with rugged topography of Gilgel Gibe watershed and the erosive rainfall in the area increases the rate of soil erosion/sedimentation and threats the lifespan of Gilgel Gibe 1 hydropower reservoir. The problem of sedimentation in Gilgel Gibe 1 will also affect Gilgel Gibe 2 as it largely uses the water released from Gilgel Gibe 1 for power generation. The sustainability of these hydropower plants needs catchment management intervention/practices that will reduce soil erosion/sedimentation problems. This paper presents the results of monthly and yearly sediment yield simulations experiments conducted for Gilgel Gibe 1 under different Best Management Practice (BMP) scenarios. The scenarios applied in this paper are (i) maintaining existing conditions, (ii) introducing filter strips, and (iii) applying stone/soil bunds iv)reforestation . The Soil and Water Assessment Tool (SWAT) was used to model soil erosion, identify soil erosion prone areas and assess the impact of BMPs on sediment reduction via simulations. The discharge data of 26years and available land use/land cover and soil data were effectively utilized. The model was calibrated for available daily and monthly observed discharge and resulted in satisfactory Nash-Sutcliffe efficiencies of 0.684 and 0.711 respectively. The correlation coefficients for the calibration process with daily and monthly data were found to be 0.726 and 0.751 respectively. The Validation of the model using daily and monthly discharge data also resulted in Nash-Sutcliffe efficiencies of 0.640 and 0.715 and correlation coefficients of 0.662 and 0.746 respectively. More over the PBIAS values calculated were -13.9% and -12.5% for daily and monthly calibration respectively. The PBIAS values for validation of the model were -5.2% and -5.3% for daily and monthly validation. The simulation results showed that applying filter strips, stone bunds and reforestation scenarios could reduce the current sediment yields at soil erosion prone areas and at the outlet of the catchment area which is the inlet to Gilgel Gibe 1 reservoir.

I. Introduction The Gilgel Gibe River is a right hand tributary of one of the eight major river basins in Ethiopia, the Omo-Gibe river basin. It is the major source of water for Gilgel Gibe dam reservoir project which has a live storage capacity of 657Mm3. But the storage volume of this reservoir is threatened by the soil erosion and subsequent sedimentation from the upstream of the Gilgel Gibe basin. Previous studies indicate that there is a rapid loss of storage volume due to excessive soil erosion and subsequent sedimentation in Gilgel Gibe 1 dam reservoir.[10] conducted a cross sectional study and assessed the siltation and nutrient enrichment level of Gilgel Gibe 1 dam reservoir. From their study they found that siltation and nutrient enrichment were the major problems in this reservoir.

In addition to Gilgel Gibe 1 hydropower plant, the power generation of the Cascade hydropower plant to Gilgel Gibe 1, namely Gilgel Gibe 2 Which has an installed capacity of 420MW and uses the water released from the same reservoir, will significantly be affected.

Ned

hi

Bulbul

Unta

Nad

aKallo

NadaGud

a

GilgelGibe

Jimma

Sekoru


192

NCSTI-2012 Fig.1. Location map of the Gilgel Gibe basin. Currently, the government of Ethiopia is constructing a huge hydropower plant, Gilgel Gibe 3, downstream of Gilgel Gibe 1 and 2. The Gilgel Gibe 3 dam and powerhouse are being built approximately 155km downstream of the Gilgel Gibe II plant. Up on its completion Gilgel Gibe 3 will have an installed capacity of 1870MW.There is also a plan to construct Gilgel Gibe 4 which will be the farthest downstream in the cascade. Though the Government of Ethiopia is putting an effort to construct large hydropower plants to satisfy the energy demand of the country, the rapid loss of storage volume due to sedimentation is major problem of all reservoirs.Some preliminary studies indicate that the levels of some reservoirs (e.g., Koka reservoir), lakes (e.g., Alemaya, Awassa, Abaya, and Langano) have decreased. The process is so challenging that the initial water carrying capacity the dams has reduced due to progressive silt accumulation. For example the Koka dam has accumulated about 3.5 million m3 of silt (or 2300 t.km-2) in just 23 years [15]. Thus, an insight into the soil erosion/sedimentation mechanisms and the mitigation measures plays an indispensable role for the sustainability of the existing reservoirs and newly planned projects.To develop effective soil erosion control plans and to achieve reductions in sedimentation, it is necessary to quantify the sediment yield and identify areas that are highly vulnerable to erosion. Literature review shows that there are many catchment models that include the soil erosion/sedimentation processes and simulate the effect of mitigation measures [33], [5].The range of models can be viewed in the way they represent the area to which they are applied; that is, whether the model considers processes and parameters to be lumped or distribute. With increasing computing power over the last two decades, distributed approaches have become more feasible. Distributed models reflect the spatial variability of processes and outputs in the catchment analysis. A distributed approach seems particularly applicable to sediment transport modelling [33].Some of the soil erosion models are AGNPS (Agricultural Non-point Source Pollution Model) [58]. ANSWERS (Areal Nonpoint Source Watershed Environmental Response Simulation)[4], CREAMS (Chemicals, Runoff, and Erosion from Agricultural Management Systems) [27] EPIC (Erosion Productivity Impact Calculator)[55], EROSION-3D[45], EUROSEM (European Soil Erosion Model)[37],SWAT (Soil and Water Assessment Tool) [1], WEPP (Water Erosion Prediction Project) [28], and so on. However, there are a few applications of erosion

modelling in Ethiopia and most of them concentrate on-Blue Nile basin. In the Blue Nile Basin[59] simulated soil loss in the Dembecha catchment using WEEP,[20] applied AGNPS and predicted sediment yield in Augucho catchment. The same AGNPS model was used by [35] to simulate sediment yield in the kori catchment.[22]- applied LISEM to simulate effect of reforestation on soil erosion in the kushet-Gobo Deguat catchment .[47] calibrated and validated a Simple soil erosion model in the Abbay(Upper Blue Nile)basin and obtained a reasonable result, and [46] applied SWAT for simulation of a sediment yield in the Anjeni gauged catchment and obtained quite acceptable result. SWAT has been successfully applied by different researchers in Ethiopia. Most of the SWAT model applications in Ethiopia concentrate on the Blue Nile river basin. For instance, [48] applied SWAT model to evaluate the effectiveness of different scenarios in reducing runoff, sediment and soil nutrient losses in northern Ethiopia. [3] applied the SWAT model to establish the spatial distribution of sediment yield and to test the potential of watershed management measures to reduce sediment loading from hot spot areas in Gumara watershed (Blue Nile) and [14] also applied the SWAT model to assess the impact of BMPs on sediment reductions in the Upper Blue Nile River Basin. Though the SWAT model is widely applied in Ethiopia, particularly on the Blue Nile river basin, there is no literature that indicates the SWAT model application on Omo-Gibe basin in general and Gilgel Gibe 1 basin in particular. In this study, the SWAT model will be applied to the Gilgel Gibe river basin with specific focus on BMPs application. Currently, there is recommendation to protect the buffer zone around Gilgel Gibe 1 dam reservoir from agricultural practices. In addition to the buffer zone protection, the Oromiya environmental protection bureau is also implementing watershed development through the community based participatory approach of [38] management practices to reduce soil erosion and conserve soil and water under its basin development programme. Such basin development programme should be aided by powerful modelling tools such as SWAT. Therefore, the objective of this study is to model the spatially distributed soil erosion/sedimentation process in the Gilgel Gibe basin at monthly and yearly time step and assess the impact of different catchment management interventions applied on hot spot areas on sediment yield. A brief description of the Gilgel Gibe basin is given in the next section, followed by a discussion on the methodology used. The third section presents the model results and discussion of different land management scenarios. Finally, the


193

NCSTI-2012conclusion summarizes the main findings of the investigations.

II. Description of the study area The Gilgel Gibe project is situated in the south-western part of Ethiopia, in Oromia Regional state. The project is purely a hydropower scheme, with an installed capacity of 180Mw, aimed to increase energy and power supply to the national grid. The reservoir has a live storage capacity of 657Mm3. The Gilgel Gibe River is a right hand tributary of one of the 8 major river basins in Ethiopia, the Omo-Gibe river basin. The catchment area of the Gilgel Gibe basin is about 5125Km2 at its confluence with the great Gibe River and about 4337km2 at the dam site. The basin is generally characterized by high relief hills and mountains with an average elevation of about 1700m above mean sea level. The basin is largely comprises of cultivated land. In general terms, the Gilgel Gibe basin is characterized by wet climate with an average annual rainfall of about 1550mm and average temperature of 190c. The seasonal rainfall distribution takes a uni-modal pattern with maximum during summer and minimum during winter, influenced by the inter-tropical convergence zone (ITCZ).

III. Methodology A.SWAT model description The Soil and Water Assessment Tool (SWAT) is a physical process based model to simulate continuous –time landscape processes at a catchment scale [1], [40]. The catchment is divided into hydrological response units (HRU) based on soil type, land use and slope classes. The major model components include hydrology, weather, soil erosion, nutrients, soil temperature, crop growth, pesticides agricultural management and stream routing. The model predicts the hydrology at each HRU using the water balance equation, which includes daily precipitation, runoff, evapotranspiration, percolation and return flow components. The SWAT model has two options for computing surface runoff: (i) the Natural Resources Conservation Service Curve Number (CN) method [49] or (ii) the Green and Ampt method [17]. The flow routing in the river channels is computed using the variable storage coefficient method [53], or Muskingum method [8]. SWAT includes three methods for estimating potential evapotranspiration: (i) Priestley-Taylor [42] (ii) Penman-Monteith [36] and (iii) Hargreaves [21]. The SWAT model employs the Modified Universal Equations (MUSLE) to compute

HRUs level soil erosion. It uses runoff energy to detach and transport sediment [56]. The sediment routing in the channel [2] consists of channel degradation using stream power [54] and deposition in channel using fall velocity. Channel degradation adjusted using USLE soil erodibility and channel cover factors.

B. SWAT model setup SWAT model inputs are Digital Elevation Model (DEM), land use map, soil map, and weather data. There is a considerable amount of data available on the web, and Map window SWAT used in this study used this advantage. Map Window SWAT (MWSWAT) is delivered along with the following data [9].DEM maps: SRTM project [26] ,Land: Global Land Cover characterization [19], Soil maps: FAO [12].The Step by Step Geo-Processing &Set-up an the Map window interface for SWAT(MWSWAT) document s [29] and [30], has been followed to extract the required watershed data and to set up the SWAT model for Gilgel Gibe basin. The DEM was used to delineate the catchment and provide topographic parameters such as overland slope and slope length for each subbasin. The catchment area of the Gilgel Gibe was delineated and discretized into 51 subbasins using a 90m DEM (http://srtm.csi.cgiar.org) through an MWSWAT interface. The Land use data which has been constructed from the USGS Global Land Cover Characterization (GLCC) database (http://edcsns17.cr.usgs.gov/glcc/glcc.html), by Abbaspour is used. This map has a spatial resolution of 1 km and 24 classes of land use representation. The parameterization of the land use classes (e.g. leaf area index, maximum stomatal conductance, and maximum root depth, optimal and minimum temperature for plant growth) is based on the available SWAT land use classes. The land cover classes derived are Dryland Cropland and pasture(CRDY),36.68%, Grassland(Gras),15.56%, Savanna(SAVA) 14.45%, Evergreen Forest(FOEB)22.65%, Mixed Forest(FOMI) 9.92% and Cropland/woodland mosaic(CRWO),0.74%. The Soil map was produced by the Food and Agriculture Organization of the United Nations [11]. Almost 5000 soil types at a spatial resolution of 10 kilometers with soil properties for two layers (0-30 cm and 30-100 cm depth) are provided. Further soil properties (e.g. particle-size distribution, bulk density, organic carbon content, available water capacity, and saturated hydraulic conductivity) were obtained from [44].The soil data is also available from the WaterBase web site http://www.waterbase.org/download_data.html, and was extracted for the study area. FAO soil and the slope class


194

NCSTI-2012maps were overlaid together to derive 410 unique HRUs. Although the SWAT model provides an option to reduce the number of HRUs in order to enhance the computation time required for the simulation, we considered all of the HRUs with landuse of Dryland, Cropland and pasture of to evaluate the management intervention impact. The daily precipitation, maximum and minimum temperature, wind speed, average relative humidity data from Jimma and Sekoru stations were used to run the model. In addition as Jimma and Sekoru meteorological stations have daily data on duration of sunshine hours, the Angstrom formula which relates solar radiation to extraterrestrial radiation and relative sunshine duration is used to estimate the daily solar radiation to be used in the model. The missed sunshine data were filled by non-linear regression analysis with other auxiliary climate variables such as relative humidity and temperature data before it was used to estimate solar radiation. The solar radiation data is required by SWAT and if not supplied SWAT generates this data. Though all the daily weather data which are required to run the model have been supplied, the weather generator file was also prepared using the 20 years daily data from these two stations and included in the project file. Daily river flow data measured at Asendabo gauging station was used for model calibration and validation. The flow observations were available throughout the year, but the sediment concentration data was not available for Gilgel Gibe basin. The model was run using daily data of 26 years. The daily meteorological data from 1980 to 2005 was used to run the model. The three years data from 1980 to 1982 was used to warm up the model. Whereas, the data from 1983 to 1992 was used to calibrate the model and the data from 1993 to 2000 was used to validate the model. The modeling period selection considered discharge data quality and availability. A daily flow was used to calibrate and validate the model at Asendabo gauging station and sediment discharge was simulated at the outlet of the Gilgel Gibe watershed which is in turn an inlet to the Gilgel gibe 1 hydropower reservoir. Sensitivity analysis was carried out to identify the most sensitive parameters for model calibration using One-factor-At-a-Time (LH-OAT), an automatic sensitivity analysis tool implemented in SWAT 2005. SWAT 2005 editor is used to read the project database generated by

Map window SWAT interface to edit SWAT input files, execute SWAT, and perform sensitivity, auto calibration and uncertainty analysis. Based on the sensitivity analysis results, we identified 8 parameters of interest for this basin. We started with all 27 hydrological flow-related parameters and ranked by their order of sensitivity in simulating the basin hydrology. It resulted in about 8 parameters as the most sensitive ones for this basin. Followed by the sensitivity analysis, the most sensitive parameters were calibrated by both manual calibration (expert) and automatic calibration. Appropriate lower and upper ranges in parameter values have been assigned prior to initiating the auto calibration process.

C. Model performance Evaluation. Model evaluation is an essential measure to verify the robustness of the model. In this study, the following methods were used (i) Nash-Sutcliffe efficiency (NSE), (ii) percent bias (PBIAS), and (iii) correlation between observed and simulated flows. The Nash-Sutcliffe efficiency (NSE) is computed as the ratio of residual variance to measured data variances [39]. The NSE simulation coefficient indicates how well the plot of observed versus simulated values fits the 1:1 line. The Nash-Sutcliffe is calculated using Eq.

(1)

n

i

meanobsi

n

i

simi

obsi

XX

XXNSE

1

2

1

2

)(

)(1 (1)

Where: - observed stream flow in m3/s obs

iX

- simulated stream flow in m3/s simiX

- meanX mean of n values.

- mean of simulated values simmeanX

obsmeanX mean of observed values

- =number of observations The NSE can range from n to +1, with 1 being a perfect agreement between the

model and real (observed) data. The simulation results were

considered to be good if NSE≥0.75, and satisfactory if 0.36≤NSE≤0.75 [51].The percent bias (PBIAS) measure the average tendency of the simulated data to be larger or smaller than their observed counterparts. A positive value indicates a model bias toward underestimation, whereas a negative value indicates a bias toward overestimation

[18].The PBIAS< ±25% is satisfactory [50]. The PBIAS is calculated with eq.2.

n

i

obsi

n

i

simi

obsi

X

XXPBIAS

1

1

)(

100)( (2)


195

NCSTI-2012

The coefficient of determination R2 value is an indicator of the strength of the linear relationship between the observed and simulated values. It ranges from 0.0 to 1.0, with higher values indicating better agreement. The R2 is calculated with eq.3.

n

i

n

i

obsmean

obsi

simmean

simi

n

i

obsmean

obsi

simmean

simi

XXXX

XXXXR

1 1

22

2

12

)()(

)(( (3)

D. Catchment management intervention scenarios Agricultural conservation practices, often called best management practices or BMPs, are widely used as effective measures for preventing or minimizing pollution from nonpoint sources within agricultural watersheds. SWAT already has an established method for modeling several agricultural practices including changes in fertilizer and pesticide application, tillage operations, crop rotation, dams, wetlands, and ponds. The model also has the capacity to represent many other commonly used practices in agricultural fields through alteration of its input parameters [6]. Ten important agricultural conservation practices were selected for representation with the SWAT2005 model and a number of previous modeling studies have used SWAT to evaluate conservation practices around the globe [31]. However, selection of BMPs and their parameter values are site specific and should reflect the study area reality [14]. For this study, we selected BMPs based on the previous traditional soil and water conservation practices on Ethiopian highlands. Currently, some of these practices are largely under implementation through community based participatory watershed development of [38], Ethiopia. The baseline values for the input parameters could be selected by (i) a model calibration procedure; or (ii) a ‘suggested’ value obtained from the literature, previous studies in the study area, or prior experience of the analyst [31]. For this study the baseline values which will represent the basin existing condition (Scenario 0) for the input parameters have been selected based on the suggested value obtained from the literature. For scenarios 1 and Scenario 2 the BMPs were represented in SWAT model by modifying the SWAT parameters to reflect the effect the practice has on the processes simulated within SWAT [6].The scenarios simulated and representation of BMPs in the SWAT are depicted in Table 2.

Table 2: Scenario description and SWAT parameters used to represent BMPs. *The average values taken from Community Based Participatory Watershed Development Guideline. **the calibration value for discharge is maintained.***Min: minimum value of SLSUBBSN in SWAT model.****assigned by SWAT model. In scenario 1, filter strips were placed on all Dry land, Cropland and Pasture (CRDY), all soil types and slope classes. The effect of filter strip is to reduce sediment, dissolved contaminants, and sediment adsorbed organics in runoff [43]. Appropriate model parameter for representation of the effect of filter strips is width of filter strip (FILTERW). The filter width value, FILTERW, of 1m was assigned to simulate the impact of filter strips on sediment trapping. The FILTERW value was assigned based on local research experiences in the Ethiopian highlands [25], [23].In scenario 2, stone/soil bunds were placed on all Dry land Cropland and Pasture (CRDY), all soil types and slope classes. This practice has a function to reduce overland flow, sheet erosion and reduce slope length [6]. This BMP was selected as it was the most widely and most intensively used soil conservation practice in the area [57]. Appropriate parameters for representing the effect of stone bunds are the Curve Number (CN2), average slope length (SLSUBBSN) and the USLE_P support practice factor (USLE_P).The SWAT assigned value of the USLE_P value of 1.0 is used prior to the application of BMPs. The modified value/Post-BMP value for USLE_P of 0.5 was assigned based on [25] being the P factor recommended for all types of bunds in Ethiopia. The average slope length (SLSUBBSN) for slopes 0-10% and 10-20% is taken from the community based participatory watershed development guideline which is currently under implementation in Ethiopian highlands. The minimum acceptable SLSUBBSN by SWAT is model 10m and this

SWAT parameter used Scenarios

Description of BMP

Parameter name

input file PRE BMP/Calibration value

Post-BMP/Modified value

Scenario 0

baseline

Scenario 1

Filter strip

FILTERW

0 0 1m

Scenario 2

Stone/soil bund

SLSUBBSN CN2 USLE_P

0-10% 10-20% 20-260%

30m 30m 30m ** 1.0

17.5m* 11m* 10min***

** 0.5

Scenario 3

Reforestation

**** ****

Scenario 4

Further subdivision

Similar to the previous scenarios

0-5% 5-10% 10-20% 20-260%



maps were overlaid together to derive 410 unique HRUs. Although the SWAT model provides an option to reduce the number of HRUs in order to enhance the computation time required for the simulation, we considered all of the HRUs with landuse of Dryland, Cropland and pasture of to evaluate the management intervention impact. The daily precipitation, maximum and minimum temperature, wind speed, average relative humidity data from Jimma and Sekoru stations were used to run the model. In addition as Jimma and Sekoru meteorological stations have daily data on duration of sunshine hours, the Angstrom formula which relates solar radiation to extraterrestrial radiation and relative sunshine duration is used to estimate the daily solar radiation to be used in the model. The missed sunshine data were filled by non-linear regression analysis with other auxiliary climate variables such as relative humidity and temperature data before it was used to estimate solar radiation. The solar radiation data is required by SWAT and if not supplied SWAT generates this data. Though all the daily weather data which are required to run the model have been supplied, the weather generator file was also prepared using the 20 years daily data from these two stations and included in the project file. Daily river flow data measured at Asendabo gauging station was used for model calibration and validation. The flow observations were available throughout the year, but the sediment concentration data was not available for Gilgel Gibe basin. The model was run using daily data of 26 years. The daily meteorological data from 1980 to 2005 was used to run the model. The three years data from 1980 to 1982 was used to warm up the model. Whereas, the data from 1983 to 1992 was used to calibrate the model and the data from 1993 to 2000 was used to validate the model. The modeling period selection considered discharge data quality and availability. A daily flow was used to calibrate and validate the model at Asendabo gauging station and sediment discharge was simulated at the outlet of the Gilgel Gibe watershed which is in turn an inlet to the Gilgel gibe 1 hydropower reservoir. Sensitivity analysis was carried out to identify the most sensitive parameters for model calibration using One-factor-At-a-Time (LH-OAT), an automatic sensitivity analysis tool implemented in SWAT 2005. SWAT 2005 editor is used to read the project database generated by

Map window SWAT interface to edit SWAT input files, execute SWAT, and perform sensitivity, auto calibration and uncertainty analysis. Based on the sensitivity analysis results, we identified 8 parameters of interest for this basin. We started with all 27 hydrological flow-related parameters and ranked by their order of sensitivity in simulating the basin hydrology. It resulted in about 8 parameters as the most sensitive ones for this basin. Followed by the sensitivity analysis, the most sensitive parameters were calibrated by both manual calibration (expert) and automatic calibration. Appropriate lower and upper ranges in parameter values have been assigned prior to initiating the auto calibration process.

C. Model performance Evaluation. Model evaluation is an essential measure to verify the robustness of the model. In this study, the following methods were used (i) Nash-Sutcliffe efficiency (NSE), (ii) percent bias (PBIAS), and (iii) correlation between observed and simulated flows. The Nash-Sutcliffe efficiency (NSE) is computed as the ratio of residual variance to measured data variances [39]. The NSE simulation coefficient indicates how well the plot of observed versus simulated values fits the 1:1 line. The Nash-Sutcliffe is calculated using Eq.

(1)

n

i

meanobsi

n

i

simi

obsi

XX

XXNSE

1

2

1

2

)(

)(1 (1)

Where: - observed stream flow in m3/s obs

iX

- simulated stream flow in m3/s simiX

- meanX mean of n values.

- mean of simulated values simmeanX

obsmeanX mean of observed values

- =number of observations The NSE can range from n to +1, with 1 being a perfect agreement between the

model and real (observed) data. The simulation results were

considered to be good if NSE≥0.75, and satisfactory if 0.36≤NSE≤0.75 [51].The percent bias (PBIAS) measure the average tendency of the simulated data to be larger or smaller than their observed counterparts. A positive value indicates a model bias toward underestimation, whereas a negative value indicates a bias toward overestimation

[18].The PBIAS< ±25% is satisfactory [50]. The PBIAS is calculated with eq.2.

n

i

obsi

n

i

simi

obsi

X

XXPBIAS

1

1

)(

100)( (2)


196

NCSTI-2012value is assigned for slopes greater than 20%.In Scenario 3, we simulated the impact of reforestation on sheet erosion. The reforestation has a function to reduce over land flow and rainfall erosivity .The reforestation effect was simulated by introducing land use change. Thus we replaced 10% of the area occupied by Dry Land Cropland and Pasture (CRDY) in to Evergreen forest.

IV. Results and discussion A. Model calibration and validation. The most sensitive parameters for flow predictions were curve number(CN2), baseflow alpha factor(ALPHA_BF), groundwater delay time (GW_DELAY), ground water ”re-vap” co-efficient (GW_REVAP), threshold water depth in the shallow aquifer for “revap” (REVAPMN), soil evaporation compensation factor(ESCO), available water capacity (SOL_AWC) and maximum canopy storage (CANMX). These flow parameters were adjusted within the given limits to initiate auto calibration. As measured data is not available on sediment yield, only the

modeled data has been used to identify the impact of adjusting a parameter value on some measure of simulated sediment output. Accordingly most sensitive parameters ranked 1 to 3 were USLE support practice factor(USLE_P), USLE land cover factor(USLE_C), and Ch_K2 respectively. The parameters channel cover factor (Ch-Cov), channel erodibility factor(Ch-erod), exponent of re-entrainment parameter for channel sediment routing (spexp) and linear re-entrainment parameter for channel sediment routing (spcon) were equally important with rank 8. The SWAT flow predictions were calibrated against daily and monthly average flows with a warm up period of 3 years from 1983 to 1992 and validated from 1993 to 2002 at Asendabo gauging station, as shown in fig. 2. The simulated daily flow matched the observed values for calibration period with NSE, PBIAS and R2 equal to 0.684, -13.9% and 0.726 respectively. For the validation period, and the observed daily flows showed acceptable agreement as indicated by NSE, PBIAS and R2 values equal to 0.640, -5.2% and 0.662 respectively.

.

Fig.2. observed and simulated daily hydrographs at Asendabo Station,calibration(top),validation(bottom).


197

NCSTI-2012

Fig.3. observed and simulated monthly hydrographs at Asendabo Station,calibration(top),validation(bottom). The simulated monthly average flow values also matched the observed values for calibration period with NSE and R2 values equal to 0.54 and 0.886 respectively. The calibration parameters were checked for the validation period and found to be 0.629 and 0.696 for NSE and R2 respectively. Table .3.SWAT sensitive parameters and fitted values.

parameter name Description parameter value

m-CN2.mgt* a-ALPHA_BF.gw** r-GW_DELAY.gw r-GW_REVAP.gw r-REVAPMN.gw r-ESCO.hru m-SOL_AWC.sol a-CANMX.hru

Curve number Baseflow alpha factor Groundwater delay time Groundwater revap co-efficient Threshold water depth in the shallow aquifer for revap Soil evaporation compensation factor Available water capacity Maximum canopy storage

0.8 0.302 45 0.20 0.15 0.25 1.67 4

*The extension (e.g.,.mgt) refers to the SWAT input file where the parameter occurs. **The qualifier (a-)refers to the substitution of a parameter by adding the parameter values indicated in table 3 and (m-)refers to the relative change in the parameter where the value from the SWAT database is multiplied by the values in the above table. And (r-) refers to

replacement in the parameter from the SWAT database by the values indicated in the table. The model simulated well the discharge on the rising limb of the hydrograph. While, the falling limb of the hydrograph indicated that the simulated discharge is slightly greater than the observed discharge data for the whole calibration and validation period, and the crest segment of the hydrograph show the simulated peak discharge to be slightly less than the observed peak discharge. Generally, as it was shown by model performance evaluation criteria, the SWAT model performed well in simulating stream flow hydrograph for this study. Besides, the performance of SWAT model, [41], indicated that the SWAT model can satisfactorily estimate sediment yield for even poorly gauged catchments of East African countries.

B. scenario analysis. The assessment of the spatial variability of soil erosion is useful for catchment management planning [14].The soil erosion prone areas in the Gilgel Gibe 1 basin are shown in Fig.5. The SWAT model simulation shows erosion extent varies from negligible erosion to 39 t/ha. Based on the classification of erosion rates in the Ethiopian highlands [24] the erosion level which are classified as high and very high in sub-basin 1, 3, 5 and 8 of Gilgel Gibe 1 basin corresponds to moderate erosion level (20-70t/ha/yr). The erosion level in the sub basin 1, 3, 5, and


198

NCSTI-2012

8 is in the range of 20 t/ha to the maximum value of 39 t/ha. The erosion level which are indicated as medium in sub-basins 2, 38 and 46 are relative to remaining sub-basins and their erosion level is in the range of 10 to 20t/ha. Generally, the SWAT simulation results for Gilgel Gibe 1 basin indicate that the sub-basins 1,2,3,5,8,38 and 46 have the high rate of erosion relative to the remaining sub-basins. The subbasins with high rate of erosion have a maximum percentage of nearly 60% landuse of Dry land Cropland and Pasture (CRDY) while the subbasins with very low soil erosion rate have got 0-9% Dry land, Cropland and Pasture as their landuse. These simulation results show the relative variations of soil erosion level within a sub-basin. These results are helpful to prioritize BMPs implementation area. Moreover, these results showed that the sediment yield to Gilgel Gibe 1 reservoir is mainly from sub-basins of the tributaries of Nedhi, and Bulbul which are to the left side of Gilgel Gibe River and at a close proximity to the reservoir.

±

0 19,000 38,0009,500 Meters

LegendSYLDsc0

<all other values>

ErosionHigh

Low

Medium

V.High

Fig.4. Erosion prone areas in Gilgel Gibe basin.

The SWAT model simulation for the existing condition predicted the sediment yield at the outlet of Gilgel Gibe 1

basin, which is an inlet to Gilgel Gibe dam reservoir to be 122.73x103t/yr. However, running the model with different catchment management scenarios provided interesting results. The simulation of filter strips scenario reduced the total sediment yield to 79.82x103t/yr from current condition at the same outlet location, which is equivalent to 35% reduction. The simulation of stone/soil bunds reduced the sediment yield to 23.26x103t/yr from the current conditions, which is equivalent to 81% reduction. This result is comparable to the results reported in the literature. [23] Reported 72-100% sediment yield reductions by stone bunds at plot scale in Ethiopian and the Eritrean highlands. The simulation of reforestation scenario (Scenario 3) showed the average reduction of sediment yield by 9.1% for subbasins 1,3,5,8,2,38 and 46 from the current condition. This less sediment reduction under scenario 3 as compared to scenario 1 and 2 could be attributed to smaller implementation area. The average sediment reduction at subbasin level where the subbasin has got dry land crop land and pasture(CRDY) greater than 10% of its total area under filter strip scenario was 35%.This is comparable with results reported by[14].[14], reported the sediment reductions under filter strip scenario ranged from 29% to 68%. In this study, the percentage sediment yield reduction per ha at subbasin level increased with an increase in the percentage area of CRDY which was provided with filter strip width of 1m. The subbasins such as subbasins 15,16,23,48 and 50 with percentage area of Dry land, Cropland and Pasture (CRDY) less than 10% showed the sediment yield reduction efficiency of 0% under filter strip scenario. The effectiveness of BMPs per hectare for the sub-basins with different percentage of Dry land, Cropland and Pasture (CRDY) is shown in fig.5.

Fig.5.Percentage of CRDY in the subbasin and sediment reduction efficiency of scenarios 1 and 2

V. Conclusions and recommendation The SWAT model was applied to assess the impact of the three Best Management Practices (BMPs) scenarios on sediment reduction in the Gilgel Gibe river basin. The


199

NCSTI-2012

impact of further subdivision of the subbasins in to more number of HRUs on the effectiveness of BMPs on sediment reduction was also checked. The model showed that the erosion prone areas at subbasin level, which is useful information for catchment management planning and for the implementation of the watershed development programme. The watershed development programme through community based participatory approach under implementation throughout the country. This study result showed that the implementation of the three BMPs could reduce the soil erosion and sediment yield at the subbasin and basin level. One of the three BMPs, namely soil/stone bunds has been practiced and implemented in some of the districts in the study watershed. The same practice is widely under implementation in Gilgel Gibe basin as per the decision made by the Ethiopian Government to promote and expand community watershed development in the country. However, the cost of implementing the BMPs should be evaluated. Additional BMPs should also be investigated and the best ones combined to form other scenarios which reduce soil erosion and sedimentation. This study shows as the modeling approach could be helpful for decision makers to prioritize the areas of intervention. In order to obtain a better estimate of the effectiveness of the filter strips, further investigation should be undertaken by using the improved vegetative filter strip (VFS) sub-model of SWAT2009 version. Furthermore, SWAT2009_LUC, a tool to activate the land use change module in SWAT2009 should be applied for further investigations. REFERENCES [1] Arnold, J.G.,Srinivasan, R.,Muttiah, R. S., and Williams, J.R.:Large area hydrologic Modelling and assessment Part I: Model development J.Am.Water Resour. As.,vol.34, pp.73-89,1998.

[2] Arnold,J.G.,Williams,J.R., and Maidment,D.R.: continuous time water and sediment-routing model for large basins, J.Hydraul.Eng- ASCE,vol.121,pp.171-183,1995.

[3] Asres,M.T. and S.B. Awulachew.: SWAT based runoff and sediment yield modeling, a case study of the Gumera watershed in the Blue Nile Basin, Ecohydrology and Hydrobiology,vol 10, pp.191-200,2010

[4] Beasley, D.B. and Huggins, L.F. 1982. ANSWERS (Areal Nonpoint Source Watershed Environ-ment Response Simulation): User's Manual, U.S. Environmental Protection Agency, Chi-cago, Illions.

[5] Borah, D.K. and Bera,M.: Watershed-scale hydrologic and nonpoint-source pollution models: Review of mathematical bases,T.ASAE,vol46,pp.1553-1566,2003

[6] Bracmort,K.,Arabi,M.,Frankenberger,J.,Engel,B.,and Arnold J.: Modeling long-term water quality impact of structural BMPs, T.ASABE,vol49,pp.367-374,2006.

[7]. Chekol, D.A., Tischbein, B., Eggers, H. and Vlek, P. (2007). Application of SWAT for Assessment

of Spatial Distribution of Water Resources and Analyzing Impact of Different Land Management

Practices on Soil Erosion in Upper Awash River Basin Watershed, FWU Water Resources Publications. Vol 06,ISSN No. 1613-1045. pp 110-117, 2007.

[8] Chow, V.T.: Open channel Hydraulics, McGraw-Hill Book company, New York,1959.

[9]. C.George and L.F.Leon.:WaterBase:SWAT in an Open source GIS, The Open Hydrology Journal,vol2,pp.1-6,2008

[10]. Devi et al., Assessment of siltation and nutrient enrichment of Gilgel Gibe dam, Southwest Ethiopia. Bio resource Technology vol.99,pp.975–979, 2008.

[11]. FAO:Digital soil map of the World and derived soil properties Food and Agricultural Organization of the United Nations,Rome,1995.

[12]. FAO/UNESCO. Digital Soil Map of the World and Derived Soil Properties. Rev. 1. (CD Rom), 2003. Available from http://www.fao.org/catalog/what_new-e.htm

[13]. Gessese,A. and Yonas M. Prediction of Sediment inflow to Legedadi Reservoir Using SWAT watershed and CCHEID Sediment Trasnport Models, Nile Basin Water Engineering Scientific Magazine,vol1,pp.65-74,2008.

[14]. G.D.Betrie, Y.A.Mohamed, A. van Griensven, and R.Srinivasan.: Sediment management modeling in the Blue Nile Basin using SWAT model, Hydrol. Earth Syst. Sci.,vol.15,pp. 807-818,2011.

[15]. Gizaw et al.,2004.E. Gizaw, W. Legesse, A. Haddis, D. Deboch and W. Birke, Assessment of factors contributing to eutrophication of Abasamuel water reservoir in Addis Ababa Ethiopia, Journal of Health Science vol.14 (2) pp. 112–113.

[16] Global Land Cover Characterization (GLCC):http://edesns17.cr.usgs.gov/glcc/glcc.html.

[17]. Green, W.H.and Ampt,C.A.: Studies on Soil physics: I.Flow of air and water through soils, J.Agr. Sci.,vol4,pp.1-24,1911

[18]. Gupta,H.,Sorooshian,S., and Yapo,P.: Status of automatic calibration for hydrologic models: Comparison with multilevel expert calibration, J.Hydrol.Eng.vol4,pp.135-143,1999.

[19]. Hansen, M., R. DeFries, J. Townshend, and R. Sohlberg. 1 Km Land Cover Classification Derived from AVHRR, 1998. http://glcf.umiacs.umd.edu/data/landcover.

[20]. Haregeweyn, N.and Yohannes,F.:Testing and evaluation of the agricultural non-point source pollution model(AGNPS) on Augucho catchment, western Hararghe,Ethiopia,Agr.Ecosyst.Environ.,vol99,pp.201-212,2003.

[21]. Hargreaves , G., Hargreaves,G.,and Riley,J.: Agricultural benefits for Senegal River Basin, J.irrig. Drain. E-ASCE, vol.111,pp.113-124,1985.

[22]. Hengsdijk,H.,Meijerink,G., and Mosugu,M.: Modeling the effect of three soil and water conservation practices in Tigray,Ethiopia, Agr.ecosyst.Environ.vol.105,pp.29-40,2005.

[23]. Herweg,K. and Ludi,E.: the performance of Selected soil and water conservation measures –case studies from Ethiopia and Eritrea, Catena,vol.36,pp.99-114,1999.

[24] Hurni,H.: Soil erosion and soil formation in agricultural ecosystems: Ethiopia and Northern Thailand, Mt.Res.Dev.vol.3,pp.131-142,1983.

[25] Hurni, H.: Erosion-Productivity-Conservation systems in Ethiopia, in: Proceedings of the 4th International

conference on Soil Conservation, Maracay, Venezula, pp.654-674, 1985.

[26] Jarvis, A., H.I. Reuter, A. Nelson, E. Guevara, 2008, Hole-filled Seamless SRTM data Version 4,


200

NCSTI-2012International Centre for Tropical Agriculuture(CIAT), available from (http://srtm.csi.cgiar.org).

[27] Knisel, W.G.. CREAMS: A field scale model for chemicals, runoff, and erosion from agri-cultural management systems. USDA Conservation Research Report; No. 26. Washing-ton, D.C. ,1980.

[28] .Laflen, J.M., Lane, J.L., Foster, G.R. 1991. WEPP (The water erosion prediction project) - a new generation of erosion prediction technology. J. Soil and Water Conserve .vol.46(1): pp.34–38.

[29] Luis F.Leon.:Step by Step Geo-Processing and Set-up of the Required Watershed Data for MWSWAT(MapWindow SWAT), Version 2,2011.

[30] Luis F.Leon.:Map window interface for SWAT(MWSWAT)version 1.8,2010.

[31] Mazdak Arabi, Jane R. Frankenberger, Bernie A. Engel2 and Jeff G. Arnold.: Representation of agricultural conservation practices with SWAT, Hydrol. Process. DOI: 10.1002/hyp (2007)

[32] Mekonnen,M.A., A.Worman, B.Dargahi and A.Gebeyehu.: Hydrological Modelling of Ethiopian catchments using limited data, Hydrol . Process.,vol.23,pp.3401-3408,2009.

[33] Merrit,W.S.,Letcher,R.A., and Jakeman,A.J.: A review of erosion and sediment transport models,Environ.Modell.softw.vol.18,pp.761-799,2003

[34]. Michael W. Van Liew and Jurgen Garbrecht.:Hydrologic Simulation of The Little Washita River,Journal of The American Water resou. Ass.,39,413-426,2007.

[35] Mohammed,H.,Yohannes,F., and Zeleke,G.:Validation of agricultural non-point source(AGNPS) pollution model in Kori watershed,South Wollo,Ethiopia, Int.J. Appl.Earth.Obs., 6, 97-109,2004.

[36] Monteith, J.L.: Evaporation and environment, symp.soc.Exp.Biol.,19 205-234,1965.

[37] Morgan, R.P.C., Quinton, J.N., Smith, R.E., Govers, G., Poesen, J.W.A., Auerswald, K., Chisci, G., Torri, D., Styczen, M.E. 1998. The European Soil Erosion Model (EUROSEM): a dynamic approach for predicting sediment transport from fields and small catchments. Earth Surface Processes and Landforms 23(6): 527 – 544.

[38] .MoARD.:Community Based Participatory Watershed Development,Part 1 A Guideline,Addis Ababa,Ethiopia, 2005

[39] Nash,J.E. and Sutcliffe,J.V.: River floe forecasting through conceptual models part I-A discussion of principles, Journal of Hydrolo.vol.10,pp.282-290,1970.

[40] Neitsch, S.L.,Arnold, J.G.,Kiniry,J., and Williams,J.R.: Soil and water assessment tool theoretical documentation(version 2005), USDA Agricultural Research Services and Texas A &M Blackland Research Center , Temple,Texas,2005.

[41] P.M.Ndomba and Ann van Griensven.: Suitability of SWAT Model for Sediment Yields Modeling in the eastern Africa, Advances in Data, Methods, Models and Their Applications in Geoscience.

[42]. Priestley, C. and Taylor,R.: On the assessment of surface heat flux and evaporation using large-scale parameters, Mon. Weather Rev.vol.100,pp.81-92,1972.

[43]. P. Tuppad, N.K.R. Srinivasan, C.G.Rossi, J.G.Arnold.: Simulation of Agricultural Management Alternatives for watershed Protection,Water Resour manage. DOI 10.1007/s11269-010-9598-8,2010.

[44]. Reynolds, C.A., T. J. Jackson, and W.J. Rawls. Web site. Estimated Available Water Content from the FAO

Soil Map of the World, Global Soil Profile Databases, and Pedo-transfer Functions, 1999. http://www.ngdc.noaa.gov/seg/cdroms/reynolds/reynolds/reynolds.htm

[45] Schmidt, J., Werner, M.V., Michael, A.. Application of the EROSION 3D model to the CATSOP watershed, The Netherlands. In: De Roo, A. (Ed.): Modelling Soil Erosion by Water at the Catchment Scale. Catena 418: pp.449-456. 1999.

[46] Setegn, S.,Dargahi, B.,Srinivasan,R., and Melesse,A.: Modeling of Sediment Yield From Anjeni-Gauged Watershed,Ethiopia Using SWAT model, Journal of American Water resources As.vol.46, pp.514-526,2010.

[47] Steenhuis,T.,Collick,A.,Easton,Z.,Leggesse,E.,Bayabil,H.,White,E.,Awulachew,S.,Adgo,E.,and Ahmed,A.:Predicting discharge and sediment for the Abay(Blue Nile) with a simple model, Hydrol.Process.vol.23,pp.3728-3737,2009.

[48] .Tesfahunegn, G.B., P.L.G. Vlek and L.Tamene.: Management Strategies for reducing soil degradation through modeling in a GIS environment in northern Ethiopia catchment, Nutrient cycling in Agroecosystems. In press,1-16,2012

[49] US department of Agriculture-Soil Conservation Service(USDA-SCS): National Engineering Handbook,Section IV, Hydrology,vol. 4-102pp.,1972.

[50] Van Liew, M.W.,T.L.Veith,D.D. Bosch, and J.G.Arnold.: Suitability of SWAT for the conservation Effects Assessment Project: comparison on USDA Agricultural Research Service watersheds .J.Hydrol.Eng. Vol12(2):pp.173-189,2007.

[51] Van Liew,M.W.,and J.Garbrecht.: Hydrologic Simulation of the Little Washita river experimental watershed using SWAT. J. Am.Water Resour.Assoc. Vol 39(2), pp. 413-426,2003

[52]. Waterbase http://www.waterbase.org/download_data.html,

[53]. Williams,J.R.: Flood Routing with variable travel time or variable storage coefficients, T.ASAE.vol.12,pp.100-103,1969.

[54 Williams,J.:SPNM, a Model for predicting sediment,Phosphorus, and nitrogen yields from agricultural basins,J.Am.Water.Resour. As.vol.16,pp.843-848,1980.

[55] Williams, J.R. Chapter 25: The EPIC model. In: V.P. Singh (eds.), Computer models of Watershed hydrology, Water resources Publications, Highlands Ranch, CO. pp. 909-1000. 1995.

[56] Williams,J. and Berndt,H.: Sediment Yield Prediction based on watershed hydrology,T.ASAE,20,pp.1100-1104,1977

[57] Y. ANLEY A., BOGALE AND A. HAILE-GABRIEL.: Adoption Decision and Use Intensity of Soil and Water conservation Measures by small holder subsistence farmers in Dedo district, western Ethiopia,Land degrad. develop. 18, pp.289-302, 2006

[58] Young, R.A., Onstad, C.A., Bosch, D.D., Anderson, W.P. AGNPS: A non point Sourse Pollution Model for Evaluating Agricultural Watersheds. Journal of Soil and Water Conser-vation .vol.44(2): pp.168-173, 1989.

[59] Zeleke,G.:Landscape dynamics and soil erosion process modelling in the north-western Ethiopian highlands,in: African Studies Series A, Geographica Bernensia,Berne,2000.

Assessing Reliabilities of Multipurpose Reservoir Using Stochastic Modeling and scenario ...

201

NCSTI-2012

Assessing Reliabilities of Multipurpose Reservoir Using Stochastic Modeling and Scenario Based Simulations of Net

Inflows, the Case of Lake Tana, Ethiopia Mulugeta Azeze1; Hartmunt Eckstädt2; Yilma Seleshi3

School of Civil and Water Resources Engineering, Bahir Dar University, Bahir Dar, Ethiopia1, Institute of Sanitary Engineering, Rostock University, Rostock, Germany2, Civil Engineering Department, Addis Ababa University, Addis Ababa, Ethiopia3 Abstract — Assessing the reliabilities of multipurpose reservoir is very instrumental and paramount importance for effective and efficient management of the resource. Lake Tana, the biggest fresh water body in Ethiopia, is a natural Lake serving multiple purposes. Recently it has been identified as one of the most important sources of water for hydropower, irrigation, and fishery development. However, the Lake is being exploited based on results of rough estimates obtained from the analysis of its water balance. The purpose of this article is to evaluate the yield of the Lake and its reliabilities using the monthly partial sum series of the Lake water balance terms. The Lake is simulated with 100years synthetically generated net inflow series of different realizations. The simulation results show that the Lake has a safe annual yield of 2700*106m3 of water. In its current state, the Lake can satisfy the requirements for both Tana-Beles hydropower project supply and Lake Navigation at 98% reliability. However, the realization of the planned irrigation projects in the Lake sub-catchments, which are deemed to use nearly 600 *106m3 of water, will reduce the reliabilities of Tana-Belese hydropower project supply and Lake Navigability to 86% and 93% respectively. Keywords- Lake Tana, Partial sum series, net inflow, yield, reliability, simulation, multipurpose reservoir

1. INTRODUCTION Lake Tana is the largest fresh natural water body in Ethiopia with a maximum depth of 14m and mean depth of 9m. It is located at 110 36' N and 370 23' E with average natural altitude of 1786meter above sea level (masl). At similar elevation the lake has a surface area of about 3156km2 with length of 74km and 68km width and storage capacity of 29km3. It is the head water of the Blue Nile River. The lake is shallow and freshwater with weak seasonal stratification. More than 40 rivers and streams drain into the lake. However, it's most significant inflows come from four major perennially flowing rivers: Gilgel Abay, Rib, Gumera and Megech. The only natural out flowing river is the Blue Nile which is named Abay locally. These four major rivers contribute 93% of the inflow to the Lake [5] There is an increasing demand and effort for irrigation and hydropower development in Ethiopia. The catchments of Lake Tana and the adjacent river Beles have been identified by the Ethiopia Federal Democratic Republic (EFDR) government as the first growth corridor of the country [8]. Huge investment plans have been prepared to develop the water resources of these sub-basins. In line with this the constructions of some water conveyance and storage structures for hydropower and irrigation development have been started in Lake Tana and Beles sub-basins. Tana-Beles hydropower projects, that has an installed capacity of 460MW and supplied by

Lake Tana at an average supply rate of 77m3/s [10], is the biggest among the planned projects in the sub-basins. The construction of this hydropower project is completed and commissioned in year 2010. The construction of a storage Dam for irrigation of 7000ha of land on Koga River, a tributary to Gilgel Abay, is also completed. The constructions of some other Dams in Tana sub-catchments are still going on. Effective management of water resources of a river basin requires an understanding of the amount of water which can be provided under various conditions. Yield estimates are key element in practically all studies and decisions involving water supply. However, hydrology of Lake Tana is not well documented in the scientific literatures except in very few journal articles and reports that present estimate of the water balance of the Lake. Kebede et al. (2005) have estimated the water budget component of the lake, and make a preliminary sensitivity analysis of Lake Level and Lake Outflow to rainfall changes. However, the estimate is based on the assumptions that net groundwater flux is negligible and rainfall at the closest station to the Lake represents the average direct rainfall on the Lake. The Lake is large to represent its areal precipitation using a single nearby station. Nevertheless, they have estimated annual values for the most uncertain water balance terms as 1451mm, 1162mm, and 1478mm for the Lake Precipitation, inflows, and evaporation respectively with a balance error of + 22 mm/year. Wale (2008) has estimated the runoff


202

NCSTI-2012from ungauged catchments using regionalization as 880mm per year with a water balance closure error of 5%. But SMEC (2008a) has indicated the tricky in regionalization approach as the catchments generating the base flow and surface run off for some of the major tributaries like Gilgel Abay are not concurrent. SMEC (2008a), however, has estimated the water balance using a reverse Lake water balance calculation in which the Lake inflows were estimated from records of water level, outflow, evaporation, and direct rain fall on the Lake. Consequently, it has indicated that some 25% of the Lake inflow is from ungauged catchments. This is a significant amount which was considered as 7% in kebede et al’s. (2005) estimate. Chebud et al’s (2010) has also estimated the hydrological balance of Lake Tana by neglected the interaction between the Lake and the surrounding ground water. Everyone involved in water resource systems development and management has an obligation to see that those systems provide sufficient quantities and qualities, at acceptable price and reliabilities, and at the same time protect the environment and preserve the biodiversity and health of ecosystems for future generation [6]. Existing water resources development in Lake Tana for hydropower generation has, however, modified flows downstream of the Lake, reduced water levels of the lake and significantly decreased flows over the Tis Issat waterfall [7]. Therefore, in response to the increased rate of development in Lake Tana there is an intensified need for reliable and more accurate hydrologic models to be used in applications such as planning, operation, and environmental impact assessment of the interventions. The purpose of this paper is to assess the yields of the reservoir and their corresponding reliabilities.

II. CONCEPTUALIZATION AND MODELING APPROACHES

In Ethiopia the network of hydro-meteorological stations is very scanty. Lake Tana has, however, relatively the longest hydrologic records. Mean monthly outflow from the Lake, through the natural outflow river Abay, and Lake Levels are available intermittently since 1959. The EFDR Ministry of Water and Energy (MWE) is responsible for the collection and management of hydrological and meteorological data. The hydro-meteorological data pertinent to this paper was collected from MWE. More than 60% of the surface inflows to the Lake are not measured and there is no data that shows the groundwater inflow to the Lake. Relatively sufficient and reliable records were obtained for the Lake Level and Lake Outflow only. In deriving and applying reservoir storage

yield relationships, it is desirable to use unregulated natural flows in order to apply correlation procedures [12]. However, a weir has been constructed and commissioned in the year 1996 at the outlet of the Lake for the purpose of supplying a regulated flow to Tis Abay I and II hydropower stations which are found at 35km downstream of the weir. The Lake Level and Lake Outflow from year 1996 onwards are affected by the regulation rules of the weir gates. The effect of the regulation on both the Lake Level and Lake Outflow can be seen clearly from the time series plots on Figure 1. Hence, monthly records of the Lake Level and Lake Outflow from 1976 to 1995 have been identified as natural flow and selected for further analysis. Missed data for the selected data range have been filled by interpolation and the presence of outliers were detected and corrected using the box plot rule [14]. Stream flow, precipitation, evaporation, water demand, and other variables pertinent to yield determinations are highly stochastic. The stochastic nature of these variables must be reflected in methods quantifying yield [15].

Figure 1. Time series Plot of Lake Level and Lake Outflow Using stochastic hydrology, sequences of any specified length for hydrological variable such as monthly stream flow can be synthesized. The synthetically generated data will then be used to simulate the physical process under consideration. Usually in reservoir design and analysis synthetic inflows are generated to determine either the storage capacity given the demand or the demand given the storage. But in Lake Tana there are no sufficient representative records of inflows. Hence to use the classical approaches of reservoir design or analysis one has to look either for methods to estimate the inflows from the unguaged catchments or device a proper conceptualization that uses only the available but reliable records. One of such conceptualization is the use of partial sum series. Response of reservoirs to their inputs and outputs is revealed in the change of water levels in the reservoirs. Among the water balance terms of Lake Tana, only the Lake Level and Lake Outflow have relatively sufficient and reliable records. Using these terms and the water


203

NCSTI-2012

balance equation a partial sum (Ps) for the other most uncertain terms can be obtained as:

Qin + P - E ± Qg - Qout = ΔS Qin + P - E ± Qg = ΔS + Qout

Let Ps = Qin + P - E ± Qg Ps = ΔS + Qout

Where Qin is the inflow to the Lake, P is the direct precipitation, E is the evaporation, Qg is either ground water inflow to the Lake or Lake Outflow to the surrounding aquifer, Qout is the outflow from the Lake through the out flowing river, ΔS is the change in Lake storage, and Ps is the partial sum of the water balance terms. The partial sum is the net inflow to the Lake. The concept of stochastic hydrology can be applied on the partial sum series (Ps) i.e. the net inflow and an evaluation of the yield of the reservoir and its corresponding reliability can be obtained. This approach can eliminate the approximation made for most uncertain water balance terms like the Lake inflow and it is comprehensive as it includes all possible inputs and outputs. Furthermore, it is easy to account the aggregated serial correlation and seasonality of the uncertain and unmeasured water balance terms of the reservoir in the partial sum series. It is conceptually more realistic and will give accurate results.

III. MODELING THE PARTIAL SUM SERIES

A. Lake Storage and Elevation Relationship Studio Pietrageli (1990) has developed Bathymetry of Lake Tana. A relationship between the Lake storage capacity and its water surface elevation is developed for the active storage zone using the data obtained from Studio Pietrageli (1990). The sill level for natural outflow river from the Lake was at 1785masl. But after the construction of the Chara chara weir, a weir constructed at the outlet of the Lake to regulate its outflow, the outflow sill level has been changed to 1784masl [9]. The crest level of the weir is set at an elevation of 1787.00masl. The Lake storage below 1784.00masl is a dead storage. Taking the zone of the Lake above 1784masl as an active storage, a linear relationship with R2 = 0.999 is developed between the storage capacity and the water surface elevation of the Lake. This relationship is used to convert Lake Level to equivalent storage volume. The relationship has the form:

V = 3073.2*H -5459500.4 Where V is the volume of the Lake in Million

cubic meters, and H is the elevation of the Lake water surface in masl. B. Model Identification and Calibration Monthly data for the Lake Level and Lake Outflow have been used to generate the partial sum series. Visual inspection of the partial sum series plot on Figure 2 shows clear seasonality. Besides, the Partial sum series has some negative values that can be explained by the excess of evaporation and/or recharge of ground water over precipitation and inflows in some months of the year. This is apparent in the dry seasons when no rainfall and inflows are very minimal. A Box-Cox transformation is used to maintain normality and constant variance in the model residual and to eliminate the negative terms in the partial sum series. The transformation has the form:

Yt = ln (Zt + C) Where zt is the partial sum series, Yt is the transformed partial sum series, and C is a constant which is chosen as 7 *108 to make all values in the series positive. The transformed partial sum series has been deseasonalized by subtracting the seasonal mean from each series and also dividing this by the seasonal standard deviation [3].

Yea rM onth

1994199119881985198219791976FebFebFebFebFebFebFeb

5000000000

4000000000

3000000000

2000000000

1000000000

0

Net

Inflo

w

T ime s e r ie s plot o f N et inflow

Figure 2. Time Series Plot of partial Sum Series (Net Inflow) The focus is to model the transformed and deseasonalized partial sum series (Yt) shown on Figure 3.

Y e a rM o n th

1 9 9 41 9 9 11 9 8 81 9 8 51 9 8 21 9 7 91 9 7 6F e bF e bF e bF e bF e bF e bF e b

4

3

2

1

0

- 1

- 2

- 3

- 4

Ran

d

T i m e S e r i e s p l o t o f th e R a n d o m C o m p o n e n t o f th e N e t I n f l o w

Figure 3. Time series Plot of random component of Net Inflow The plot of the autocorrelation function, shown on Figure 4, of the transformed and


204

NCSTI-2012deseasonalized series is used to identify the suitable model. The plot indicates first order Moving Average, MA (1), model as initial choice to represent the series. The stochastic model we are looking to employ is, therefore, based on the ideas that a time series in which successive values are highly dependent and can frequently be regarded as generated from a series of independent "shocks" at. These shocks are random drawings from a fixed distribution, usually assumed normal and having mean zero and variance σa

2. Such a sequence of random variables at, at-1, at-2... is called a white noise process or innovation [1]. The white noise process at is transformed to the process Yt by a model of the form:

Yt =µ + at +θ1 *at-1 + θ2*at-2 +…+ θ k *at-k Where µ and θi are model parameters which will be determined from the data and k is the order of the model. When µ=0 and k=1, the above equation will reduce to:

Yt =at +θ1 *at-1

Figure 4. Plot of Autocorrelation Function The length of the selected series has been checked for sufficiency before proceeding to further analysis. The ratio of the number of observations to the number of model parameters can be used as a rough check for the adequacy of the data length [3]. The ratio is well above four for the selected model and hence the data length is considered as sufficient. The values of the model parameters, θ1 and σa

2, are estimated and found to be -0.2866 and 0.9368 respectively. The adequacy of this model has been assessed and confirmed using various model diagnostic methods. To ensure the adequacy of the model in describing the time series under consideration, a graph of the residual autocorrelation function (RACF) shown on Figure 5 has been plotted and examined. The RACF indicates that the chosen model for the partial sum series of Lake Tana water balance satisfies the whiteness assumption. This fact is also confirmed by Li and McLeod modified χ2 distributed Portmanteau statistic QL whose calculated magnitude for QL is 32.90 for 34 degrees of freedom and is therefore, not significant.

Figure 5. Plot RACF and ±95% confidence

43210-1-2-3

99.9

99

9590

80706050403020

10

5

1

0.1

Res

Perc

ent

M ean -0.009010S tD ev 0.9993N 239A D 0.486P -Valu e 0.225

N ormality test for R esidualsNorm al

Figure 6. Normality plot of Residuals The assumption of normality of the residuals of the fitted MA (1) model is ascertained by calculating the skewness coefficient (g1) and using the transformation method given by DAgostino [3]. For 95% significance level the g1 statistics shows that g1 has a value of 0.2148 which is less than 1.96 times its standard error (SE) and this indicates that the residuals are normally distributed. This has been also check using normalty plot shown on Figure 6.

IV. SIMULATION Usually Storage Yield (SY) analysis is carried out to determine the smallest volume of storage required to meet given demand with a stated reliability. In the case of natural reservoirs like Lake Tana and existing man made reservoirs the analysis is carried out to determine the yield of the reservoir or reassessing the demand with a stated reliability given the storage. Generally, the amount of water which can be supplied by a reservoir may be analyzed in simulation study in terms of a firm or safe yield, percent of time specified quantities of water are available, reliability of meeting various demand levels, risk of shortages, or likelihood of various reservoir storage levels occurring. The focus of this paper is on yield versus reliability relationships and determination of safe yield. Safe yield or firm yield is defined as the estimated maximum release or withdrawal rate which can be maintained continuously during a repetition of the hydrologic period of record [15]. Reliability can be defined as the


205

NCSTI-2012

percentage of time that a specified demand level can be met [15]. Simulation techniques are the most commonly used tool for analyzing reservoir yield. This technique can realistically and conveniently be used to examine and evaluate the performance and response of the reservoir to a set of alternative options. Currently the Lake is used for two major purposes: to supply the required amount of water to Tana-Beles hydropower project and for navigation. The simulation run will help us to assess the Lake capacity in providing these services. The operation of Lake Tana is simulated using different realizations of 100 years synthetically generated partial sum series and with a set of predefined rules. First, the minimum outflow required for environmental flow and the minimum water level required for Boat and Ferry transportation are determined. The minimum water level required for Boat and ferry transportation is 1784.75masl. Following the construction of Chara chara weir and the regulation of the Lake outflow, an estimate for the environmental flow using the Desktop Reserve Model (DRM) has been made by McCartney et al. (2010).The study indicates the need of allocating 22% of the natural mean annual flow to environmental flow. SMEC (2008b) have also presented summarized result of other studies and indicates the need of releasing a fixed regular flow of 17m3/s during the non spill period for environmental flow. Based on these estimates approximately 20% of the demand, which is on average equivalent to 20m3/s, is assumed to be released to the natural out flowing river as environmental flow for all the periods except for the periods when the reservoir is full and spilled over. The lake will spill over when its Level is above the Chara Chara weir crest level. After determining the environmental flow and the minimum level required for navigation purpose the Lake is simulated with different demands and various operation Levels. Base on the Tana-Beles average supply and the environmental flow, various simulation flows that are above and below the sum of these flows have been used to assess the capacity of the Lake. For each selected demand and operation level the Lake is simulated in such a way that at each time interval an attempt is made to satisfy the demand to the extent possible. During the simulation run, if there is no enough water in the reservoir to meet the demand i.e. when the Lake level is below the chosen operation level, the amount required only for environmental flow is released and the period is treated as failure to satisfy the demand. Beside, in any simulation run the period with a Lake level below 1784.75masl is treaded as failure for navigation requirement. If there is so much water during the period that there is no place to keep it even

after meeting all the demands, i.e. if the reservoir level is above 1787masl, the extra water over the storage capacity is allowed to spilled over. For the chosen demand values and operation levels the corresponding reliabilities for supplying the Tana-Beles hydropower project and transportation services are estimated. The reliabilities are estimated using:

R = [1 – (FAIL (F) /N)] * 100 Where R is the reliability of reservoir supply or navigation service, FAIL is the number of failure (number of months when the release is less than the required demand or the Lake level is below 1784.75masl). N is the number of periods (months) used in the simulation analysis.

V. RESULTS AND DISCUSSIONS The assessment of supply and navigation reliabilities shows that limiting the supply will reduce the rapid decline of Lake Level and hence increase reliability for Lake Navigation services. But the reliability for Lake Supply increases as the minimum operating level decreases. Plots of reliabilities of lake supply capacity and Lake Navigability versus Lake Levels, shown on Figure 7, help to easily identify the minimum optimum operation level. From these plots it is identified that the Lake has a minimum optimum operation level of 1784.62masl. If the Lake is operated at its minimum optimum operation level 98% reliability for both the Lake supply capacity to Tana-Beles hydropower project and Lake Navigation can be obtained. The effect of the planned irrigation projects on the reliabilities of Lake Supply capacity and transportation is assessed further by making various levels of reduction in the net inflows series to the Lake. Assuming that the planned reservoirs in the inflowing rivers to the Lake have flood controlling purpose in addition to storing water for irrigation and with the fact that significant refill of the Lake is obtained in the months of August and September, only these months net inflow series have been reduced by a factor that corresponds to the planned irrigation demand. Thus, simulating the Lake with the modified net inflow series indicates that the reliabilities of Lake Supply to Tana-Beles hydropower project and Lake Navigability will reduce to 86% and 93% respectively.


206

NCSTI-2012

1784.8

0

1784.7

5

1784.7

0

1784.6

5

1784.6

0

1784.5

5

1784.5

0

1784.4

5

1784.4

0

1784.3

5

1.000

0.995

0.990

0.985

0.980

0.975

HP ReliabilityNAV Reliability

Reliability Curves

[2] Chebud, Y. A. and Melesse, A. M.: “Modelling lake stage and water balance of Lake Tana, Ethiopia”, Hydrol. Process, 23, 3534–3544, 2009.

[3] Hipel, K.W; McLEOD, A.I (1994): Time Series Modelling of Water Resources and Environmental Systems. Amsterdam: Elsevier (45).

[4] Jain, S.K; Singh, V.P (2003): Water Resources System Planning and Management. First Edition. Amsterdam: Elsevier (51).

[5] Kebede, S.; Travi, Y.; Alemayehu, T.; Marc, V. (2005): “Water Balance of Lake Tana and Its sensitivity to Fluctuations in Rainfall, Blue basinn, Ethiopia”. In Journal of Hydrology, 316, pp. 233-247.

Figure 7. Hydropower and Navigation Reliabilities

[6] Loucks, Daniel P. (2000): Sustainable Water Resources Management. In water International, 25 (1), pp. 3–10. VI. CONCLUSION

The planning and management of Lake Tana water resource systems should be based on good decision support tools. Without having and being informed with these tools the endeavour in developing the Lake and its catchment resources will miss the target and end up with an overall loss of socio-economic benefit and environmental degradation. The Lake is a finite resource with good but limited potential for various developments. The yield-reliability study shows that the Lake has an annual safe yield of 2700millions cubic meters of water. The simulation analysis made with a release of constant environmental flow of approximately 20% of the demand in the non spill periods to the natural out flowing river indicates that 98% reliability for both Lake supply to Tana-Beles hydroelectric power project and Lake navigation can be obtained if the Lake is operated at its minimum optimum operating level of 1784.62masl. However, if all the planned irrigation projects in the sub-catchments of Lake Tana are implemented the reliabilities of the Lake Tana supply to Tana-Beles hydroelectric power project and Lake Navigation will reduce to 86% and 93% respectively. This is a significant reduction that seriously affects the sustainable utilization of the Lake resources. It will violate the obligation that anyone involved in the development and management of water resource systems has to see the system’s capacity in providing sufficient quantities and qualities, at acceptable price and reliabilities, and at the same time protect the environment and preserve the biodiversity and health of ecosystems for future generation. Developing a decision support system based on a system approach is paramount important and urgently need to Lake Tana water resources systems planning and management.

[7] McCartney, Matthew; Alemayehu, T.; Shiferaw, A.; Awulachew, S.B (2010): “Evaluation of Current and Future Water Resources Development in the Lake Tana Basin, Ethiopia”. IWMI (134).

[8] Ministry of Finance and Economoc Development (MoFED) (2006): “A Plan for Accelerated and Sustained developmnet to End Poverty (PASDEP)”. Ministry of Finance and Economic Development. Addis Ababa.

[9] SMEC (2008a): “Hydrological Study of the Tana-Beles Sub-Basins, Surface Water Investigation”. SMEC international Pty Ltd. Addis Ababa.

[10] SMEC (2008b): “Hydrological Study of the Tana-Beles Sub-Basins. Ecological Studies with Emphasis on Biological Resources”. SMEC international Pty Ltd. Addis Ababa.

[11] Studio Pietrangeli (1990):Tana –Beles Project. Hydrological Report. Addia Ababa.

[12] US Army Corps of Engineers (1975): “Reservoir Yield”. Hydrologic Engineering Center

[13] Wale, Abeyou (2008): Hydrological Balance of Lake Tana Upper Blue Nile Ethiopia. Master´s dissertation. ITC, Enschede. International Institute for Geo-information Science and Earth Observation.

[14] Wilcox, Rand R. (2009): Basic Statistics. Understanding Conventional Methods and Modern Insights. New York: Oxford University press.

[15] Wurbs, R.A; Bergman, C.A (1990): “Evaluation of Factors Affecting Reservoir Yield Estimates”. In Journal of Hydrology, 112, pp. 219–235.

REFERENCES [1] Box, George E.P et al. (1994): Time Series

Analysis Forecasting and Control”. New Jersey: Prentice Hall.

lake level

Relia

bilit

y

Calibration and Validation of SWAT Model and Estimation of Water Balance Components ...

207

NCSTI-2012

Calibration and Validation of SWAT Model and Estimation of Water Balance Components of Shaya River Watershed, Genale-

Dawa Basin, South-Eastern Ethiopia

Alemayehu Abate1, Tena Alamirew2 and Megersa Olumana3

1 Department of Natural Resources Management, Madawalabu University, Bale Robe, Ethiopia E-mail: [email protected], PO Box 247 Bale Robe

2 Ethiopian Institute of Water Resources, Addis Ababa University, Ethiopia E-mail: [email protected]

3School of Natural Resource and Environmental Engineering, Haramaya University, Ethiopia E-mail:[email protected]

Abstract — Water resource development is certainly the basic and crucial infrastructure for a

nation’s sustainable development. To utilize water resources in a sustainable manner, it is necessary to understand the quantity and quality in space and time. This study was initiated with the objective of evaluating the performance and applicability of the physically based Soil and Water Assessment Tool (SWAT) model in analyzing the influence of hydrologic parameters on the streamflow variability and estimation of monthly and seasonal water yield at the outlet of Shaya river watershed in Bale mountainous area. Sensitivity analysis, model calibration and validation were made to evaluate the model performance for simulation of streamflow. The calibrated SWAT model performed well for simulation of monthly streamflow. Statistical model performance measures, coefficient of determination (r2) of 0.71, the Nash-Sutcliffe simulation efficiency (ENS) of 0.71 and Percent difference (D) of 3.69, for calibration and 0.76, 0.75 and 3.30 respectively for validation, indicated good performance of the model simulation on monthly time step. Mean monthly and annual water yield simulated with the calibrated model were found to be 25.8 mm and 309.0 mm, respectively. Overall, the model demonstrated good performance in capturing the patterns and trend of the observed flow series, which confirmed the appropriateness of the model for future scenario simulation. Therefore, it is recommended that SWAT model can be taken as a potential tool for simulation of the hydrology of unguaged watershed in mountainous areas, which behave hydro-meteorologically similar with Shaya watershed. Keywords: Hydrological modeling, calibration, SWAT, water balance, sensitivity analysis, Shaya River, Bale Mountains.

I. INTRODUCTION

Understandings on hydrological processes to develop suitable models for a watershed are the most important aspect in water resource development and management programmes. Water resource development is the basic and crucial infrastructure for a nation’s sustainable development. To utilize water in a sustainable manner, it is necessary to understand the quantity and quality in space and time through studies and researches [1]. Major hydrological processes can be quantified with the help of water balance equations. The component of water balance of a watershed is influenced by climate, and the geophysical characteristics of the watershed such as topography, land use and soil. Consideration of the relationship between these physical parameters and hydrological components is very essential for any water resource development related work [2]. Since the hydrologic processes are very complex,

their proper comprehension is essential and therefore, watershed based hydrological models are widely used.

Mountainous watersheds are the origin of many of the largest rivers in the world and represent major sources of water availability for many countries [3]. They represent not only local water resources (local freshwater supply, hydropower generation, irrigation, etc), but also considerably influence the runoff regime of the downstream rivers. [4] described that the Bale Mountain National Park (BMNP) is a source of over 40 streams on which more than 12 million people are dependent. The importance of the hydrological services that the area provides to south-eastern Ethiopia and parts of Somalia and Kenya have gradually been recognized over the subsequent years and their conservation is now a primary purpose of the park. Shaya is one of a river which originates from afroalpine area of the BMNP among many other rivers. Expansion and


208

NCSTI-2012

encroachments of agriculture, settlements and livestock, however, are the main causes that make the hydrological system of the area in under continuous transformation.

At the downstream parts of the river, there are different water based projects which are attached to the flow of Shaya River. The projects include existing and proposed irrigation schemes, tourism and fish farming at different parts of the river. Hence, estimation of monthly, seasonal and long term runoff yield helps to identify the best and sustainable land use and management options in the area. Therefore, the output of this study can be taken as an input to plan and implement effective land and water resources development and management.

There are a number of integrated physically based distributed models. Among which, researchers have identified SWAT as one of the most promising and computationally efficient model [5]-[6]. Therefore, to test the capability of the model in determining the effect of spatial variability of the watershed on streamflow, SWAT 2005 with ArcGIS interface was selected. The time series data on climate and runoff yield were available at the gauging stations of the watershed and these were used to calibrate and validate the SWAT model and to assess its applicability in simulating runoff yield from the Shaya watershed. The objective of this study was to evaluate the performance and applicability of the physically based SWAT model for estimation of the streamflow variability in Bale mountainous region of the Shaya river watershed with the specific objective: To perform calibration and validation of SWAT model at the outlet of Shaya watershed in the Bale Mountainous region and to estimate water balance components of the watershed.

II. MATERIALS AND METHODS A. Description of the Study Watershed

Shaya watershed is found in south-eastern part of Ethiopia in Oromia Regional State, Bale Zone. The watershed is situated in Genale-Dawa basin at the upper most parts of the Weyb basin, located between 6° 52' - 7° 15' N latitudes and 39° 46' - 40° 02' E longitudes as shown in Fig. 1. It covers a total drainage area of 503.5 km2. The Shaya River originates from the northern flanks of the Bale Mountains and first flows generally north-eastwards before joining the Weyb river which flows to east and south eastwards for the remainder of its course. Finally, it joins with Genale and Dawa River near Ethiopia-Somalia border to strengthen its journey to Somali lowlands. It

originates from an elevation of 4,343 meter above sea level (m.a.s.l), in the Bale Mountains extreme point locally called Sanetti Mountains to an elevation of 2,357 m.a.s.l at the outlet of the watershed. The average annual rainfall distribution is 1071 mm and the annual maximum and minimum temperature of the watershed area is about 19.7 ºC and 6.1 ºC, respectively.

The Shaya river watershed is a region of rich environmental diversity, but with increasing levels of environmental stress in recent years from a rapidly expanding human population [4].

B. General Description of SWAT Model

SWAT is the acronym for Soil and Water Assessment Tool, a river basin scale, continuous-time and spatially distributed model developed for the USDA Agricultural Research Service (ARS). It was developed to predict the impact of land management practices on water, sediment and agricultural chemical yields in large complex watersheds with varying soils, land use and management conditions over long periods of time [5]. In recent years, SWAT model has gained international acceptance as a robust interdisciplinary watershed modeling. SWAT is currently applied worldwide and considered as a versatile model that can be used to integrate multiple environmental processes, which support more effective watershed management and the development of better informed policy decision [7]. The review of SWAT model applicability to Ethiopian situations at relatively larger watersheds [8]-[9] indicated that the model is capable of simulating hydrological processes with a reasonable accuracy. In SWAT, a watershed is divided into multiple sub- watersheds, which are then further subdivided into hydrologic response units (HRUs) that consist of homogeneous land use, management, and soil characteristics [5].In the land phase of hydrological cycle, SWAT simulates the hydrological cycle based on the water balance equation:

t

1igwseepasurfdayot QWEQRSWWS (1)

where, SWt is the final soil water content (mm), SWo is the initial soil water content on day i (mm), t is the time (days), Rday

is the amount of

precipitation on day i (mm), Qsurf is the amount of

surface runoff on day i (mm), Ea is the amount of evapotranspiration on day i (mm), Wseep

is the

amount of water entering the vadose zone from the soil profile on day i (mm), and Qgw is the amount of return flow on day i (mm).


209

NCSTI-2012

Figure 1: Major River basins of Ethiopia and location map and sub-basins of Shaya watershed

Figure 2: Soil map (left) and land use map (right) of Shaya watershed

C. SWAT Model Inputs

Digital Elevation Model (DEM): The DEM forms the base to delineate the watershed boundary, stream network and create sub-basins. This was performed by the pre-processing module of the SWAT but requires a so-called minimum threshold area. Topography was defined by a DEM which describes the elevation of any point in a given area at a specific spatial resolution as a digital file. It was also used to analyze the drainage patterns of the land surface terrain. And sub-basin parameters such as slope, slope length, and defining of the stream network with its characteristics such as channel slope, length, and width were derived from the DEM. For this specific study a DEM with a resolution of 30 m was used, which was sourced from ASTER GDEM official website, released by ERSDAC (Earth Remote Sensing Data Analysis Center).

Soil Properties: Soils in the study watershed are classified on the basis of the revised FAO/UNESCO-ISWC [10] classification system. The soil data was extracted from the 1:250,000 scale of soil map (Fig. 2) developed by Ministry of Water and Energy (MoWE) [11]. Basic physico-chemical properties of major soil types in the watershed were mainly obtained from the following sources: Genale-Dawa river basin integrated resources master plan, Soil database and digital soil map from the MoWE produced between the year 2004 and 2007; Soil and Terrain Database for north-eastern Africa CD-ROM [12]. In addition to these sources, some soil properties were estimated based on available soil parameters. Major soil types of the study watershed are described in Table 1.

Land Use and Land Cover (LULC) Data: The LULC map and all datasets were obtained from [11]. This spatial database was derived from


210

NCSTI-2012satellite imagery and field data collected between year 2004 to 2007 [11] and is the most current and detailed LULC data known to be available for the study watershed (Fig. 2). The reclassification of the land use map was made to represent the land use according to the specific LULC types and the respective crop parameter for SWAT database as shown in Table 1. A lookup table that identifies the SWAT land use code for the different categories of LULC was prepared so as to relate the grid values to SWAT LULC classes.

Meteorological Data: The SWAT model requires daily meteorological data that could either be read from a measured data set or be generated by a weather generator model which include precipitation, maximum and minimum air temperature, solar radiation, wind speed and relative humidity. Meteorological data collected from National Meteorological Service Agency of Ethiopia (NMSA) for Bale Robe, Dinsho, Agarfa, and Goba stations are within the watershed and some are in the vicinity of watershed boundary as shown in Fig. 1. The SWAT weather generator model WXGEN was used to fill missing values in weather data. All weather stations provided precipitation and maximum and minimum temperature while solar radiation, wind speed and relative humidity were obtained from Robe weather station with varying recording periods. The Penman-Montheith method which utilizes the solar radiation, relative humidity and wind speed data records was employed for estimation of PET for this specific study. Meteorological stations were geo-referenced using latitude, longitude, and elevation data.

The typical quality of rainfall data was checked by cross correlation between the stations. The correlation coefficient among the stations on monthly rainfall amount ranged from 0.81 to 0.97. This implied a good agreement or consistency of record on the monthly rainfall series of the gauging stations. The climate data for study periods were finally prepared in .dbf format and imported to the SWAT model database.

Hydrological Data: The hydrology of the watershed reflects the rainfall pattern with river flow peaking firstly in April to May and subsequently in August and October. Daily river discharge data of the Shaya River were obtained from the Hydrology Department of the MoWE. It was used for performing sensitivity analysis, calibration and validation of the SWAT model.

An automated baseflow separation and recession analysis technique [13] was employed to separate the baseflow and surface runoff from the total daily streamflow records. This information was then used in order to get SWAT to correctly reflect basic observed water balance at the outlet of the watershed.

D. Model Set Up

Watershed Delineation: The first step in creating SWAT model input is delineation of the watershed from a DEM. Inputs entered into the SWAT model were organized to have spatial characteristics. Before going in hand with spatial input data i.e. the soil map, LULC map and the DEM were projected into the same projection called UTM Zone 37N, which is a projection parameters for Ethiopia. A watershed was partitioned into a number of sub-basins, for modeling purposes. The watershed and sub-basin delineation was done using DEM data. A mask was first created over the DEM around the study watershed, to reduce the processing time of the GIS functions. The watershed delineation process include five major steps, DEM setup, stream definition, outlet and inlet definition, watershed outlets selection and definition and calculation of sub-basin parameters. For the stream definition the threshold based stream definition option was used to define the minimum size of the sub-basins. About three fourth (3/4) of suggested threshold area by the ArcSWAT interface (default) was used for the delineation of sub-basins to increase the number of sub-basins for a more detailed analysis of the hydrologic processes.

Table 1:Land use and Major soil unit of Shaya watershed and their areal coverage in the watershed

Land Uses SWAT code

Area [ha]

[%] in wat.

Soil Unit Area [ha]

[%] in wat.

Residential-Medium/Low Density (Towns ) URML 1197.0 2.39 Chromic Cambisol 11027.86 22.02

Residential-Low Density(Farm Villages) URLD 1264.09 2.52 Dystric Cambisol 7743.29 15.46Agricultural Land-Generic(Mixed Cropping) AGRL 11099.67 22.17 Eutric Cambisol 13154.78 26.27Pasture PAST 10667.39 21.30 Haplic Luvisol 2946.76 5.88Range-Brush RNGB 17707.53 35.36 Eutric Regosol 1891.49 3.78Forest-Mixed FRST 2390.7 4.77 Vertic Luvisol 13313.20 26.59Range-Grasses RNGE 5510.7 11.00Forest-Evergreen FRSE 240.31 0.48


211

NCSTI-2012

Hydrological Response Units (HRUs): The land area in a sub-basin was divided into HRUs. The HRU analysis tool in ArcSWAT helped to load land use, soil layers and slope map to the project. The delineated watershed by ArcSWAT and the prepared land use and soil layers were overlapped 100%. HRU analysis in SWAT includes divisions of HRUs by slope classes in addition to land use and soils. The multiple slope option (an option which considers different slope classes for HRU definition) was selected. The LULC, soil and slope map was reclassified in order to correspond with the parameters in the SWAT database. After reclassifying the land use, soil and slope in SWAT database, all these physical properties were made to be overlaid for HRU definition. For this specific study a 5% threshold value for land use, 20% for soil and 20% for slope were used. The HRU distribution in this study was determined by assigning multiple HRU to each sub-basin.

Sensitivity Analysis: After a thorough preprocessing of the required input (temporal and spatial) for SWAT model, flow simulation was performed for an eight years of recording periods starting from 1992 through 1999. The first three years of which was used as a warm up period and the simulation was then used for sensitivity analysis of hydrologic parameters and for calibration of the model. The sensitivity analysis was made using a built-in SWAT sensitivity analysis tool that uses the Latin Hypercube One-factor-At-a-Time (LH-OAT) [14].

Calibration: For this study manual and automatic calibration method were applied. First the parameters were manually calibrated for the period of 1995 through 1999 until the model simulation results were acceptable as per the model performance measures. Next, the final parameter values that were manually calibrated were used as the initial values for the autocalibration procedure.The graphical and statistical approaches were used to evaluate the SWAT model performance a number of times until the acceptable values were obtained for surface runoff and baseflow independently. The flow calibration procedure made by SWAT developers in [15] was carefully followed. For each calibration run and parameter change, the corresponding model performance statistics r², ENS and D were calculated.

Validation: Streamflow data of three years from 2003 to 2005 were used for validation. The three statistical model performance measures used in calibration procedure were also used in validating streamflow.

Model Performance Evaluation: The regression coefficient (r2) describes the proportion of the total variance in the observed data that can be explained by the model. The closer the value of r2 to 1, the higher is the agreement between the simulated and the measured flow and is calculated as follow:

2

avi2

avi

2aviavi2

YYXXYYXX

r (2)

where: Xi is measured value, Xav is average measured value, Yi is simulated value, Yav is average simulated value, the same holds true for equation (2) and (4).

Nash and Sutcliffe simulation efficiency (ENS) indicates the degree of fitness of observed and simulated data and given by the following formula.

2

avi

2ii

NS XXYX

1E (3)

The value of ENS ranges from 1 (best) to negative infinity. If the measured value is the same as all predictions, ENS is 1. If the ENS is between 0 and 1, it indicates deviations between measured and predicted values. If ENS is negative, predictions are very poor, and the average value of output is a better estimate than the model prediction [16].

The percent difference (D) measures the average difference between the simulated and measured values for a given quantity over a specified period were calculated as follows:

i

ii

XXY

100D (4)

A value close to 0% is best for D. However, higher values for D are acceptable if the accuracy in which the observed data gathered is relatively poor.

III. RESULTS AND DISCUSSIONS A. Sensitivity Analysis

The model considered twenty seven flow parameters for sensitivity analysis from which twenty one of them were found to be relatively sensitive with the category of sensitivity ranging from very high to small. Among the sensitive flow parameters the ground water parameters were found to be more sensitive to streamflow. Deep aquifer percolation fraction; Rchrg_Dp, Initial curve number (II) value; Cn2, Baseflow alpha factor [days]; Alpha_Bf, Threshold water depth in the shallow aquifer for flow [mm]; Gwqmn, Soil evaporation compensation factor; Esco, Soil depth [mm]; Sol_Z, Threshold water depth in the shallow aquifer for "revap" [mm]; Revapmin, Maximum potential leaf area index; Blai, Available water


212

NCSTI-2012

capacity [mm water / mm soil]; Sol_Awc, Maximum canopy storage [mm]; Canmx, Groundwater Delay [days]; Gw_Delay, Saturated hydraulic conductivity [mm/h]; Sol_K and Surface runoff lag time [days]; Surlag were found to be the most effective hydrologic parameters for the simulation of streamflow. A brief description of each hydrologic parameter is listed in the SWAT model user’s manual [5].

B. Model Calibration

Model calibration followed sensitivity analysis. Calibration for water balance and streamflow was first done for average annual conditions. After the model was calibrated for average annual conditions, we shifted into monthly and daily records to fine-tune the calibration processes. Model efficiency measures for initial monthly default simulation r2, ENS and D were 0.60, 0.16 and 96.6 respectively which were beyond the acceptable ranges. Thus, model parameter adjustments were undertaken for a realistic hydrologic simulation.

The baseflow separation technique indicated about 60% of the total water yield was contributed from the subsurface water source which was more than surface runoff involvement for the total water yield at the outlet of the watershed.

The calibration processes were considered 13 most sensitive flow parameters (Table 2) and their values were varied iteratively within the allowable ranges until satisfactory agreement between measured and simulated streamflow was obtained. The autocalibration processes significantly improved model efficiency. The result from different statistical method of model performance evaluation met the criteria of ENS > 0.5, r2 > 0.6 and D ≤ ±15%. The statistical results of the model performance for both calibration and validation periods on monthly time steps are summarized in Table 3.

A rigorous hydrologic calibration resulted good SWAT predictive efficiency at the monthly time step of the watershed when compared to measured flow data (Fig. 3). The hydrograph of observed and simulated flow indicated that the SWAT model is capable of simulating the hydrology of Shaya mountainous watershed. However, the model was unable to capture some extreme values mainly, observed discharge on the month of November, 1997 and October 1998. The under prediction of flow during peak events by the SWAT model has been reported in many studies [2]-[6]-[17].

C. Model validation

It was found that the model has strong predictive capability with r2, ENS and D values of 0.76, 0.75

Table 2: Sensitive flow parameters and finally calibrated values

Parameters Sensitivity

RankUpper and

lower Bounds Calibrated

values Alpha_Bf 3 0.0 to 1.0 0.96Blai 5 0.0 to 1.0 0.27Canmx 10 0.0 to 10 7.55Cn2 2 ±25.0 -15.9Esco 4 0.0 to 1.0 0.16Gw_Delay 13 ±10.0 5.47Gwqmn 6 ±1000.0 854.02 Revapmin 9 ±100.0 92.71Rchrg_Dp 1 0.0 to 1.0 0.48Sol_Awc 8 ±25.0 15.43Sol_K 12 ±25.0 21.66Sol_Z 7 ±25.0 6.17Surlag 11 0.0 to 10 5.09

Table 3: Summary of model performance for calibration and validation periods

Mean annual water yield (mm)

Monthly model efficiency measures Period

Observed Simulated r2 ENS D (%) Calibration

330.30 336.09 0.71 0.71 3.69

Validation

251.89 319.87 0.76 0.75 3.30

and 3.30, respectively. Statistical model efficiency criteria fulfilled the requirement of r² > 0.6 and ENS > 0.5 which is recommended by SWAT developer [15]. This showed the model parameters represent the processes occurring in the watershed to the best of their ability given available data and may be used to predict watershed response for various outputs. The model validation results for monthly flow (Fig. 4) indicated generally a good fit between measured and simulated output. Since the model performed as well in the validation period, as for the calibration period hence, the set of optimized parameters listed in Table 2 during calibration process for Shaya watershed can be taken as the representative set of parameters for the watershed.

D. Monthly and seasonal water yield simulation

The water yield was simulated for the base period of twelve years 1995 to 2006 on monthly time step at the outlet of Shaya watershed. Moreover, the result was summarized in a monthly and seasonal bases as comparatively Dry (Feb-May), Wet (Jun-Sep), Intermediate (Oct-Jan) and annual bases, after an intensive model calibration for a sensitive flow parameters. Mean monthly and annual water yield simulated for a base period found to be 25.8 mm and 309.0 mm, respectively. Seasonal water yield simulation resulted 56.6 mm, 89.7 mm and 155.0


213

NCSTI-2012

Figure 3: Hydrograph of the observed, calibrated and default simulated monthly flow for the calibration period.

Figure 4: Hydrograph of the observed and simulated monthly flow for the validation period. mm for Dry, Intermediate and Wet seasons respectively. It indicated that the south-western part of the watershed has a larger contribution to the total water yield of the watershed.

Table 4: Average annual water balances simulated for a base periods of 1995-2006.

E. Average Annual Water Balance Components of the Watershed

The SWAT model estimated other relevant water balance components in addition to the daily and monthly discharge of the watershed. Average annual basin values for different water balance components during a base simulation periods shows average annual watershed gain and losses with change in soil water storage Table 4. From these components Actual evapotranspiration contributed a larger amount of water loss from the watershed and total water yield is the amount of streamflow leaving the outlet of watershed during the time step.

Water Balance Components Amount (mm)

Precipitation; Precip 1038.80

Surface runoff ; Sur_Q 83.24

Lateral soil flow contribution; Lat_Q 13.56

Ground water contribution to

streamflow; Gw_Q

212.46

Revap or shallow aquifer recharges 20.56

Deep Aquifer Recharges 174.12

Total water yield; Twyld 309.03

Percolation out of soil; Perc 435.80

Actual evapotranspiration; ET 518.7

Potential evapotranspiration; PET 606.8

Transmission losses; Tloss 0.22

Change in soil water storage 15.94


214

NCSTI-2012

III. CONCLUSION

Understandings on hydrological processes and develop suitable models for a watershed is the most important aspect in water resource development and management programmes. Watershed based hydrologic simulation models are likely to be used for the assessment of the quantity and quality of water. The performance and applicability of SWAT model was successfully evaluated through sensitivity analysis, model calibration and validation. Subsurface flow parameters were found to be more sensitive to the streamflow of the watershed, signifying the watershed is rich in ground water as a result of good recharge capacity. SWAT model was found to produce a reliable estimate of monthly runoff for Shaya watershed which was confirmed by various model efficiency measures. Therefore, the calibrated parameter values can be considered for further hydrologic simulation of the watershed and the model can be taken as a potential tool for simulation of the hydrology of unguaged watershed in mountainous areas, which behave hydro-meteorologically similar with Shaya watershed. However, for a more accurate modeling of hydrology, a large effort will be required to improve the quality of available input data. Future studies on Shaya watershed modeling should address the issues related to water quality and evaluate best management practices to address different watershed management issues.

REFERENCES [1] P. G. McCornick, A. B. Kamara and Girma

Tadesse (eds), "Integrated water and land management research and capacity building priorities for Ethiopia", Proceedings of a MoWR/EARO/IWMI/ ILRI international workshop held at ILRI, Addis Ababa, Ethiopia, 2003.

[2] K. Sathian and P. Symala, "Application of GIS integrated SWAT model for basin level water balance", Indian Journal of Soil Conservation vol. 37, no. 2, pp. 100-105, 2009.

[3] K. Sanjay, T. Jaivir and S. Vishal, "Simulation of runoff and sediment yield for a Himalayan watershed using SWAT model", Journal of Water

Resource and Protection, vol. 2, no. 3 pp. 267-281, 2010.

[4] Farm Africa-SOS Sahel Ethiopia, "Participatory Natural Resource Management Programme and Oromia Bureau of Agriculture and Rural Development", Six Months Report, Bale Robe, Ethiopia, 45p, 2007.

[5] S. I. Neitsch, J. G. Arnold, J. R. Kinrv and J. R. Williams, "Soil and Water Assessment Tool, Theoretical Documentation, Version 2005", Temple, TX. USDA Agricultural Research Service Texas A and M Black land Research Center, 2005.

[6] W. P. Gassman, M. R. Reyes, C. H. Green and J. G. Arnold, "The Soil and Water Assessment Tool: Historical development, Applications, and future Research Direction", Transactions of the ASABE vol. 50, no.4 pp. 1211-1250, 2007.

[7] W. P. Gassman, M. R. Reyes, C. H. Green and J. G. Arnold, "SWAT peer–reviewed literature: A review, Proceedings of the 3rd International SWAT conference", Zurich, 2005.

[8] Dilnesaw Alamirew, "Modeling of hydrology and soil erosion of Upper Awash River Basin", PhD Dissertation, University of Bonn, Germany, 236p, 2006.

[9] Setegn Shimelis, "Modeling hydrological and hydrodynamic processes in Lake Tana Basin, Ethiopia", KTH. TRITA-LWR PhD Dissertation, 74p, 2010.

[10] FAO/UNESCO-ISWC, "The World Reference Base for Soil Resources", Rome, Italy, 1998.

[11] MoWE, "Genale-Dawa River Basin Integrated Resources Development Master plan Study, Sector Reports", Addis Ababa, Ethiopia, 2007.

[12] FAO, "The Soil and Terrain Database for northeastern Africa (CDROM)” FAO, Rome, 1998.

[13] J. G. Arnold and P. M. Allen, "Automated methods for estimating baseflow and ground water recharge from streamflow records", J. Am. Water Resour. Assoc. vol. 35, no. 2 pp. 411-424, 1999.

[14] A. Van Griensven, "Sensitivity, auto-calibration, uncertainty and model evaluation in SWAT 2005", UNESCO-IHE. 48p, 2005.

[15] C. Santhi, J. G. Arnold, J. R. Williams, W. A. Dugas, R. Srinivasan and L. M. Hauck, "Validation of the SWAT model on a large river basin with point and nonpoint sources", J. Am. Water Resour. Assoc. vol. 37, no.5 pp. 1169-1188, 2001.

[16] J. E. Nash and J. V. Sutcliff, "River flow forecasting through conceptual models, part I- a discussion of principles", Journal of Hydrology, vol. 10, pp. 282-290, 1970.

[17] M. P. Tripathi, R. K. Panda and N. S. Raghuwanshi, "Identification and prioritization of critical sub-watersheds for soil conservation management using the SWAT model", Bio-systems Engineering vol. 85, no. 3 pp. 365-379, 2003.

Changes in sediment transport and channel morphology in a micro-scale Experimental ...

215

NCSTI-2012

Changes in Sediment Transport and Channel Morphology in a Micro-Scale Experimental Braided River

Michael M.M.

Bahir Dar University, Institute of Technology, Department of Civil and Water Resources Engineering, Bahir Dar, P.O.Box-26, Ethiopia, email: [email protected]

Abstract — The findings from a series of micro-scale laboratory experiments designed to examine

the effects of aggradation and degradation in the hydraulic geometry and sediment transport capacity of braided channels are discussed. The experiments were conducted in the University of Exeter Sediment Research Facility using a 5 m long and 2.7 m wide laboratory flume; hence situations are considered where channel width is unrestricted by the experimental setup. The experimental model is generic and is not scaled to a real world prototype. The evolution of river morphology was recorded using high resolution laser profiling to quantify channel changes (fill, incision and lateral erosion) and section geometry. During all the runs the evolution of the channel was recorded continuously using a Canon HG10 digital video camera and still imagery was collected at 2 minute interval using Canon EOS10D digital cameras. All cameras were mounted overhead. The quantitative observations from five experimental scenarios are discussed and evaluated. Rapid incision in the upstream portion of the channel resulted in the development of a single channel. However, the downstream reach remained braided as a result of continued delivery of sediments from the incising reach. The experimental work presented here has shown that despite the lack of dynamic similarity conditions and simplification of overall similarity criteria, fairly consistent results could be obtained which can be interpreted in a generic sense. The similarity between the laboratory channels from this experiment and those previously investigated by different researchers with or without a field prototype utilizing the Froude modelling principle suggest that the laboratory channels are at least qualitatively transferable to the field-scale.Keywords- micro-scale modelling, aggradation, degradation, river morphology

I. INTRODUCTION

Although a vast amount of research has been directed toward understanding braided rivers, we still do not completely understand the relationship between controlling independent variables (such as changes in sediment load) and braided river morphology. Relatively little has been done regarding the effects of sediment load on braided river pattern and evolution of longitudinal profile. Recent studies on bedload pulses in laboratory braided rivers [1, 2] provides further insight into some of the morphological changes attendant on changes in sediment load in rivers that are fundamentally in equilibrium with overall sediment load. From a practical perspective, insight into the response of a braided system to changes in sediment load would be desirable for interpreting or predicting channel responses to climate change, upstream flow regulation by dams or any other hydraulic structure, land use change or river management. Laboratory studies of this type range from true scale models, in which a valid model must exhibit geometric, kinematic and dynamic similarity

between model and prototype as expressed by dimensionless ratios [3], to a more liberal approach where models are viewed as small prototype channels [4]. Moreover, [5] argues that geological models could meet similitude criteria based upon similarity of processes, provided that gross scaling relationships are met and the model reproduces morphologic characteristics of the prototype, which are produced by similar processes in both model and prototype. The formal Froude modelling procedure in engineering and geomorphology in which the theorems of similarity mechanics observed in studies performed as a rule with water, offers certain advantages. A great number of river problems have been solved using this approach during recent decades. The major advantage of this type of model is that the results can, to some extent, be scaled to compare to the field prototype to validate the findings. However, this does restrict the applicability of the model, as it is limited to one field prototype and the results cannot easily be scaled to other areas. Nevertheless, an increasing number of problems have been encountered for which the solution obtained by this method is either insufficiently accurate, too time consuming, or too expensive. A typical example is modelling of braided rivers. Models usually need to be verified


216

NCSTI-2012

using prototype data. One of the problems with this is total braided river flood plain width. Although part of the problem may be attributed to difference in bank strength and difference in flow history (time of model rum)[6], laboratory models are often limited by flume walls to describe the complex geometry of braided rivers or it should be accepted that lateral development of braids will be limited by the flume width. This may not by itself pose a big problem for some braided rivers as flow rarely fills the full width of the braid plain and the active width, where much of the changes is occurring, is much smaller than the full braid plain width[7]. However, in extreme discharges which are very important because of their high transporting capacity, the braided rivers usually occupy the full bed width[8]. In most of these cases even if the planform geometry is to some extent statistically similar to the prototype, it is hardly possible to exactly reproduce the geometry of a given braided river reach[8]. Morphological modelling or assessing changes in downstream river morphology as a result of flow obstruction or any hydrological disturbance and long profile evolution data are very difficult to compare. Undistorted Froude scaled models require similar slopes in models and prototype. Practically there is limitation in flume size, sediment management and cost because of the requirement of rough turbulent flow in Froude scaled models. Comparison of sediment transport processes is also another challenge in Froude scaled models. The considerable variability of bed load transport in braided rivers made it less meaningful to compare instantaneous bed load transport in model and prototype[8]. Direct measurement of bed load transport in the field is an almost impossible task due to the inherent temporal variability and errors associated with the sampling devices although it is easier in laboratory flumes. This often limits model scales to a linear scale of not more than 1:50, restricting modelling to typical reach of smaller braided rivers [8]. All this often limits the application of the Froude scaled models and in recent days the desire to at least reduce the cost and time of model testing methods, with reasonable accuracy comparable to Froude scale modelling, has been one of the basic motivations to develop methods of investigation in a smaller scale (micro-scale) without much attention being given to most of the dimensionless numbers. With this in mind, the purpose of the present laboratory study was two-fold: to investigate the application of micro-scale physical model in braided river studies and to use this modeling approach and provide some insight into the effects of changes in sediment supply on braided river morphology.

II. EXPERIMENTAL DESIGN

This project uses a generic micro-scale hydraulic model to observe changes in channel morphology and sediment transport with changes in controlling variables. The model is not scaled to any specific prototype. Such experimental systems are based on the assumption that aspects of a natural system can be reproduced in a laboratory setting and that the processes responsible for producing features in nature will be similar to those operating in the laboratory environment. The basic requirements for these experimental systems are that a number of gross scaling relationships be met, that the system reproduces some morphological characteristic of the landform in question, and that the processes that produced this characteristic in the experiment can logically be assumed to have the same affect on the original landform [9].

Although the experimental River is based on the ‘similarity of processes and performance’ approach, to assist interpretation with respect to natural rivers it is useful to review briefly the initial and boundary conditions employed. In case of application of micro-scale modelling for a specific site, the approach to be followed as given by [10] involves adjustment of channel bed slope and water discharge to attain a chosen level of sediment transport intensity in the model. Once the model has reached this level of transport, it will be calibrated and a base test will be conducted. Deviation of all successive runs from the base test is assumed to occur in the prototype [11]. The approach taken in this study entails a programme of flume experiments that does not follow a formal framework of similitude consideration. Although this experiment is based on the use of generic micro-scale physical model and not fully scaled to a prototype, an utmost effort was put to ensure realistic representations of sediment grain size and water discharge in terms of capability to transport sediments and formation of important morphodynamic phenomena. Based on this, simple scaling relations were used to test and select model input conditions. Published data on grain-size distribution of gravel-bed Rivers is considered to have an order of magnitude of the model grain-size[1, 12].Several trial experiments were also run to arrive at those experimental conditions and ensure their suitability.

The parameters for this project are calculated for sand with grain size distribution varying from 0.25mm-0.71mm (d50~0.47mm). This size range is relatively narrow compared with the sediment size characteristics of natural gravel bed Rivers, which often have a range of grain sizes extending from boulders to silt. The decision was taken to exclude


217

NCSTI-2012

the silt and clay sized fractions, which would be required to represent finer (sand) sizes present in natural Rivers, because inter-particle cohesion in these fractions limits the capacity for these fractions to represent non-cohesive sediment. In other studies, this problem has been addressed by using light-weight sediment like crushed coal and other non-cohesive fine sediment to represent the finer non-cohesive fractions [13] and [14], however, it was considered that omitting fine sediments was an adequate solution for this research. The experimental sediment grain size and water discharge relationship is established mainly using the Shield’s theory [15] for initiation of bed materials. Even if there is debate on the scatter of the data and the interpolation of the trend line in the original Shields diagram [16], former research has shown that it is possible to create a completely movable bed for τc (critical shields parameter) more than 0.056. The selected critical bed shear stress will vary depending on the hydraulic conditions in the flume. So, to create a moderately moving bed for d50 = 0.47mm, critical bed shear stress (τc) values are chosen (for water at 20oc) from the Shield’s diagram [15]. Flow depth is calculated using the uniform flow approach and roughness is estimated using a Keulegan type roughness estimator. Based on this and a preliminary channel width of 0.2m, the threshold discharge is calculated. The channel dimension is chosen to just accommodate the imposed discharge and keep the influence of the flume walls to a minimum later in the experiment when the river braids and expands in width. This approach gives order of magnitude estimates of model parameters to be used under a specified condition of grain Reynolds number. Experimental discharge values are selected based on this reference and slightly refined based on prevailing experimental conditions and characteristics of available water pump. The work for this project was carried out in the University of Exeter Experimental Landscapes Facility in Table 1 Sequence of the experimental series.

Figure 1 Gradation curve of the sand used in the experimental model runs.

Figure 2 The experimental apparatus.

a multi-purpose contemporary terrain modeller (sedimentation tank). The purpose-built sedimentation tank, has dimensions of 5m length, 2.7m clear internal width and 1m depth. Rails running longitudinally on either side provide location and measurement datum to within +/-3mm under any loading condition. A false floor was constructed up to 0.75 m depth to reduce the depth of sand required to 0.25 m. The tank is pivoted at its lower edge and can be angled by a maximum of 150 from the horizontal. The tank is equipped with four 150mm wide cuts to simulate lateral flows if required and if not there are plates to cover the cuts. To facilitate base level change studies, the elevation of both the upstream run-on plane/inlet channel and the vertical downstream weir can be adjusted remotely using the hard-wired computer controller with adjustable vertical height for base level change studies (see Figure 2). Bed topography can be measured by a laser micro-topographic scanner which can traverse the whole 5m length of the tank along a carriage mounted on a high accuracy rail system. The scanner measures output bed topography to a resolution of 1mm in cross stream (y) and downstream (x) directions and 0.1mm in elevation (z). The carriage and laser were both automatically controlled by software. Water is pumped to the upstream inlet from a smaller water tank connected with an online tap avoiding the labour work of filling the tank. The tank has a float controlled inlet valve so that a constant pumping head could be maintained. Sediment is supplied


218

NCSTI-2012

of the basin a plastic gutter

t the start of each experiment the sediment and water

III. RESULTS AND DISCUSSION

A) Variability and Changes in sediment output

ts taken at the outlet of

at a controlled rate to the upstream inlet using a variable speed gravity-fed sediment hopper. The sediment drops into the inlet groove made of steel where it mixes with the water, being fed at a specified discharge, before reaching the basin. At the point where the mixed water and sediment drop vertically onto the flume surface, smaller gravel is placed on the surface to dissipate the energy of the incoming water and thereby to prevent bed disturbance. Potassium permanganate solution was added regularly to the flow to facilitate flow visualization and identification in imagery. At the downstream end conveyed both the liquid and sediment flux to a constant head overflow tank. The tank rested on a high precision laboratory scale (accuracy 0.1g) which was used to measure the cumulative weight of exported sediment at 5 minute intervals. These data were used with observations of channel width at the downstream flume to derive sediment transport per unit river width. But both sediment and water is not recirculating in the system. Adischarge rates were set to the predetermined values and held constant throughout the experiment, unless the experiment dictated otherwise. An order of magnitude of sediment feed rate to use in subsequent experiments was arrived at after a period of time during which the sediment collected in the overflow tank was monitored. The aim was to achieve and maintain a balance in feed and transport rate and to avoid long term aggradation at the entrance of the river channel. This experiment (RunS1) was carried out with a discharge of 3.5 l/min and gave a feeling of the amount of sediment transport to be expected in subsequent experiments. However, this does not imply that the channel is in absolute equilibrium as there was no long term monitoring of either longitudinal slope or channel configurations. There seems to be a small variation in sediment feed rate in the long term which proved to be very difficult to quantify. Accordingly, an effort was made to ensure that it is as constant as possible by checking the feed rate regularly and top-up continually to keep the level of sediment in the hopper constant.

and Channel storage Sediment transport measurementhe flume are displayed in Figure 3. When sediment output was first measured during the period when a stable braided channel was being established, a time lag of about 50min to 60min was observed before the output rate began to increase. Once this occurred, sediment output immediately increased and the rate was never constant. For the remainder of the run, sediment output in RunS1 fluctuated with greater amplitude until approximately 500min. but after that it became relatively consistent with minor fluctuation. Excluding the initial phase and from the exponential fit in Figure 0 3, there

seems to be a gradually decreasing average transport rate during the development of the braided network. At the beginning of the first aggradation cycle (from 0 to 275min in Figure 3, RunS2), the transport rate was generally higher than the transport rates observed in the later stages of the previous degradation run, with a mean of 0.205g/s, which was nearly equal to the sediment feed rate of 0.2 g/s and statistically indiscernible with the mean transport rate of RunS1 (samples within the same standard deviation). Channel storage was almost constant during this time. However, as channel aggradation progressed, sediment outputs reduced gradually and channel storage suddenly stepped-up and began to increase continuously (Figure 3). Observation of the changes in the long profile and cross section of the main channel (Figure 7C, D, L) shows that at the beginning of the first aggradation cycle (S2), the sediment input was probably stored in the actively incised upstream reach (section between 4m and 5m upstream of flume outlet). Around the middle of the flume, there was some erosion but this is intensified further downstream close to the flume outlet (1m upstream of flume outlet).The mean sediment output rate for RunS2 excluding the first 275min. of the flume run was 0.148 g/s, which is well below the feed rate. When sediment feed ceased during the second degradation (S3), channel storage reduced abruptly. Similar sediment output continued despite cessation of sediment input at the upstream end of the flume. The mean output rate for RunS3 is 0.138 g/s, which is not very different from 0.148 g/s, as measured in RunS2 excluding its initial phase. With an increase in sediment feed again to 0.2 g/s, sediment output reduced for thefirst 340 min. and varied with lesser amplitude having a mean of 0.094 g/s.

Figure 3 Measured sediment transport as a function of time throughout all runs.

But, when sediment feed continued, the output started to increase slightly with a mild fluctuation around a mean of about 0.105 g/s and standard deviation of 0.042 g/s. Moreover, as channel aggradation continued, sediment storage increased at a steeper rate than during the previous aggradation cycle (RunS2). During the last degradation cycle (RunS5), similar sediment output continued for about 360 min (around 5300 min in Figure 3). The mean output rate in this period was 0.088 g/s. With continued degradation and cessation of sediment feed from upstream, a slight increase in sediment output is observed. The overall mean of this last degradation cycle is 0.129 g/s.


219

NCSTI-2012

d to increase ightly towards the end of the experiment.

me despite a change in sediment pply conditions.

e one used in this e

Overall, a positive relationship between sediment output from flume and channel storage was observed through the experiment. Although the relations between sediment output and channel storage in RunS2 and RunS4 displayed similar styles, the magnitude and patterns of response differed early in the cycle. At the start of RunS2, when sediment feed was increased, storage was almost constant while output was steadily increasing. In RunS2, storage then increase steadily while output rates fluctuated around 0.15 g/s. When feed was stopped in RunS3, storage continuously decreased as output remained in a similar state as in the previous run. There is a large loss of storage volume of around 10,000g during RunS3. In RunS4, storage increased greatly as a slight reduction in output was observed up to around 3900min in Figure 3. After this time, sediment output remained fairly high until the end of this run even though storage continued to increase. The sediment storage in the second aggradation cycle (RunS4) increased at a much steeper rate than the first aggradation cycle (RunS2). This may be due to the fact that sediment storage may persist long after aggradational events and channel responses may be affected as a result of repeated cycles of aggradation and degradation [17]. At the beginning of RunS5, transport rates were the same as in RunS4, as storage declined and the channel bed started to degrade. This decline in storage continued up to the end of the run and transport rates also appearesl When sediment input ceased in the degradation runs, the channel actively incised and channel morphology changed to more of a single thread near the flume inlet and remained braided close to the outlet. However, there was not significant variation between the mean sediment output rates for the degrading runs (S1, S3 & S5) and mean sediment output rates of aggradation runs (S2 and S4). This suggests that transport rates respond weakly to sediment supply variations and hence sediment transport appears to be more closely related to transport capacity than sediment supply. This is in agreement with flume experiments of [18] but counter to the experiments of [17], who observed high sediment output rates for aggradation runs when the sediment feed rate was high. In fact, the mean sediment output for the first aggradation cycle (S2) was higher than the next degradation cycle (S3), but this is again changed and mean sediment output rate went further down for the second aggradation cycle (S4). The sediment output rate was partly restored in the last degradation cycle (S5). There is also a tendency for transport rates to remain similarly for sometisu Temporal variability in bed load transport rates has been observed in various laboratory studies of braided river dynamics [1, 2, 12, 19, 20] and field studies e.g. [21, 22]. This temporal variability in sediment transport was attributed to the downstream migration of bedforms at different temporal scales, migration of large bars or groups of bars and localized incision [23, 24] and has

often led to difficulty in measuring both total and local bed load transport in braided rivers e.g. [25]. Assessment of the time series shown in Figure 3 confirms that bed load transport rates are variable and shows a series of fluctuations. Although within-run transport rates sometimes vary from close to zero to just twice the mean rate, the extent of variability observed here is much lesser than in previous investigations e.g. [2, 19, 20, 26]. This is probably due to the fact that most previous investigators used much higher water discharge leading to higher transport capacities and hence possibilities of large scale incision or bank erosion. At smaller discharges (similar to thexperiment), the sedimentlower with less variability [27

transport capacity will b].

Figure 4 Histograms of sediment transport rates.

The original time series is used to derive the histograms shown in Figure 4 to show the variation of bed load transport rate within each run. A close look at the histograms shows that they are slightly skewed towards the lower transport rates (S3, S4, and S5) with the exception of RunS1. RunS1 is almost symmetrical with most output values varying around the middle and dying off out near the tails. Mean transport rate decreases from RunS1 to S4 and is restored slightly during RunS5. Transport rates in RunS2 are heavily influenced by RunS1 in that more than 50 samples collected were all relatively high and close to the mean transport rate of RunS1, which changes the shape of the histogram.

f the correlation between

Excluding this initial phase, the shape of the histogram of RunS2 would have been very similar to the later runs. Similarity of sediment output rates over a period of time despite a changing input condition (as was observed in almost all of these experiments) is an indication of ‘’persistence’’ (a tendency for a system to remain in the same state from one observation to the next). This was found to be a very common characteristic of the bed load transport processes in braided river models [20]. The best way to describe this statistically is using an autocorrelation function, because visual assessment from the time series is too subjective. The autocorrelation function provides a measure otransport rates in the series at positions separated by a time interval along the series.


220

NCSTI-2012

our is common to all runs. eriods ranging from 2 to 10 hours were observed in

].

Auto-correlograms of the time series of bed load output rate shown in revels strong and positive autocorrelation with significant periodicity in the fluctuations of bed load transport rates. This is observed in almost all runs. The interpretation of this is that there is a greater persistence and some statistical dependency on the previous transport rates of the series as positive departure from the mean is followed by positive departure from the mean and vice versa. The frequency of periodicity varies between runs. However, a peak in the autocorrelation function at lags of about 1 hPbraided river models of [19

B) Changes in channel pattern and morphology The photographs taken at the end of each run

shown in Figure 6 illustrate changes in channel p

fact that b

ge downstream, where the channel was u

an

f others e.g. [2] as a re

co

lanform during aggradational and degradational cycles.

Although quantitative evaluation of the intensity of braiding proved difficult due to the narrower nature of the channels, visual inspection of the channel in each experiment and investigation of the DEMs demonstrated that aggradation and sediment storage changed the channel morphology in each aggrading run by increasing the number of braid bars and widening the active channel itself. Similar morphological changes have been reported to some sections of laboratory braided channels as a result of bed load pulses moving into the reach [1]. The magnitude of aggradation and degradation varied throughout the length of the flume, but it was greater near the flume entrance and there were minimal changes near the outlet due to the

ed level is fixed at the downstream end, hence damping out any major elevation changes.

Figure 7shows typical cross-sectional profile at various distances down the flume during all the runs and illustrates the changes that occurred throughout the experimental run, and how this varied with time. Generally, the further downstream the cross section is located, the less channel change in terms of degradation and aggradation occurred. Figure 7(A,

B, K) represents cross sections taken at the channel bed before sediment feed began and a braided channel had formed. Major changes in the first run include considerable channel widening, degradation of the channel near flume inlet and deposition of eroded materials in the channel midway down the flume. The channel were transformed to single-thread system in the upstream reach and remained braided in the downstream reach. This may be attributed to the fact that the amount of sediment supplied from the rapidly degrading and widened upstream reach resulted in an increase in sediment dischar

nable to transport all the sediment supplied to the reach.

Consequently, despite the absence of sediment feed to the upstream end of the channel, the downstream part continued to aggrade.During the second run with sediment feed (RunS2), the thalweg and active channel bed filled at the upstream cross sections and the channel widened by more than 10cm (Figure 7c). Channel fill and widening midway down the flume at this time was insignificant (Figure 7d). Moreover, as can be seen from the plot of long profile evolution Figure 8, thalweg elevations in the downstream section remained well below elevations during the previous degradation run. Aggradation was greatest near the inlet, associated with development of new braid bars which forced the flow into the channel banks

Figure 5 Auto-correlation functions of the original sediment transport time series. Dashed line represents 95% confidence

d resulted in lateral erosion and an increase in overall channel width by more than 25% to 0.74 m.

This value was more or less consistent throughout the length of the channel. The CBRI (cross-channel bed relief index) dropped from nearly 20.1mm at the end of RunS1 to 11.7mm because of deposition in the active flow channels apparently reducing the relief between channel beds and bar tops. This is in agreement with [2, 17]but against the laboratory braided channel experiments of [18]. [18] attributed the different responses measured between their experiment and those o

flection of different rates and magnitudes of aggradation.

It is obvious that the bed relief index increases when a braided channel is degraded as a result of a rise in the difference between the incising channel bed and stable bar tops but [18] also illustrated the greater number of braid bars that formed during their aggradation experiment contributed to an increase in BRI. They found that the bed relief index could increase if aggradation occurred over the entire cross section and produced new bars. The BRI may remain constant or decrease if agradation is confined to the channels thalwegs. [18]

ncluded that both aggradation and degradation can increase the BRI, although for different reasons.

During RunS3 (Figure 7, E, F and M) the main channel continued to migrate and shifted to the right in the upstream parts of the channel. By the end of

limits.


221

NCSTI-2012

the run, it had migrated by more than 35cm, while at the same time the thalweg deepened by more than 1cm close to the upstream end of the flume. Degradation appeared to have been concentrated in the upper reaches of the channel and terraces appeared to have been formed. Towards the middle of the flume the channel migrated slightly to the left and the thalweg somewhat filled up. However, further downstream thalweg bed elevations remained below those in RunS2 for most of the reach. Mean morphologically active channel width, which is defined as the parts of the channel actively changing its elevation by erosion and deposition, declined and CBRI increased to almost pre-RunS2 conditions. As in the case of degradation in the first run, the channel here also experienced a transformation towards a single thread pattern in its upstream reach. More t

Channel morphological characteristics of the perimental series.

Fi

to increase throughout the flume, although many those channels wershallow.

uns in the flume, measured at 1000mm (K

han half the length of the part of channel remained braided in the downstream

the flume. Table 2ex

gure 6 Photographs of channel pattern taken towards the end of each Run.

Initially, the channel response to the second period of sediment feed was similar to that during the first, with a small amount of widening and minor channel fill (Figure 7, G, H and N). But as sediment feed progressed, the channel completely filled in. A relatively large mid-channel bar formed near the middle of the flume (from 1000mm- 2500mm downstream the flume entrance), dividing the channel and hence the flow. The CBRI declined from 19.4mm at the beginning of RunS4 to 14.2mm at the end of RunS4. The number of channels appeared

e relatively narrow and

Figure 7 Typical cross-sectional changes during aggradational and degradational r

,L,M,N,O), 2500mm (B,D,F,H,J) and 4500mm (A,C,E,G,I) upstream of outlet.

The most significant channel change occurred in RunS5. Incised channels formed on the true left of the flume towards the upstream end and on the true right towards the middle of the flume. The depth of incision in this channel decreased in the downstream direction as can be seen from the long profile evolution. Upstream the channel was actively incising (Figure 7I) and by the end of the experiment it had incised by more than 2cm. Mid-flume degradation appeared to be concentrated on the true right side of the channel with a reduced magnitude of about 1cm. However, further downstream channel changes did not appear to be very significant owing to the effects of the fixed bed elevation at the flume outlet. The CBRI rose to its highest value in the experiment of 26.1mm. Channel bed slope was reduced significantly, but mean sediment transport rate increased, although it


222

NCSTI-2012

maintained smaller than in the corresponding d

w. During aggradation runs, a greater proportion of the

relative to

gth indicating that the channel is unable to recover from the disturbances (change in sediment supply).

Figure 8

egradation runs without sediment feed. Differences in channel form exist between the

five experimental runs, with the most noticeable changes occurring in the upstream one-third of the flume. Channel pattern in downstream areas remained similar for all the runs. In runs with sediment feed, cross section elevations clearly demonstrate the aggrading nature of the channels. In RunS2, the magnitude of aggradation varied from around 1.7cm (measured close to flume inlet at X=4500mm) to almost nothing (measured around mid-flume at X=2500mm). Small amounts of degradation (about 0.6cm) were measured close to flume outlet at X=1000mm. Overall, amounts of aggradation were almost three times as great as amounts of degradation. The trend was the same for the second aggradation cycle (RunS4) though with a lesser magnitude. Absence of sediment feed essentially reversed the trend. Most parts of the upstream flume reach were degraded; the highest degradation of around 3cm being for the last experiment (RunS5), measured at X=4500mm. Further downstream changes were reduced in magnitude. The notable difference between RunS5 and RunS3 was that in RunS5 degradation was highest and continuous throughout the length of the flume except at the very downstream end. Moreover, degradation was concentrated over less than 50% of the channel bed occupied by flo

bed accommodated changes in sediment volume.

C) Development of longitudinal profile and channel bed slope

Figure 8shows the development of the longitudinal profile throughout the course of the experiment. The channels observed here typically exhibited convex upward profiles marked by several distinct inflections. The slope of the channel declined as the sediment feed rate ceased for RunS1, RunS5 and to some extent RunS3. Longitudinal profiles of rivers are most commonly concave upwards. However, many dry land rivers are either less concave than those located in humid temperate areas [28] or upward convex [29]. This may be attributed to the fact that the ratio of sediment to stream flow often increases downstream due to transmission losses [28]. This means that sediment is transported by progressively less flow, leading river bed aggradation and development of convex profile. Likewise, when more sediment flows out of the reach than is fed in, the channel is forced to degrade. This transient period of degradation at the upstream boundary assisted by the transmission losses and inability of flow to transport sediment, will force the profile of the channel to be upward convex. In the current experiments, incision was

very rapid and intense for degradation runs in the upstream part of the channel. Thus, even when sediment feed stopped, the downstream reach continued to aggrade because of continued delivery of sediment from the rapidly degrading reach in the upstream portion of the channel [30]. This is probably what happened especially in RunS1, RunS3 and to a lesser extent in RunS5. [31] also modelled and analysed the effects of major variables like changes in water discharge, sediment discharge, sediment caliber on the long profile evolution of both sand and gravel beds by holding the other variables constant. They observed that downstream variation in sediment discharge

flow discharge has strong potential to influence profile form and produce convex profiles.

Fill and scour of the bed varied for short lengths of the flume during aggradation and degradation runs. Most changes were accommodated between the flume entrance and middle point. Changes in thalweg elevation from 2.5m to the outlet were very minor. During the first aggradation cycle (RunS2) bed elevation increased and most changes in elevations were accommodated at the entrance of the flume (from 3.5m up to the inlet) while the rest of the flume exhibited little change. During all the degradation cycles erosion largely contributed for most of the changes in bed elevation especially from entrance to mid-flume. At the end of the final degradation cycle (RunS5), the bed elevation remained lower than the original for most of the flume len

Longitudinal profile of channel thalweg measured at the end of each Run.

The continued delivery of sediments from the actively incising upstream reach during the degradation experiments reduced the capacity of the channels to transport sediments further downstream. This has kept bed elevations in the downstream portion of the channel to remain high although sediment feed has stopped (Figure 8). Examining the evolution of the longitudinal profile over the course of the experiment on Figure 9 two distinct areas can be noted (indicated by a dashed line in Figure 9A, B and C). Area 1 represents the section where the long profile of the degrading channel lies completely below the following or previous aggrading channel (in most of the cases this is located close to entrance of the flume). Area 2


223

NCSTI-2012

(Table 2), leading to migration of the

ling variables (w

from regression of thalweg elevations of cr

ge in value between 0.020 and 0.027;

cy

corresponds to the section where the long profile of the degrading channel lies above and/or below the following or previous aggradation phase. The two regions are separated by an inflection point: the point that separates the rapidly degraded upstream area from the braided downstream reach as reported by laboratory experiments of [18, 30]. This point also seems to migrate through time as illustrated in Figure 9. Initially, the division between the rapidly degrading upstream zone and the aggrading braided zone occurred close to the flume inlet (4m from the flume outlet as indicated in Figure 9A). The channel was then rapidly filled by sediments during the first phase of sediment feed. Intensive filling of this rapidly degraded zone has occurred for the first 500min and after that only minor changes happened in this reach. The inflection point seems to move further downstream with cessation of sediment feed in RunS3, though it had only advanced not more than 50cm from its former location (Figure 9B). However, as the gradient continued to decrease in the upstream degrading zone (see upstream section of RunS5 in Figure 8) bed load transport also reducedinflection point further downstream closer to the outlet. Very similar results have been reported by [30]. In the experiments of [30] there is a clear trend that indicates the downstream migration of the inflection point as the channel continued to degrade. However, in this case it appears that phases of aggradation and degradation are alternating in the downstream section, with the exception of RunS1-S2, where in this case the long profile of RunS2 lies entirely above that of RunS1 below the inflection point (see Figure 9). This small difference between the results is due to the fact that [30] reported continuously degrading channel where as in this case continuous degradation is interrupted by supplying sediment and initiating aggradation in between degradation experiments. Moreover, there are certainly differences in control

ater discharge, sediment feed rate and grain-size) between the two sets of experiments.

Several researchers have studied relationship between channel pattern and longitudinal slope both quantitatively and qualitatively [4, 32-35]. Physical model studies have also been used in the past to demonstrate the existence of pattern threshold related particularly to slope [36]. In general, for a certain discharge, braided streams are characterized by greater slopes than meandering streams. Different channel patterns can also occur at different sections of the same river depending on longitudinal slope and sediment characteristics. [37] observed an abrupt change in channel pattern of the Red river from non-braided to braided while in that same section the longitudinal slope doubled from 0.034% to 0.068%. Similarly, [4] also observed a change in channel pattern in Horse Creek from

single thread to braided where the longitudinal slope changed from 0.0022 to 0.0073. In experimental series 1, channel bed slope is calculated

oss sections taken every8mm in the stream-wise direction.

Furthermore thalweg elevations of channel cross sections acquired at 25cm longitudinal interval are also used to calculate temporal variation of longitudinal slope within an experiment. Channel slopes calculated by regression of thalweg elevations ranthe maximum being attained in the first aggradation

cle Runs2.

Since bed elevation is fixed at the downstream end of the flume, significant increase in bed elevation occurred in the upstream end of the flume. For all degradation runs a break in slope occurred together with a change in channel pattern. This was greatest in RunS1 and RunS5. In RunS1 there was a significant change in slope (from 0.002 to 0.028) when the channel pattern changed from single thread to braided although it was only in the upstream 1m of the channel that this occurred (Figure 9A). In RunS3, however, the change in slope was reduced (from 0.0065 to 0.028) owing to interruption of degradation by sediment feed from upstream. Generally, absence of sediment feed and

A

B

C

Figure 9 Relationship between channel pattern, longitudinal slope and migration of inflection point.


224

NCSTI-2012

g p

increased by more than double to 0.014 before

oth of which ultimately reduced the channel slope.

V

similar to the laboratory experiments ported here.

s hown that despite the lack of dynamic similarity

ling principle ggest that the laboratory channels are at least

q e to the field.

thank the department for providing financial support for this research through an internal scholarship.

FER1.

gravel-bed streams.

2. ed rivers: a

3. ysical Modelling. in

4. s: braided, meandering, and straight: US

5. arid-

6. hology in a model

channel degradation throughout the experimental series was associated with the channel moving towards a single thread, meandering system (Figure 9) with milder slope close to the inlet. Durin

eriods of sediment feed the slope of the channel was generally steeper and the pattern was braided.

The most abrupt change in gradient occurred in the area near the flume inlet or from 3.5m to 5m upstream of the flume outlet. When the upstream reach underwent a transition from braided to single-thread (RunS2 to RunS3), it experienced a reduction in longitudinal slope from 0.017 to 0.0065. When sediment feed was restarted in RunS4, the channel pattern in the same reach changed to braided and its longitudinal slope

finally attaining the minimum slope of 0.0046 in the course of the experimental run. In real terms, the increase in slope translates into aggradation and development of multiple threads at least for this experiment. In response to the cessation of sediment feed from upstream, an increase in transport capacity is observed which resulted in vertical incision with different magnitude and channel migration through bank erosion in all degradation runs; b

I. CONCLUSION The morphological changes reported in this

experimental series were mostly similar to those seen in previous laboratory studies of braided channels e.g.[1, 2, 17, 18, 30]. In most of these studies the braiding intensity, number of braid bars, pattern complexity and morphologically active braid plain width all increase when a channel experiences aggradation. The same relationship is recorded in this experiment. However, unlike other similar laboratory experiments the bar size observed here do not show significant differences between aggrading and degrading channels. Degradation essentially reversed the morphological changes associated with aggradation. Initiation of degradation by stopping sediment feed resulted in vertical incision close to the flume inlet that sometimes extended to middle of the flume. Flow became concentrated into a single channel flanked by erosional terraces that ultimately left a greater proportion of the active channel exposed. This is the most prominent feature of the evolution of sand bed laboratory channels. However, the downstream portion of the channel did not degrade due to the resultant influx of sediment arriving from the actively incised reach upstream. This is also in agreement with the laboratory experiments of [30] and [18]. The experiments of [30]have also been compared with Ash Creek, a degrading ephemeral- flow braided channel, on the east flank of the Mazatzal Mountains in central Arizona. The stream was degraded due to a reduction in sediment delivery as a result of re-vegetation of the drainage basin. The channel responses to this reduction in sediment yield included incision progressing from upstream to downstream and continued aggradation downstream, maintaining the braided pattern. These trends are veryre The experimental work presented in this paper has

conditions and simplification of overall similarity criteria, fairly consistent results could be obtained which can be interpreted in a generic sense. The similarity between the laboratory channels from this experiment and those previously investigated by different researchers with or without a field prototype utilizing the Froude modesu

ualitatively transferabl

Acknowledgements This work formed part of a PhD project at the

Department of Geography, University of Exeter, United Kingdom and Michael would like to

RE ENCES Ashmore P: Channel morphology and bed load pulses in braided, Geografiska Annaler Series A Physical Geography 1991:37-52. Hoey TB, Sutherland AJ: Channel morphology and bedload pulses in braidlaboratory study. Earth Surface Processes and Landforms 1991, 16(5):447-462. Shen HW: Principles of PhShen, H W, ed, Modelling of Rivers: New York, Wiley, P6-1 to 6-27 1979. Leopold LB, Wolman MG: River channel patternGovernment Printing Office Washington (DC); 1957. Hooke RLB: Steady-state relationships onregion alluvial fans in closed basins. American Journal of Science 1968, 266(8):609-629. Warburton J: Active braidplain width, bed load transport and channel morp


225

NCSTI-2012

7.

river. New Zealand, Geological

8. modelling of braided gravel-bed

9. al fans: Effect of discharge and sediment

10. icroscale loose-bed

11. small-scale movable bed model.

12.

13.

14.

15.

bewegung. Mitt Preuss

16.

ations and unresolved issues. The

17. p

nt modeled after Redwood

18. gy resulting from aggradation

and degradation. The Journal of geology

20. ed

21.

22. ry of the

23.

24. d sediment storage in gravel-

25.

sition in braided rivers. Geological Society,

26. d

27. of a cellular

28. D: Geomorphology of

29. and deposition in ephemeral-stream

chronous

31. d RL: Mathematical

32. tural streams flowing in erodible

34. der geometry

35. nnels in Kansas,

, 278.

37. ofiles and channel pattern for

the Red River. Journal of Hydrology 1977, 35(1-2):191-201.

braided river. Journal of Hydrology New Zealand 1996, 35:259-286. Warburton J, T. Davies and M. Mandl: A meso-scale field investigation of channel change and floodplain characteristics in an upland braided gravel-bed Society, London, Special Publications 75(1): 241 1993. Young WaJW: Principles and practice of hydraulic rivers. Journal of Hydrology (NZ) 35(2): 175-198 1996. Le Hooke RB, Rohrer WL: Geometry of alluvisize. Earth Surface Processes 1979, 4(2):147-166. Gaines RA, Maynord ST: Mhydraulic models. Journal of Hydraulic Engineering 2001, 127:335. Maynord ST: Evaluation of the micromodel: An extremelyJournal of Hydraulic Engineering 2006, 132:343. Young W, Davies T: Bedload transport processes in a braided gravel‐bed river model. Earth Surface Processes and Landforms 1991, 16(6):499-511. Whipple KX, Parker G, Paola C, Mohrig D: Cha nel dynamics, sediment transport, ann d the slope of alluvial fans: Experimental study. TheJournal of geology 1998, 106(6):677-694. Sheets B, Hickson T, Paola C: Assembling the stratigraphic record: Depositional patterns and time‐scales in an experimental alluvial basin. Basin research 2002, 14(3):287-301. Shields A: Anwendung der Ähnlichkeitsmechanik und Turbulenzforschung auf die Geschiebe

30.

Versuchsanstalt für Wasserbau und Schiffbau, No 26, Berlin. 1936. Peakall J, Ashworth P, Best J: Physical modelling in fluvial geomorphology: principles, applicScientific Nature of Geomorphology 1996:221-253. Madej MA, Sutherland DG, Lisle TE, Pryor B: Channel responses to varying sediment in ut: A flume experimeCreek, California. Geomorphology 2009, 103(4):507-519. Germanoski D, Schumm S: Changes in braided river morpholo

1993:451-466.

19. Ashmore PE: Bed load transport in braided gravel‐bed stream models. Earth Surface Processes and Landforms 1988, 13(8):677-695. Warburton J, Davies T: Variability of bedload transport and channel morphology in a braidriver hydraulic model. Earth Surface Processes and Landforms 1994, 19(5):403-421. Goff JR, Ashmore P: Gravel transport and morphological change in braided Sunwapta

River, Alberta, Canada. Earth Surface Processes and Landforms 1994, 19(3):195-212. Griffiths G: Recent sedimentation histoWaimakariri River, New Zealand. Journal of Hydrology(New Zealand) 1979, 18(1). Gomez B, Naff RL, Hubbell DW: Temporal variations in bedload transport rates associated with the migration of bedforms. Earth Surface Processes and Landforms 1989, 14(2):135-156. Hoey T: Temporal variations in bedload transport rates anbed rivers. Progress in physical geography 1992, 16(3):319. Bridge JS: The interaction between channel geometry, water flow, sediment transport and depoLondon, Special Publications 1993, 75(1):13-71. Bertoldi W, Ashmore P, Tubino M: A methofor estimating the mean bed load flux in braided rivers. Geomorphology 2009, 103(3):330-340. R. T: Development and evaluationmodel to simulate braided river dynamics. PhDthesis, University of Exeter 2003. Parsons AJ, Abrahams Adesert environments. Geomorphology of Desert Environments 2009:3-7. Schumm SA: Effect of sediment characteristics on erosion channels. Geological Survey professional paper 1961, 352. Germanoski D, Harvey MD: Asynterrace development in degrading braided channels. In.: DTIC Document; 1993. Snow RS, Slingerlanmodeling of graded river profiles. The Journal of geology 1987:15-33. Lane EW: A study of the shape of channels formed by namaterial: US Army Engineer Division, Missouri River; 1957.

33. Henderson F: M., 1966: Open channel flow. In.: New York: Macmillan. Ackers P, Charlton F: Meanarising from varying flows. Journal of Hydrology 1970, 11(3):230-252. Osterkamp W: Gradient, discharge, and particle-size relations of alluvial chawith observations on braiding. American Journal of Science 1978

36. Schumm S, Khan H: Experimental study of channel patterns. 1971. Lee LJ, Henson BL: The interrelationships of the longitudinal pr


226

NCSTI-2012

Intensity-Duration-Frequency (IDF) Analysis of Point Rainfall for Selected Meteorological ...

227

NCSTI-2012

Intensity-Duration-Frequency (IDF) Analysis of Point Rainfall for Selected Meteorological Stations in Illu Aba Borra Zone,

Oromia Region Elias Jemal 1, Desalegn Chemeda 2

1School of Natural Resource and Environmental Engineering, Haramaya University, Dire Dawa, P.O.Box 138, Ethiopia, [email protected]

2 School of Natural Resource and Environmental Engineering, Haramaya University, Dire Dawa, P.O.Box 138, Ethiopia,

Abstract — the annual maximum rainfall data of varying duration were extracted from rainfall

charts and fitted to theoretical frequency distribution (EVI, Log Normal and Log Pearson) to construct IDF curves for two stations in Illu Aba Borra Zone and generate IDF parameters. The most commonly used plotting position formula, the Weibull formula, was used to identify the best fitting probability distributions. The chi-square test confirmed the appropriateness of EVI and Log Normal distribution to explain the rainfall data from the two stations. Because lognormal distribution has only two parameters and simpler mathematical expression it is proposed to work for Bedele and Gore stations. Extrapolation of values for larger return periods was made. The analysis of rainfall intensities was expressed using the IDF equation of the generalized mathematical form and families of curves. The performance of this method was evaluated based on the historical and computed values of intensities using coefficient of determination. The results of this evaluation showed that the computation of rainfall intensities using the parameters derived adequately describe the observed data, confirming that the method is reliable.

Keywords- IDF curve, Weibull formula, theoretical frequency distributions

I. Introduction Shortage of rainfall and its presence in abundance

especially if not controlled have their consequences in every environment. Extreme high precipitation amounts are among environmental events with the most disastrous consequences for society. Estimates of their return periods and design values are of great importance in hydrologic modeling, engineering practice for water resource systems and reservoirs design and management, planning for weather-related emergencies, etc.

Virtually all hydraulic designs require an estimation of

extreme high precipitation events. Hydraulic structure design is based on a certain return period which is the frequency of an extreme event of a given magnitude. Accurate prediction of return period enables the designers to prescribe the most economical structure consistent with public safety. Generally, prediction of rainfall is done using isopluvial maps and Intensity-Duration-Frequency (IDF) curves. These two tools are used by engineers to design safe and cost effective structures for certain return periods.

Many structures constructed using the scarce financial

resource and borrowed capital are not functioning as expected. A number of technical and management related factors could contribute to this under performance of water resource development schemes. The technical performance is mainly due to lack of information for the design of structures. One of those design parameters

needed, but often not available, is the rainfall Intensity-Duration-Frequency relationship.

The rainfall-intensity data, which include

measurements of rainfall depth and time, are used to construct Intensity-Duration-Frequency curves. According to Nhat` et al. (2006), the Intensity-Duration-Frequency relationship is an important hydrologic tool which will bridge the gap between the design need and the availability of design information especially in planning and design of water resources projects. This relationship is used to estimate peak run-off rates for instance as in the Rational formula. The curves also used as input to rainfall-runoff models that are used to simulate large floods for bridge and spillway design, soil-erosion prevention practices and irrigation-management procedures. The development of IDF curves for precipitation remains a powerful tool in the risk analysis of natural hazards. Indeed the IDF curves allow for the estimation of return period of an observed rainfall event or conversely of the rainfall amount corresponding to a given return period for different aggregation times (Gerbi, 2006)

The specific objectives were: To establish mathematical relationships among

intensity, duration, and frequency for the study area and to develop Intensity-Duration-Frequency curves for selected meteorological stations in Illu Aba Borra Zone.

A. Characteristics of Rainfall Events

a. Rainfall intensity


228

NCSTI-2012Rainfall intensity is defined as the time rate of rainfall

depth or the ratio of the total amount of rainfall (R) (rainfall depth) falling during a given period to the duration of the period (D) and is commonly given in the units of millimeters per hour. All precipitation is measured as the vertical depth of water (or water equivalent in the case of snow) that would accumulate on a flat level surface if all the precipitation remained where it fell.

Intensity is the most important of the rainfall characteristics. All other factors being equal, the more intense the rainfall, the larger will be the discharge rate from a given watershed. Intensities can vary from misting conditions to precipitation which may fall to cloudbursts where several centimeters per hour are common. The average intensity (I) can be expressed as:

DRI (1)

Where R = rainfall depth (mm), D = duration (hr).

b. Rainfall duration or time of rainfall

The rainfall duration can be determined from either cumulative mass rainfall curves or rainfall hyetographs. In the first case, the duration is the time from the beginning of rainfall to the point where the mass curve becomes horizontal indicating no further accumulation of precipitation. In second case, the rainfall duration is simply the width (time base) of the hyetograph. The most direct effect of rainfall duration is on the volume of surface runoff, with longer rainfall producing more runoff than shorter duration rainfall of the same intensity.

c. Return period

Return period (T) is a term commonly used in hydrology. It is the average time interval between the occurrence of storms of a given magnitude and duration. The exceedence probability (p) is the probability that a storm event having the specified duration and volume will be exceeded in one given time, typically in any one year. The exceedence probability (p) and return period (T) are related by:

T = 1/p (2)

The selection of the return period for design will depend on the relative importance of the facility being designed, cost (economics), desired level of protection, and damages resulting from a failure. Typical design return periods for storm sewer conveyance in Iowa (inlets and piping) vary from 2-10 years, with 5 years being most common. For culverts, design periods of 25-50 years are typical, depending on the type and level of service for the roadway. For detention basins, 25-100 years are common (CTRE, 2007).

The commonly used return periods such as for United

States Isohytal maps are 2, 5, 10, 25, 50 and 100 years (Faiers et al., 1997).

B. Rainfall Frequency Analysis

There has been a considerable interest, especially amongst engineers, in the problems of estimating the frequency with which a given amount of intense rainfall may be exceeded. We require knowledge of the characteristics of intense rainfall for a variety of design purposes. The economic design of bridges, culverts, dams and other hydraulic structures demands knowledge of the likely floods which the structure would have to withstand during its estimated economic life (Bell, 1969). Since floods are primarily due to intense and prolonged rainfall (except in the cold regions where snowmelt is an important factor), major attention is given to the estimation of the magnitude/frequency relations for rainfall. From these relations the corresponding floods can be calculated by various methods.

C. Forms and Definitions of Probability Distributions

A probability density function is defined by its characteristic form (normal, Log Normal, etc.) and by its parameters. Some distributions commonly used in frequency analysis (N, LN and EV1) have two parameters and others (LN3, GEV, W, P3 and LP3) have three. Some analysts have used distributions with as many as five parameters. In general, the greater the number of parameters, the better is the fit that can be obtained to a single data series. But this does not necessarily mean that the underlying population distribution is better represented.

A probability distribution can also be characterized by its mean (µ), variance (σ2) and coefficient of skew (γ). For some distributions, these statistics are identical to the parameters in the probability density function, but for others they are not.

Future floods cannot be predicted with certainty.

Therefore, their magnitude and frequency are treated using probability concepts. In a similar way, the analysis of rainfall data for computation of expected rainfall of a given frequency is commonly done by utilizing different probability distributions.

Thus, for determining extreme flood events, specific

extreme value distributions are assumed, and the required statistical parameters are determined from the available data from which the flood magnitude for any specific return period can be determined. The basic statistical equation is given by:

TKXX (3)

Where X is any variable, X means of variate, standard deviation of the variate and K, frequencies factor (a constant). This equation has been shown by Chow (1992) to be governing equation for almost all theoretical frequency distribution functions.

Thus, some of common probabilities models for

extreme value function have been developed by different


229

NCSTI-2012scientists; and can thus be well developed by the above equation (3). Among the different probabilistic models the Normal, log Normal, Gumbel’s and log Pearson type III distributions are used for the analysis of rainfall data in order to compute the expected rainfall of a given frequency. Normal distribution

The normal distribution is a classical mathematical

distribution commonly used in the analysis of natural phenomena. The normal distribution has a symmetrical, unbounded, bell-shaped curve with the maximum value at the central point and extending from - ∞ to + ∞. For the normal distribution, the coefficient of skewness is zero. The function describing the normal distribution curve is:

5.0

2

2

2

2

SeXf

SXX

X (4)

Note that only two parameters are necessary to describe the normal distribution: the mean value, ( X ) and the standard deviation, (S).

2. Log-normal distribution

The log-normal distribution has the same characteristics as the normal distribution except that the dependent variable, X, is replaced with its logarithm (Y = log(X)). The characteristics of the log-normal distribution are that it is bounded on the left by zero and it has a pronounced positive skew. These are both characteristics of many of the frequency distributions that result from an analysis of hydrologic data. The logarithmic transformation of the normal distribution is given as:

5.0

2

2

2

2

Y

SYY

XSeXf Y

(5) 0X

Where , XY log Y and SY are the mean and standard deviation of the reduced variates of the sample. The log-normal distribution has advantage over the normal distribution that it is bounded, as x > 0 and the log transformation tends to reduce the positive skewness.

3. Gumbel extreme value distribution

The Gumbel Extreme Value frequency distribution also referred to as the Gumbel Extreme Value type I distribution (EVI) is the most widely used probability distribution model for extreme values in hydrologic and meteorological studies and has received the highest application for estimating large events in various part of the world. This distribution has been used for rainfall depth-duration-frequency studies (Garg, 1999).

The Gumbel Extreme Value type I asymptotic distribution for maximum or minimum events is the limiting mode for the distribution of the maximum or minimum of ‘n’ independent values from an initial distribution whose right or left tails is unbounded and is an exponential type. According to Reddi (2001), theory

of maximum events, the probability of occurrence (p) of an event equal to or larger than a value xo is:

yeo exxp

1 (6)

Where y is the reduced variate, given by xy , in which and are parameters of the distribution, and μ and σ are the mean and standard deviation,

28255.1 and 45005.0 .

The cumulative distribution function is given by:

yeexF (7)

It may be noted that β is the mode of the distribution point of maximum probability density and ‘x’ is the variant (historically observed data). Simplifying and solving equation (7) for y gives:

xFy 1lnln (8)

Substituting F(x) of equation (7) into equation (6) yields: pxF 1 (9)

1

lnln1

1lnlnT

Tp

yT (10)

Where p is probability of occurrence and yT is reduced variate for different return period (T). Therefore, for the extreme value distribution, xT is related to yT as follows:

TT xyxy (11) Hence, xT, the expected rainfall event is given as:

xTTT

T SKxxyx

(12)

Where KT is the frequency factor and x and Sx is the sample mean and standard deviation, respectively.

4. Log-Pearson type III distribution

This distribution is the standard distribution for frequency analysis of annual maximum rainfall events and has got also wide application in the analysis of rainfall intensities. The probability density function is given by (Adeboye and Alatis, 2007):

xx

eyxfy

log1

(13)

Where λ, β and ε are parameters of the distribution and y

= log x, Г (β) = (β-1)! ,

yS ,

22

yCS

and

ySy

The skew coefficient is determined using the expression:


230

NCSTI-2012

321)( 1

3

nnn

yynyC

n

iS (14)

Where n is number of observations and the parameters λ, β, and ε are used to compute the mean μy, standard deviation σy, and coefficient of skew CS (y) of sample estimates of the population as follows:

yy and (15)

This is referred to as the three parameter fit. Due to its performance in stochastic hydrology, it has been adopted in a number of countries as a standard distribution for flood frequency analysis.

The log-Pearson type III distribution differs from most of the distributions discussed above in that three parameters (mean, standard deviation, and coefficient of skew) are necessary to describe the distribution. By judicious selection of these three parameters, it is possible to fit just about any shape of distribution.

Frequency Factor

For given distribution, a relationship can be determined between the frequency factor (K) and the return period (T). This relationship can be expressed in mathematical terms or given in table. The value of xT of a hydrologic event may be represented by the mean (μ) plus departure of the variate (Δ xT) from the mean (Garg, 1999).

TT xx (16.a)

Assuming the departure to be equal to the product of the standard deviation (σ) and the frequency factor KT (i.e. Δ xT = σ KT) then, xT becomes:

TT Kx (16.b)

The value ΔxT and KT are functions of the return period (T) and the type of probability distribution to be used in the analysis. If the relationship of the data analysis is in the form of , the same style is applied to the statistics for the logarithm of the data using,

xy log

yT SKT and the required value of xT is found by taking the antilog of yT.

yy

The relationship between the frequency factor (KT) and the return period (T) is given for different type of distribution as follows: For normal distribution, the frequency factor (KT) is

given by the following formula (Suresh, 2005):

TT

xK (17)

Where xT is the variate, μ is the mean and σ is the standard deviation of sample data. For Lognormal distribution, the following formula

can be used for determining the frequency factor (KT) (U.S. Army Corps of Engineers, 1994; Bhakar et al., 2006):

32

2

0013.01893.04328.110103.08028.0516.2

wwwwwwzKT (18)

Where w is intermediate variable and p is probability of exceedence; w can be calculated using the following formula:

5.001ln2

1

2

p

pw (19)

When p > 0.5, 1-p is substituted for p and the value of z is computed by equation (18) is given a negative sign. For Gumbel Extreme Value type I distribution, the

frequency factor (KT) can be computed by using the following formula (Das, 2004):

n

nTT

yyK

(20)

Where yT is the reduced variate of a given return period T and ny and σn are the reduced mean and the reduced standard deviation as a function of sample size n, respectively. The values for ny and σn can be read from many hydrology texts (Reddi, 2001).

The value of the reduced variate yT for a given return

period T is given as:

1lnln

TTyT (21)

For log-Pearson type III distribution, the frequency factor corresponding to the annual maximum rainfall magnitude as (Suresh, 2005; Adeboye and Alatise, 2007):

53

1432233

12 161 kzkkzkzzkzzKT

(22)

6SCk (23)

Where KT is frequency factor used for Log Pearson type III distribution (when Cs = 0, then KT = z), Cs is the coefficient of skewness, and z is standard normal variable or frequency factor.

Probability Plotting

The probability distribution of hydrologic data can be determined by computing their plotting positions for a given length of record. The main purpose of probability frequency analysis is to obtain a relation between the magnitude of a storm and its probability of occurrence.

A number of different formulas have been proposed for

computing plotting position probabilities, with no unanimity on the preferred method. The simplest technique is arranging the event in descending order of magnitude and assigning the rank number (m) to each event. The severest event will be placed at the top with its ranking as 1. The lightest event will be placed at the last place and its ranking will also be n. The one most


231

NCSTI-2012

commonly used formula is the Weibull formula given by Takara and Stedinger (1994):

m

nTn

mxxp m1

1

(24)

Mathematical form of IDF

The IDF relationship can be expressed in the form of empirical equation rather than reading the rainfall intensities from graphs or maps. From the mathematical relationship, the IDF parameter can be estimated. The empirical formula is given in the following form (WMO, 1994):

n

r

TTbaPi

)1(log

(26)

Where Pi is the maximum mean rainfall intensity for duration of T and return period of Tr and a, b and n are parameters that vary from station to station with selected frequency of occurrence. Koutsoyiannis et al., (1998) proposed a new approach to the formulation of IDF curves using efficient parameterization equation of the form:

)( vd

wI (27)

Where I denote the rainfall intensity for duration d and ω, ν, θ and η represents non-negative coefficients. A numerical exercise proposed by Koutsoyiannis et al., (1998) shows that the errors resulting from imposing ν = 1 in equation (2.27) is much smaller than the typical parameter and quantile estimation errors from limited size samples of rainfall data. Considering the specification of ν 1 results in over-parameterization of equation (2.27), He suggested the following equation as a general expression of IDF relationships for given return period:

)(

dwI (28)

Parameter estimation methods

There are two methods of parameter estimation suggested by Koutsoyiannis et al., (1998). These are the robust estimation and the one-step least squares methods based on optimization technique. The second method estimates all parameters (θ, η and ω) in one-step, by minimizing the total square error of the fitted IDF relationship to the data. The minimization of the total error can be performed using the embedded solver tools of the MS-Excel spreadsheet by a trial and error procedure.

There are also various IDF Curve-Fitting softwares

developed to solve the parameters (a, b, and c) from a set of pairs of data values of an intensity-duration equation of the general form (EXACT, 2006):

cd bT

aI

(29)

Where I is rainfall intensity (mm/hr), Td is time duration (minute), ‘a’ is coefficient with the same unit as I, b is time constant (minute) and c is an exponent usually less than 1.

Generally, equation (29) is identical with equation (28) which indicating that the parameters ω; θ and η in equation (28) are equivalent to a, b and c of equation (29).

Description of the Study Area

The study was carried out in southwestern part of Oromia region specifically in Illu Aba Borra Zone. Bedele and Gore meteorological stations were considered for this study. The rainfall data of these stations were collected from National Meteorological Service Agency (NMSA) for durations of 1, 2, 3, 6, 12 and 24 hours. Selection of the stations was based on the fact that these stations are first class (automatic recording stations) from which rainfall intensity can be directly derived from automatically recorded rainfall chart. Basic information of the raingauge stations considered for the study is presented in Table 3.1 below. Table 3.1 Stations location (NMSA)

No

Station Code

Station Name

Latitude ( 0N )

Longitude ( 0E )

Altitude (meters)

Years of record

1 ILBEDE11 Bedele 8.27 36.20 2030 10

2 ILGORE12 Gore 8.09 35.32 2002 12

RegionsOromiaIllu-Aba-Borra

400 0 400 800 Kilometers

N

EW

S

Figure 3.1 Location of the rain gauge Stations

Types of data collected

The data from both stations (Bedele and Gore) were daily-based and contain information about the time of beginning and end of the individual rainfall events. The daily rainfall data were read for each station for durations of 1, 2, 3, 6, 12 and 24 hours directly from daily recorded rainfall charts. Since the most direct effect of storm duration is on the volume of surface runoff, with longer storms producing more runoff than shorter duration storms of the same intensity, in this study the above durations were used (Oyebande, 1982; Mohaymant et al., 2004; Cherkos et al., 2006; Raiford et al., 2007).


232

NCSTI-2012The annual maximum daily rainfall values were then

extracted from the daily rainfall data read for the selected duration. Finally, the extracted annual maximum daily rainfall values were arranged in descending order by their magnitude for constructing an array and distribution of each data in the form of frequency of their occurrence.

Fitting the probability distribution function

The extracted annual maximum daily rainfall data arranged in descending order were fitted to the probability distribution functions. The rainfall depths at the durations of 1, 2, 3, 6, 12 and 24 hours were fitted to EVI (equation 10), lognormal (equation 18) and Log Pearson type III (equation 22) using frequency analysis technique. The reason for selecting these distributions for analysis is due to the fact that they are commonly used in rainfall and flood studies (Hosking, 1986 and 1991; Ahmad et al., 1988; Chowdhury et al., 1991; Vogel and McMartin, 1991 and Takara and Stedinger, 1994).

In fitting the data to the probability distributions, the most commonly used plotting position formula, the Weibull formula (equation 24) given by (Takara and Stedinger 1994), was used to identify the best fitting probability distributions. The method of fitting the annual maximum rainfall data to the EVI, lognormal and Log Pearson probability distribution are given below.

(a) Fitting the data to EVI distribution function

The reduced variate (YT) was calculated using equation 10 (Cherkos et al., 2006). Weibull plotting positions were used in the computation.

(b) Fitting the data to lognormal distribution function

The standard normal variable, z, corresponding to the return periods of the ranked annual maximum rainfall was determined using equation 18 (Suresh, 2005; Bhakar et al., 2006; Adeboye and Alatise, 2007). Weibull plotting positions were used in the computation.

(c) Fitting the data to Log Pearson type III distribution function

Using the Log Pearson type III distribution, the frequency factors, KT, corresponding to the annual maximum rainfall magnitude was determined using equation 22 (Adeboye and Alatise, 2007). Weibull plotting positions were used in the computation.

Testing the goodness of fit of data to the probability distribution functions

The choice between the different probability distributions was made by means of chi-square ( ) test in order to model the data. The test was made if the annual maximum values of rainfall from the two stations follow either of the above probability distributions. The test was carried out for each duration, each station and each probability distribution (EVI, lognormal and log Pearson III). According to (Bhakar et al. 2006) chi-square ( ) test is a commonly used test for determining the goodness of fit and is expressed as follows:

2

2

m

i i

iiSc xp

xpxfn1

22

(30)

Where m is the number of class intervals, n fs (xi) is the observed number of occurrences in interval i and np (xi) is the corresponding expected number of occurrences in interval i. (Gumbel and Yevjevich 1972) suggest the choice of class intervals or cells such that all have equal probabilities Pi = 1/m or equal absolute frequencies, npi, in order to overcome much of the arbitrariness that accompanies the construction of class intervals. According to (Haan 1977) the number of classes should be 5 – 20 and also the class interval should not exceed one-fourth to one-half of the standard deviation of the data. Hence, the resulting calculated values of chi-square 2

c can be compared with the tabulated values of chi-square ( ). The degree of freedom, ν, is given by ν = m-p-1 (where ‘m’ is the number of intervals and ‘p’ is the number of parameters used in fitting the proposed distribution.

2

1, v

A confidence level, which is often expressed as 1-α

where ‘α’ is termed as the significant level, was chosen for the test. A typical value for the confidence level is 95 per cent. The null hypothesis for the test is that the proposed probability distribution fits the data adequately. This hypothesis is rejected if the value of is larger than a limiting value which is determined from the distribution with ν degree of freedom at 5 % level of significance. Otherwise, it is accepted.

2c2

1, v2

Computation of Extreme Rainfall Value (XT) Once the distribution of the observed data is known,

the extreme rainfall events (XT) for the selected return periods can be estimated numerically. For instance, in the Log Pearson type III and Log normal distribution, the natural values of variate are converted into logarithmic form and the computations are performed (Haan, 1977; Suresh, 2005).

The expected extreme value, XT, for any return periods can be computed by using the following relationship:

yTT SKyy XT = antilog yT SKy (31)

Where YT is the reduced variate, and Sy are the mean and standard deviation of the log transformed data, respectively, and KT is frequency factor.

The selection of return period for design depend on the

relative importance of the facility being designed, cost (economy), desired level of protection, and damages resulting from failure. The commonly used return periods, 2, 5, 10, 25, 50 and 100 years (Faiers et al., 1997), were used in this study.


233

NCSTI-2012Calculation of Rainfall Intensity (I)

The calculation of rainfall intensity was made by dividing the computed extreme rainfall value based on return period by the individual durations. Hence, intensity I can be calculated using the following expression:

i

T

DxI

(hr)duration (mm)depth Rainfall

(32)

3.6. Estimation of IDF Model Parameters

Finally, the above either extreme rainfall values (xT) in Section 3.4 or the intensity values (I) in Section 3.5 were used as inputs for the IDF Curve Fit Software (MIDUSS version 2.25) to estimate the IDF parameters (a, b and c) for the following equation 3.4 (EXACT, 2006). The parameters ω, θ and η in Koutsoyiannis general equation (28) are represented by a, b and c in the curve fit software.

(33a)

The equation can also be expressed in the following form:

I = exp [(ln (a) – c ln (b + Td))] (33b) Where, I is rainfall intensity (mm/hr), Td is rainfall

duration (minute), a, is coefficient with the same unit as I, b is time constant (minute) and c is an exponent usually less than 1.

The value of the ‘a’ coefficient depends on the return

period in years of the storm, and the system of units being used.

The constant ‘b’ in minutes is used to make the log-log correlation as linear as possible. A value of zero for this parameter represents a special case of the IDF equation where:-

cdTaI

)( (34)

In general, this results in poor agreement between observed values of intensity and duration and those represented by IDF equation.

The ‘c’ exponent is usually less than 1.0 and is obtained in the process of fitting the data to the power expression. Its values are usually in the range of 0.75 to 1.0.

Testing Model Performance

Visual comparison of the computed and observed intensities can be very useful in assessing the accuracy of the model output. However, additional statistical analysis is needed because visual comparison usually tends to be subjective. Objective assessment, however, generally requires the use of a mathematical estimate of the error between the computed and observed intensities. The coefficient of determinations, r2 between the computed and observed intensity was calculated using the following equation

2

1

2

1

2

12

n

ii

n

ii

n

iii

PPOO

PPOOr (35)

Where O is observed and P is model predicted values. The value of r2 lies between 0 and 1 which describes how much of the observed value is explained by the model. A value of zero means no correlation at all whereas a value of 1 means that the computed values are equal to that of the observed value. The closer r2 is to 1, the better the model explain the data. Nĕmec (1973) provided the values of coefficient of correlation as:

Coefficient of determinations r2 verifies the degree of closeness or association between two variables; it can also provide an answer to how well the regression line fits the observed data (Hashmand, 1997), therefore the closeness of model-predicted intensity and observed intensity was evaluated by this technique.

IV. Results and Discussion Selection of the Best Fitting Probability Distribution

Function

The selection from among the probability distributions were made by means of well-chosen chi-square (

2 ) test. The result of this test is given in table 4.1. This test was also used to check the appropriateness of the probability distribution for the respective durations. Table 4.1 Calculated Chi-square values

Summary of calculated chi-squared values for the indicated durations (hr)

Stat

ion

Nam

e Probability Distributions

1 2 3 6 12 24

Tabulated

2

( = 0.05)

EVI 10.71

7.01

7.88

8.05 13.24

5.57

14.07

Log normal 11.61

8.78

13.12

13.98 8.29

5.14

14.07

Bed

ele

Log Pearson III

17.05

7.69

5.04

3.76 4.76

4.73

12.59

EVI 5.71

2.65

3.41

8.45 6.89

5.52

14.07

Log normal 2.56

2.26

6.96

5.86 13.31

7.71

14.07

Gor

e Log Pearson III

2.62

1.72

5.77

5.69 5.69

16.11

12.59

The tabulated Chi-square values were 14.067 (for EVI and Lognormal) and 12.592 (for Log Pearson III). Results of the test show that all test statistics are less than the limiting values (at 95% confidence level) in the case of EVI and Lognormal distribution functions at both stations. However, if the Log Pearson III is assumed, the test statistics exceeds the limiting value for 1-hour (at Bedele) and for 24-hour rainfall (at Gore station). Therefore, both EVI and Lognormal distribution

cbdTaI


234

NCSTI-2012functions can be used to explain the rainfall data from both stations.

There is no agreement among hydrologists as to which of the many theoretical distributions available are most suitable to describe natural events and no agreement has been reached on the best techniques of fitting a distribution (Benson, 1968). Spence (1973) compared the fit of normal, 2-parameter lognormal and EVI to annual maximum flows on the Canadian prairies and found that the lognormal was the best fitting distribution. In the United States, Reich (1973) conducted a survey on views of engineers and hydrologists and found that EVI and Log-Pearson type III are preferred well. In Italy, Cicioni et al., (1973) found the 2-parameter lognormal to be the most suitable. Zalina et al., (2007) compared eight candidate distributions in providing accurate and reliable maximum rainfall estimates in Malaysia. On the basis of these comparisons, Generalized Extreme Value (GEV) distribution was found to be the most appropriate distribution for describing the annual maximum rainfall in Malaysia.

Because the simplest descriptions are the best,

lognormal distribution, which has only two parameters and simpler mathematical expression, is proposed to work for Bedele and Gore stations.

Extreme Rainfall Values (XT)

The selected lognormal probability distribution was used to calculate the extreme rainfall values (XT) at different rainfall durations and return periods to form the historical IDF curves for both stations. The computed extreme rainfall values for the required durations at Bedele and Gore stations are presented in Tables 4.2 and 4.3, respectively. Table 4.2 Computed XT at Bedele station

XT values (mm) for the Indicated Duration Di (hrs) Return Period (T) 1 2 3 6 12 242 24.10 26.43 27.41 31.07 31.18 36.55 5 29.63 33.38 34.57 38.02 38.05 47.23 10 32.65 37.18 38.43 41.50 41.69 54.30 25 35.94 41.29 42.55 45.00 45.54 63.24 50 38.08 43.95 45.18 47.11 47.99 69.87 100 40.01 46.32 47.50 48.90 50.14 76.45 Table 4.3 Computed XT at Gore station

XT values (mm) for the Indicated Duration Di (hrs) Return Period (T) 1 2 3 6 12 242 32.81 36.54 38.43 41.35 42.04 45.75 5 41.77 45.32 46.99 51.21 51.61 56.60 10 47.40 50.73 52.21 57.27 57.45 63.26 25 54.23 57.20 58.40 64.52 64.40 71.22 50 59.16 61.81 62.79 69.69 69.34 76.89 100 63.97 66.28 67.01 74.68 74.10 82.37

From the two tables above it can be noticed that the extreme rainfall values increase with an increase in return period for each duration in both stations as expected. The calculated extreme rainfall depth in Gore station is higher than at Bedele station for each selected duration and

return period. These values can also be used to calculate the rainfall intensity for the indicated durations and the required return periods and as input data for estimating the IDF curve parameters.

Extreme Rainfall Intensities (I)

The computed intensity of rainfall at Bedele and Gore stations are presented in Tables 4.4 and 4.5 for each duration (Di) and selected return period (Ti). The values were obtained by dividing extreme rainfall values in section 4.2 by their respective durations. Table 4.4 Computed Rainfall Intensity (mm/hr) at Bedele station

Intensity of Rainfall (mm/hr) for the Indicated Duration, Di (hrs)

Return Period (T) 1 2 3 6 12 242 24.10 13.21 9.14 5.18 2.60 1.52 5 29.63 16.69 11.52 6.34 3.17 1.97 10 32.65 18.59 12.81 6.92 3.47 2.26 25 35.94 20.65 14.18 7.50 3.80 2.63 50 38.08 21.97 15.06 7.85 4.00 2.91 100 40.01 23.16 15.83 8.15 4.18 3.19 Table 4.5 Computed Rainfall Intensity (mm/hr) at Gore station

Intensity of Rainfall (mm/hr) for the Indicated Duration, Di (hrs)

Return Period (T) 1 2 3 6 12 242 32.81 18.27 12.81 6.89 3.50 1.91 5 41.77 22.66 15.66 8.53 4.30 2.36 10 47.40 25.36 17.40 9.54 4.79 2.64 25 54.23 28.60 19.47 10.75 5.37 2.97 50 59.16 30.91 20.93 11.61 5.78 3.20 100 63.97 33.14 22.34 12.45 6.17 3.43

From the above rainfall intensity values, it can be noticed that similar to the extreme rainfall values the rainfall intensity increases with an increase in return period for each duration in both stations. The calculated rainfall intensity in Gore station is higher than at Bedele station for each of the selected duration and return period. These rainfall intensity values can be used for design purpose by matching the rainfall durations in the table with the required return periods. And it also used as input data for estimating the IDF parameters.

IDF curves for the stations

The IDF curves were plotted on a double logarithmic scale, using the duration, Td, as abscissa and the intensity, I, as ordinate, with the help of IDF curve fit software (EXACT, 2006). Figures 4.1 and 4.2 show the IDF curves plotted for the two stations.


235

NCSTI-2012The IDF-curve fit software (MIDUSS version 2.25)

which is a window version of the DOS program was used for this analysis. Table 4.6 presents estimated values for parameters a, b and c for the two stations and the desired frequencies.

Table 4.6 Estimated IDF parameters for different return periods for the two stations

Frequency (years) Station Name

Parameters

2 5 10 25 50 100

a 887.38 1179.88

1369.16

1506.92

1579.46 1643.65

b 0.467 2.246 3.382 2.857 2.202 1.729

Bedele c 0.8743 0.8897 0.8978 0.8984 0.8974 0.8965

a 1755.84

1987.54

2102.14

2338.49

2569.57 2762..70

b 9.111 4.382 1.729 0.472 0.472 0.138 Gore

c 0.9393 0.9275 0.9200 0.9188 0.9219 0.9227

Figure 4.2 IDF curves for Bedele station

For the two stations, the “a” coefficient increase with an increase in return period but the values of the estimated exponent “c” are numerically very similar and it is in the order of 0.85 to 0.90 for Bedele and 0.90 to 0.95 for Gore. The data present a rather long bending aspect represented by high values of the parameter “b” as seen in the IDF curves for Gore station for the 2 years return period (see Figure 4.1); for the remaining return periods the IDF curves has rather straight line aspect. In case of Bedele station, the data represent a short bending aspect represented by high values of the parameter “b” for 10 years return period as seen in the IDF curves for Bedele station (see Figure 4.2)

Rainfall intensities at each station for all durations can

be calculated using the estimated IDF parameters in the general form of equation (3.4a).

Fig. 4.1 IDF curves for Gore station For given duration of rainfall, the rate of increase of (I)

decreases with increase in return period. Similarly for given (I), the rate of increase of duration decreases with increase in return period. The plots will serve to meet the need for rainfall Intensity-Duration-Frequency relationships in the study area for return periods of 2 years to 100 years.

The resulting six equations for each station presented in Appendix E can be used for intensity calculations in the area represented by that station.

Evaluation of the models

Tables 4.7 and 4.8 present the observed and computed rainfall intensities and Tables 4.9 present the coefficients of determination values for different return periods in a given duration.

Development of IDF Models for the stations Table 4.7 Observed and computed rainfall intensities at Bedele station

Observed intensities (mm/hr) for the Indicated Duration, Di (hrs)

Model-predicted Intensities (mm/hr) for the Indicated Duration, Di (hrs)

Return Period ( T ) 1 2 3 6 12 24 1 2 3 6 12 24

2 24.10 13.21 9.14 5.18 2.60 1.55 24.02 13.15 9.23 5.04 2.75 1.50 5 29.63 16.69 11.52 6.34 3.17 1.89 29.89 16.40 11.49 6.24 3.38 1.82 10 32.65 18.59 12.81 6.92 3.47 2.08 33.01 18.15 12.72 6.88 3.71 2.00 25 35.94 20.65 14.18 7.50 3.80 2.29 36.51 20.00 13.99 7.56 4.07 2.19 50 38.08 21.97 15.06 7.85 4.00 2.43 38.79 21.16 14.79 7.98 4.30 2.31 100 40.01 23.16 15.83 8.15 4.18 2.56 40.80 22.20 15.50 8.36 4.50 2.42


236

NCSTI-2012

Table 4.8 Observed and computed rainfall intensities at Gore station Observed intensities (mm/hr) for the Indicated Duration, Di (hrs)

Model-predicted Intensities (mm/hr) for the Indicated Duration ,Di (hrs)

Return Period ( T ) 1 2 3 6 12 24 1 2 3 6 12 24

2 32.81 18.27 12.81 6.89 3.50 1.91 32.86 18.27 12.77 6.81 3.59 1.89 5 41.77 22.66 15.66 8.53 4.30 2.36 41.75 22.66 15.73 8.36 4.42 2.33 10 47.40 25.36 17.40 9.54 4.79 2.64 47.36 25.36 17.54 9.31 4.93 2.61 25 54.23 28.60 19.47 10.75 5.37 2.97 53.95 28.64 19.75 10.46 5.54 2.93 50 59.16 30.91 20.93 11.61 5.78 3.20 58.54 31.01 21.36 11.29 5.96 3.15 100 63.97 33.14 22.34 12.45 6.17 3.43 63.05 33.29 22.91 12.09 6.38 3.36

A comparison between observed intensities and model-predicted intensities for both stations shows that, there is a good match between model outputs and observed intensity data for the considered return periods. If the IDF model performs very well, the possible value of coefficient of determination will be 1 this implies that there is perfect fit between the historical or observed intensities and computed or model-predicted intensities. Coefficient of determination values are presented in Table 4.9. Table 4.9 Value of coefficient of determination (r2)

Stations Return Period (T) Bedele Gore

2 0.9990 1.0000 5 0.9990 1.0000

10 0.9990 0.9990 25 0.9990 0.9990 50 0.9980 0.9990 100 0.9980 0.9990

The values of coefficient of determination corresponding to all return periods at both stations are very close to 1. Therefore, there are good correlations between observed and computed values of rainfall intensity at both stations. Hence, it can be inferred that the IDF curves can be adequately expressed using the proposed models (through the parameters estimated) for the two raingauge stations in Illu Aba Borra zone.

V. Recommendation On the bases of the results of this study, the following

recommendations are given: 1. Water resource professionals, designers and

concerned institution in the study area can utilize the result of this study to derive the IDF value in any part of the study area. In addition, the network of automatic recording stations of this area must be improved in number as well as in type so that the IDF-relationships of the study area can be revised and updated from time to time for further improvement.

2. For those areas closer to Bedele and Gore Stations (within 25 km radius according to WMO guide line), the IDF relationships developed for these stations can be used.

3. Further research should be conducted to develop IDF curves for other stations in the zone to establish

Isopluvial maps of maximum rainfall for the area at various durations and return periods.

4. Since the changing climatic conditions alter the timing and frequency of extreme events (rainfall, floods, droughts, etc.) and therefore have the potential to seriously affect future engineering design standards. Therefore, future studies on IDF should focus on the analysis of IDF under changing climatic conditions.

Reference Ahmad M.I., Sinclair, C.D. and Spurr, B.D. 1988. Assessment of Flood Frequency Models Using Empirical Distribution Function. Journal of Water Resource 24, 1323-1328 Baghirathan V.R., Shaw E.M., 1978. Rainfall Depth-Duration-Frequency Studies for Srilanka. Journal of Hydrology 37, 223-239 Islam, A., and A. Kumar, 2003. HYDRO: A Program for Frequency Analysis of Rainfall Data. Journal of Hydrology, Vol. 84. 5p Nemec, 1973. Engineering Hydrology. McGraw-Hill Publishing Company Limited. New Delhi. Raiford, J. P., N. M. Aziz, A. A. Khan, and D. N. Powell, 2007. Rainfall Depth-Duration-Frequency Relationships for South Carolina, North Carolina, and Georgia. American Journal of Environmental Sciences 3 (2): 78-84 Vogel, R.M. and McMartin, D.E. 1991. Probability plot Goodness-of-fit and Skewness Estimation Procedures for the Pearson Type 3 Distribution. Water Resource Research 27, 3149- 3158 WMO (World Meteorological Organization), 1994. Guide to Hydrological Practices. Data Acquisition and Processing, Analysis, Forecasting, and Other Applications. WMO-No. 168. 5th Edition. Geneva, Switzerland. Zalina M.D., M.N.M . Desa, V.T.T. Nguyen and A.H.M. Kassim, 2002. Selecting Probability Distribution for Extreme Rainfall Series in Malaysia. Water Science and Technology Vol, 45, No 2, PP 63 – 68.

Microbiological Quality of Drinking Water and Hygiene sanitation Knowledge of the People ...

237

NCSTI-2012

Microbiological Quality of Drinking Water and HygieneSanitation Knowledge of the People in Adama Town of Oromia

Region, Ethiopia Dr. Sukanta Bandyopadhyay

Department of Biology, Adama Science & Technology University, Adama.P.B.1888, Ethiopia E mail: [email protected]

Abstract — The World Health Organization (WHO) estimated that up to 80% of all sicknesses and diseases in the world are caused by inadequate sanitation, polluted water or unavailability of safe drinking water. Over 75% of the health problems in Ethiopia are due to communicable diseases attributed to unsafe and inadequate water supply and unhygienic waste management. Recent evidence suggests that many improved drinking water supplies suffer from poor reliability. So, the aim of this study was to determine the microbiological quality of drinking water and investigate the hygiene-sanitation knowledge & practices of the people in Adama town of Oromia region of Ethiopia. A cross-sectional prospective study was conducted in Adama town during November – December, 2011. Water samples were collected from taps of direct pipe line water from different kebeles of the town and from the household water containers of the people of this town for microbiological analysis. Analysis of water showed that 20% of water samples from the taps of direct pipe line and 70% of the water samples from the household samples were contaminated with Total Coliforms and the hygiene sanitation knowledge and practices of most of the consumers in this town were poor. Regular microbiological quality assessment of the pipe line drinking water and initiative to give hygiene education to the people of this town particularly in slum area is necessary.

Keywords- Microbiological quality, drinking water, hygiene sanitation, adama town

I. INTRODUCTION

The World Health Organization(WHO) estimated that up to 80% of all sickness and diseases in the world are caused by inadequate sanitation , polluted water or unavailability of safe drinking water [1]. In Ethiopia, over 60% of the communicable diseases are due to poor environmental health conditions arising from unsafe and inadequate water supply and poor hygienic and sanitation practices [2]. Three-fourth of the health problems of children in this country are communicable diseases originating from the environment, specially drinking water and sanitation. 46% of less than five years child mortality is due to diarrhea in which water related diseases contribute a high proportion. In a report, the Ministry of Health, Ethiopia reported that each day about 6000 children die from diarrhea and dehydration [3]. Moreover, high fluoride concentration in the ground water has long been a recognized water related health concern in Ethiopia, particularly in this Rift Zone [4]. A number of studies have been conducted in various regions of this country in recent past. In 2004, by a cross sectional study on drinking water quality in North Gondar region, it was found that

majority of the drinking water sources were either of unacceptable quality or grossly polluted[5]. In the last year 2011, Milkiyas Tabor et al have reported that drinking water supply in the Bahir Dar City was found to be contaminated with bacteria [6]. The provision of water, sanitation and good hygiene services play a key role for the protection and development of human resources [7]. Globally about 1.1 billion people depend on unsafe drinking water sources from lakes, rivers and open wells. The majority of these are in Asia(20%) and sub-Saharan Africa(42%).Still now 2.4 billion people lack adequate sanitation worldwide [8]. Again, most of the population of Ethiopia does not have access to safe and reliable sanitation facilities [6]. Besides these, majority of the households do not have sufficient understanding of hygienic practices regarding food, water and personal hygiene. There are a number of pollution sources that continuously deteriorate the microbiological quality of surface and ground water in Adama town. In a recent study, from drinking water samples collected throughout the Ethiopian part of the Rift Valley analyzed for 70parameters, 86% of all wells yield water that fails to pass the quality standards set for drinking water in terms of European water directives and WHO guidelines [9]. Among the


238

NCSTI-2012

major rivers of Ethiopia, Awash is the most threatened river from industrialization and increasing of urbanization. This is because most of the existing industries and major towns within the upper watershed have no treatment plants for discharge of their wastes [10]. But till now no data is available about the microbiological quality of drinking water of Adama town which draws substantial drinking water from the river. So the aim of this study was to determine the microbiological quality of drinking water and investigate the hygiene sanitation knowledge and practices of the people in Adama town of Oromia region of Ethiopia.

II. MATERIALS AND METHODS

A cross sectional prospective study was conducted in Adama town of Oromia region during November- December, 2011. Thirty water samples from pipeline taps and thirty water samples from household water containers were randomly collected from all kebeles of Adama town. The selection of sampling points and frequency of sampling was determined according the guidelines of WHO [11]. Microbiological quality of drinking water from pipeline taps and household water containers were analyzed. The hygiene and sanitation practices of the resident consumers of this town were also assessed. For microbiological analysis, 500 ml of water samples were collected in sterile PVC bottle between 8.00 to 10.00 am and transported to the laboratory. The number of Total coliform counts and Faecal(thermotolerant) coliform counts were determined with the membrane filtration methods and Lauryl Sulfate-Broth medium using Potalab Wagtech (U.K) water quality analysis kit [12]. For the determination of Total coliform counts and Faecal (thermotolerant) coliform counts, incubation was carried out at 370C and 440C, respectively. Consumer’s hygiene sanitation practices were assessed through interview. The interview questions and sanitary inspection forms were adapted from WHO and assessment of the conditions of household water containers was obtained through observation checklist [11]. Finally, data were recorded, organized and summarized in the form of descriptive statistics using SAS, JMP 501 and SPSS version12. Ethical clearance was obtained from the Ethical clearance committee of Adama Science and Technology University. Data at the households were collected after informed consent was assured from the households. The study objectives were clearly explained to the households and each household was assured that the information provided would be kept confidential.

III. RESULTS AND DISCUSSION

In this study, 30 water samples were collected from each types of water sources--- i) direct pipeline taps and ii) household water containers. Pipeline water supply provides chlorinated treated water for drinking purpose. Analysis of samples collected from direct pipeline tap water demonstrated that 20% and 13.33% of the samples were contaminated with Total coliforms and Faecal(Thermotolerant) coliforms respectively. However 70% and 50% water samples collected from household were contaminated with Total coliforms and Faecal (Thermotolerant) coliforms respectively. On the other hand, 80% tap water samples and 30% household water samples were found having no Total coliforms respectively(Table I). Distribution (CFU/100ml) of the Total coliforms and Faecal(Thermotolerant) coliforms present in the tap water and household water samples is presented in the Table II. The results of sanitation and hygiene practices of the household consumers of Adama town are expressed in Table III.

A variety of intestinal pathogens are introduced by faecal pollution of drinking water [1], [3]. The presence of Total coliform organism indicates a need for further survey, investigation and examination of drinking water sources. In regularly checked chlorinated water supplied by pipeline should have total coliform count zero per 100ml [12]. But this fact is only observed in 80% tap water and 30% household water samples.

In the present study, the Total coliform and Faecal (Thermotolerant) coliform counts were higher in household water samples compared to that of direct pipeline tap water. This is in agreement to a study conducted in SriLanka that showed that microbiological quality of household drinking water samples was poor than water from the source [14].Similar result was found in other study conducted in Ethiopia [15]. In a recent study Milkiyas Tabor et al have found similar findings that Total coliform counts in household containers was higher compared to tap water in Bahir Dar City, Ethiopia [6]. Studies conducted in South Africa and Zimbabwe also reported that compliance is significantly higher for tap water (85.4%) than from household water containers (43.6%) [16].

General system failures, inefficiency in disinfection, poor maintenance are some of factors that affect the quality of drinking water in Ethiopia [15], [6].

The presence of Total coliforms and Faecal (Thermotolerant) coliforms in direct pipeline tap water samples could be due to cross- contamination between the pipeline water supply and sewer, due to old leaky pipes and lack of proper water pressure [17],[6]. Warm conditions and unavailability of residual free chlorine in pipeline water supply can favor the re-growth of organisms like


239

NCSTI-2012

Table I Microbiological Analysis Of Drinking Water Collected From Taps Of Direct Pipeline And Household In Adama Town Of Oromia Region, Ethiopia, Nov-Dec, 2011 (N=30)

Tap water Household water Type of organisms No % No %

Total coliforms (CFU/100ml) 6 20 21 70Faecal (Thermotolerant) 4 13.33 15 50Absence of total coliforms 24 80 9 30

TABLE II Distribution Of Organisms Present In The Samples Of Drinking Water Collected From Taps Of Direct Pipeline And Household In Adama Town Of Oromia Region, Ethiopia, Nov-Dec,2011 (N-30)

Tap water Household water Distribution of organisms No % No %

Total coliforms (CFU/100ml) 0 24 80 9 301 - 10 4 13.33 12 4011 - 20 2 6.66 9 30>20 - - - -Faecal (Thermotolerant) coliforms (CFU/100ml) 0 26 86.66 15 501 - 10 3 10 6 2011 - 20 1 3.33 4 13.33>20 - - 5 16.66

Table III Response Of Household Consumers Regarding Sanitations And Hygiene Practices In Adama Town, Ethiopia, Nov-Dec,2011(N=30)

ResponsesQuestionsaskedtotheconsumres Yes o(%) (%) N

1. 9(30%) 21(70%)Doyoucarrycovered/tightcapedcontainerstocollectwaterfromthetap?

2. Wasthereanycontactofyourhandstowater

duringcollection? 12(40%) 18(60%)3. Doyoucleanyourwatercollectioncontainers

withsoapeverydayortwice/week? 3(10%) 27(90%)iners4. Doyoucleanyourwatercollectionconta

5.everyday? 19(63.33%) 11(36.66%)Doyouwashyourhandswithsoapaftervisitingtoilet? 7(23.33%) 23(76.66%)

thermotolerant Faecal coliforms in the distribution systems [18], [11]. High counts of Total coliforms and thermotolerant Faecal coliforms at the household drinking water samples indicate that the water has been faecally contaminated. Poor sanitation & hygiene knowledge of the people particularly in the slum area of this town were the main factors for the contamination of water during transportation and after storage at home. In our study, it is found that 76.66% consumers do not wash hands with soap after visiting toilet. Unhygienic hose pipes to collect water by the people in slum area could also be a source of bacteriological contamination in the household

water samples. This finding is in agreement to the studies conducted in other parts of Ethiopia [15], [5], [6].

IV. CONCLUSION

Microbiological analysis of drinking water samples showed that in Adama town of Oromia region, 20% of samples collected from taps of direct pipeline and 70% of household water samples were contaminated with Total coliform bacteria and the hygiene sanitation knowledge and practices of the most of the consumers in this town were poor. Systematic and comprehensive water quality assessment is recommended for prevention of contamination in drinking water


240

NCSTI-2012supply. Initiations should be taken to upgrade the water, sanitation and hygiene knowledge of the common people of this town.

REFERENCES

[1] WHO, Basic Environmental Health, Geneva, 1997.[2] Abebe Land, “Hygienic water quality, its relation

to health and the testing aspects in tropical conditions”, Dept. of civil Engineering, Tempe ere University, Finland, 2006.

[3] Ministry of Health, Knowledge, attitude and practice of water supply, Environmental Sanitation and hygiene practice in selected zones of Ethiopia, 1997.

[4 ]Ground water quality: Ethiopia- British Geological Survey, 2001.

[5] Mengesha A, Mamo W, Baye G, “A survey of bacteriological quality of drinking water, North Gondar”, Eth J Health Dev, vol18, no 2, pp. 113-135,2004.

[6] Milkiyas Tabor, Mulugeta K, Bayeh A, “Bacteriological & Physicochemical quality of drinking water and hygiene-Sanitation practices of the consumers in Bahir Dar city, Ethiopia”, Ethiop J Health Sci, vol 22, no 1, pp. 19-26,2011.

[7] Fewtrell L, Colford J, “Water sanitation and Hygiene: Interventions and diarrhea a systematic review and meta- analysis”. The International Bank for Reconstruction and Development/ World Bank, 2004, www.worldbank.org.

[8] WHO. “Global Water supply and sanitation assessment report”,2000.

[9] Clemens Reimann, Kjell B, Bjorn F, Zenebe M, “Drinking water quality in the Ethiopian section of east African rift valley I- data and health aspects”, The Science of the Total Environment vol311,pp. 65-80,2003.

[10] Ministry of Water and Energy, “Existing water quality situation in Ethiopia”, 2010.

[11] WHO guidelines for drinking water- water quality,2nd

ed,vol 3, Surveillance and control of community supplies, Geneva,1997.

[12] POTALAB WAGTECH ( WAG- WE 10010), Instructions for water quality and environmental testing(www.wagtech.co.uk)

[13] Cheesbrough M, Medical laboratory manual for tropical countries, Butter- worth, Heine man, Oxford,II pp. 212-220,1987.

[14] Dissanayake SAMS, Dias SV, Perera MDC, Iddamalgoda IAVP, “Microbial quality assurance of drinking water supplies through surveillance”, Environment Division. National Building Research Organization, Colombo, SriLanka, Water Professional’s symposium, October 2004.

[15] Dagnew T, Assefa D, Woldemariam G, Solomon F, Schmoll O, “Drinking water quality in the federal republic of Ethiopia”. Ministry of Health,Environmental health department. Country report, Addis Ababa,pp. 19- 67,2007.

[16] Momba MNB, Yyafa Z, Makala N, Brouckaert BM, Obi CL, “Safe drinking water stills a dream in rural areas of South Africa .Case Study : The Eastern Cape Province”, J Water Sci Res Tech, vol32 ,no 5,pp. 1816-7950,2006.

[17] Semenza JC, Roberts L, Henderson A, Bogan J, Rubin CH, “Water distribution system and diarrheal disease transmission: a case study in Uzbekistan”, Am J Trop Med Hyg ,vol59,no 6, pp.941-946, 1998.

[18]Muyima N, Negcakani F, “Indicator bacteria and re-growth potential of the drinking water in Alice, Eastern Cape”.Dep J Biochem Microbiol, vol 24, no 1, pp. 29-34, 1998.

Monitoring stations for runoff and sediment transport along Gumara River (Tana Basin)...

241

NCSTI-2012

Monitoring Stations for Runoff and Sediment Transport along Gumara River (Tana Basin): Procedures and Preliminary Results

Mekete Dessie1, 2,*, Teshager Admasu3, 5, Valentijn Pauwels2, Niko Verhoest2, Enyew Adgo3, Jean Poesen4, Ruben Maes4, Jan Nyssen5

1School of Civil & Water Resources Engineering, Bahir Dar University, P.O.Box 430, Ethiopia,

[email protected] 2Laboratory of Hydrology and Water Management, Ghent University, Coupure links 653, B-9000 Gent, Belgium,

3College of Agriculture & Environmental Sciences, Bahir Dar University, Ethiopia, TeshagerAdmasu 4 Department of Earth and Environmental Sciences, KU.Leuven, Belgium,

5Department of Geography, Ghent University, Krijgslaan 281 (S8), B-9000 Gent, Belgium

Abstract - Gumara River, one of the major tributaries of Lake Tana, drains an area of 1768km2.It is within the growth corridor of Lake Tana Basin, where an increasing number of water resource development projects (mainly hydropower and irrigation) take place. Monitoring stations for runoff and sediment have been installed under Wase-Tana Project along Gumara River using automatic water level recorders to produce data of finer resolution (10 minutes duration) and ordinary staff gauges (maximum of three readings a day).Depth integrated suspended sediment samples have also been collected. This study presents an assessment of variation of different time scale discharge measurements and adjustments (refinements) for the daily discharge data acquired through manual records of the staff gauges. Preliminary analysis of the discharge and sediment data collected was made. The hydrograph shows that the mean daily discharge values from the automatic recorders are greater than the daily values of the staff gauges for almost all of the times of record (nearly 50% increment). A good correlation has been found between the two methods (the correlation coefficient, R2 =0.79) and the difference is much appreciated during the rainy seasons. A wide range of suspended sediment concentrations (6-6654mg/L) were obtained during the time of record. Keywords: Monitoring station, basin, hydrograph, runoff, sediment

1 INTRODUCTION Lake Tana, the source of Blue Nile, is the largest

lake in Ethiopia. It contributes about two thirds of the total flow of water to the Nile [1]. Four large rivers, Gilgel Abay, Gumara, Rib and Megech account for the major water inflow to the lake [2]. Extensive catchment degradation has reduced the productive capacity of land in the basin [3] and intensified erosion rates, which reduce the storage capacity of the lake due to sedimentation. Moreover, an increasing number of water resource development projects (mainly irrigation) take place in the area. Understanding discharge and sediment transport is, therefore, essential for effective planning of water resource development projects in the catchment and sustained life of the lake. This is possible when reliable and adequate hydro-meteorological data are available. There are some 33 meteorological and 24 hydrometric stations in and around the basin owned

by Hydrological Department and National Meteorological Agency of the MoWR. Currently, almost all the hydrometric stations use the staff gauge to record the water levels twice a day manually. A fair number of the stations (about 40%) have the housing for the automatic water level recorders, but these recorders were all removed around the late 1980’s, except from the key station on the Abbay at Bahir Dar [1]. The twice daily readings are clearly insufficient to estimate daily discharges as there are high chances for the peak floods to pass unattended, especially in smaller rivers where rising and falling limbs of the hydrographs are often very swift. Installation of water level recorders is essential for streams where the level is subject to abrupt fluctuations.


242

NCSTI-2012

The area for which discharge data is available in the Tana basin is about 42% of the catchment area draining towards the lake [1]. For some rivers, the number of hydrometric stations is much less than the minimum densities of stream flow stations required. Moreover, a large number of the stations have been victims of siltation, bank overflow and unstable cross-sections [1]. Studies made by SMEC [1] indicate that the rating curves of the Gumara and Ribb Rivers at the location of the stations are likely to be affected by the water level in Lake Tana during the wet season.

The lack of reliable and adequate hydro-metrological data has been discussed in many hydrological studies made in the basin, and the different techniques employed to quantify lake inflow from ungaged catchments resulted in very differing outputs: 7% by Kebede[2], 29% by SMEC [1] or even more than 42% by Abeyou [4].

To address these problems, a research project under the title of “Water and sediment budgets of

Lake Tana for optimisation of land management and water allocation” (shortened as Wase-Tana Project, “WASE” to mean Water-Sediment) is underway. The Wase-Tana project (http://geoweb.ugent.be/physical-geography/research/wase-tana), initiated by Bahir dar and Belgian Universities, has already installed 9 discharge and sediment monitoring stations equipped with automatic water level recorders, and at least 5 more stations are under installation in the whole Tana basin. Three measurement stations on Gumara River are operational since July, 2011. This paper addresses the procedures and preliminary results of two of the monitoring stations on Gumara River and also depicts an assessment of variation of different time scale discharge measurements and adjustments (refinements) for the daily discharge data acquired through manual records of the staff gauges.

2. THE STUDY CATCHMENT Gumara River and its catchment is located on

south-east of Lake Tana (Fig.1).The three discharge and sediment monitoring stations which are operational since July, 2011are indicated on the same Fig. The Wanzaye Station is located where the flood plain almost starts and drains an area of 1236 km2. The Lower Gumara Station has a catchment area of 1340 km2 and it is in the flood plain. The Kizin station, a tributary to Gumara River, has a drainage area of 8.4km2 and the stream is seasonal. Based on the Abay Master Plan [5], more than 80% of the catchment consists of

luvisols. The 30m DEM obtained from SRTM shows that the peak elevation is about 3710m a.s.l (Fig.2) and the dominant catchment slopes are 8-15% and 15-30% accounting for 25% and 32% of the total catchment area respectively. The annual isohyets prepared by SMEC [1] show that the mean annual rainfall is in the range of 1200-1800mm in the catchment and it is concentrated from June to September. Flooding is a recurrent problem in the downstream reaches of Gumara .The dominant land use is agriculture and agro-pastoral.

Fig. 1 Monitoring Stations and Gumara catchment


243

NCSTI-2012

Fig.2 DEM of Gumara catchment at the outlet point of Wanzaye Gauging Station

3. MATERIAL AND METHODS Discharge at a given time can be measured by

several different methods, and the choice of methods depends on the conditions encountered at a particular site [6]. The determination of runoff in Gumara River has been done by collecting data on water levels and velocity of flow from three runoff and sediment monitoring stations, which have been installed in the basin (Fig.1).Several sediment samples were also taken and analyzed to understand sediment transport of the river. Water level measurements have been made using automatic water level recorders (every 10 minutes) and manual readings from a staff gauge (three times a day, at about 7am,1pm and 6pm) since July 20,2011. The automatic water level recorder (or Mini-Diver) has a pressure sensor to determine water level and a temperature sensor. It can be set to any desired time interval and data is retrieved by inserting the Mini-Diver in a USB reading unit connected to a laptop or PC with Diver-Office. For each measurement, the date, time, water level and temperature are stored The installation of the monitoring stations involves construction of pillars on left and right banks, erecting staff gauges and placing the diver on the bed of the river. The diver can be placed using different techniques depending on the site condition, for example at Wanzaye Station it has been placed inside a pipe bottom, which is erected at one side of the river bank with holes on the surface to allow entry of water to the pipe (Fig 3). The same pipe has also been used as a staff gauge for the manual readings. The left and right pillars are used to stretch a suspension cable across the

river bank to establish arrangements along which the sediment sampler moves to the required section of the river to take samples during flood. The sediment sampler is local made metal frame designed to hold a plastic bottle tight and with sufficient weight to allow good immersion deep into the water enabling to take depth integrated suspended sediment samples (Fig.3). Sediment samples were collected at least 2 times a day during the rainy season and at any time when there is flood. The samples were brought to the laboratory and analyzed for sediment concentrations using the filter and evaporation methods. The filtration was carried out using Whatman filter papers with a pore opening of 2.5μm and the filtered sediments were oven-dried. The velocity measurement was done using the float method, owing to the difficult site condition to use the current meter during flood. The velocity of the float is equal to the distance between cross-sections divided by the time of travel and this velocity was adjusted using a coefficient of 0.86. According to reference [6], when the ratio of the immersed depth of float to depth of water is 0.10 or less, the coefficient for velocity adjustment is 0.86. The result was used to evaluate the value of manning’s roughness coefficient (n) obtained from suggested values for n, tabulated according to factors that affect roughness found in Chow[7], Henderson[8] and Streeter [9]. Measurement of river cross-section and longitudinal slope was made using surveying equipment (total station). Fig.4 depicts the cross-sections for Wanzaye station.


244

NCSTI-2012

Fig.3 Details of a typical monitoring station installation

Fig.4 River cross-section at Wanzaye Station

4. PRELIMINARY RESULTS AND DISCUSSION Rating curve

Since continuous measurement of discharge is not usually feasible, records of discharge are computed from the relationship between stages and discharge (rating curve). The rating curve (Fig.5) was produced after the survey of the cross-section of the river and measurement of flow velocity using the float method at different flow stages for the Wanzaye monitoring station. If Q and h are discharge and water level respectively, then the relationship can be analytically expressed as

Q=f (h) Where; f (h) is an algebraic function of water level. Using the power regression type in EXCEL, the rating curve equation for Wanzaye gauging station is found to be:

Q=36.556h2.3165 (Q in m3/s and h in m) The rating relationship thus established is used to transform the observed stages into the corresponding discharges.


245

NCSTI-2012

Fig.5 rating curve of Gumara River at Wanzaye Station

Hydrograph The transformed discharges from the measured

water levels are plotted against time and the resulting hydrographs for the two stations are given on Fig.6 and Fig.7. The figures show that the daily discharges obtained from the automatic recorders are greater than the corresponding values of the staff gauge readings for the entire period of record in all the stations. The difference is much appreciated during high flows and in smaller catchments (Fig.7), because in smaller catchments

the flood takes short time to pass without being observed by the person making records. However, staff gauges (three times reading a day) can give reliable estimates of discharge during dry seasons. The run off volume for Gumara River, according to the automatic water level recorder from July 20 to October 19, 2011, was 1.25 billion m3 at the Wanzaye Station

Fig.6 Hydrograph of Gumara River at Wanzaye Station


246

NCSTI-2012

Fig.7 Hydrograph of Kizin Station The maximum daily discharge observed for this river at Wanzaye Station was 375m3/s corresponding to a water level of 2.7m above river bed, and frequent rises and falls in the hydrographs were observed during the rainy season (July-September) in response to the rainfall conditions in both catchments. A further insight into the water level measured at 10 minutes time interval for Wanzaye station shows that the level has risen as high as 3.5m (equivalent to 650m3/s) on August 4, 2011.This typical flood hydrograph is shown on Fig.8.This fits well in relation to the heavy rainfall amount recorded at Wanzaye Station on this day (50.6mm, third

highest amount in the year 2011) in the catchment. The flood hydrograph shows a steeper gradient of the rising limb of the hydrograph as a result of rapid draining of surface run off, which is also a reflection of the slope of the catchment (mostly 15-30%). This is also a good indication of the occurrence of floods. After the peak, the recession limb is almost constant for a considerable time and falls at a gentler gradient. This is likely that most overland flow has now been discharged and it is mainly interflow and surface runoff due to rainfall events more upstream in the catchment which is making up the river water.

.

Fig.8 Maximum flood hydrograph


247

NCSTI-2012

The corresponding flood hydrograph for the smaller catchment of Kizin Stream (8.4km2) is depicted on Fig.9. The recession limb is

relatively steeper owing to the smaller area, and falls rapidly after the peak.

Fig.9 Typical flood hydrograph for Kizin Stream at the gauging station

Staff gauge and diver discharge relations To study the possibility of improving the

discharges obtained using staff gauges for stations using only this option, the daily discharges obtained from the two methods for both stations were plotted (Fig.10 and Fig.11) and linear correlations were established. A significant correlation was achieved with coefficient of

correlation, R2, of 0.79 for Wanzaye Station, whereas for Kizin Station the coefficient of correlation, R2 dropped to 0.5. Fig.10 and Fig.11 generally show that daily discharges found from staff gauges of three times reading a day are underestimated (about 30% for the bigger catchment and 50% for the smaller one).

Fig.10 Correlation of staff gauge and diver readings at Wanzaye Station


248

NCSTI-2012

Fig.11 Correlation of staff gauge and diver readings at Kizin Station

Sediment concentrations

Fig.12 illustrates a plot of suspended sediment concentrations against discharge. A wide range of suspended sediment concentrations (maximum 6654 mg/l and minimum 6.3mg/l) were noticed during the time of record. A preliminary correlation between SSC and discharge was made, but with many scatters and poor correlation. Similar results were found

for a study in Geba catchment [10], when all SSC samples of a gauging station are considered. The correlation can be improved by stratifying SSC samples of all measuring campaigns into different stages of the rainy season like beginning, middle and end of rainy season [10].

Fig.12 Relationship between discharge (Qw) and suspended sediment concentrations (SSC) for 49 observations at the gauging

station of Wanzaye


249

NCSTI-2012

Investigations were also made to understand sediment transport of a particular flash flood event. A flash flood hydrograph in terms of water level of the Gumara River at Wanzaye

station during August 17-18, 2011 is shown on Fig.13.17 depth integrated samples (indicated by numbers 1-17 on the same figure) were taken in this flood event.

Fig.13 Water levels of a typical flood event and sediment sampling (Wanzaye station) Results of the analysis of the suspended sediment concentrations (SSC) of this flood hydrograph are indicated on Fig.14. The sample numbers shown on Fig.13 are also indicated here for easy comparison of the hydrograph and the sediment concentration variations. It can be remarked that at a certain water level, suspended sediment concentrations during the rising limb are higher than the falling limb of

the flood hydrograph (hysteresis effect). The type of hysteresis may assist in determining the sediment source area in a basin [11].The clockwise hysteresis, observed here, results from sediment depletion during floods. In addition the increased portion of base flow during the recession limb may cause clockwise hysteresis loops [11].

Fig.14 variation in SSC and water level of Gumara River at Wanzaye Station during a typical flood event


250

NCSTI-2012 5. CONCLUSION The discharges obtained using staff gauges are smaller than the corresponding values from divers during the rainy season. However, staff

gauges can give reasonably good results during the dry season. Flash flood events are associated with high sediment export rates.

6. REFERENCES [1] SMEC International Pty Ltd, 2009.

“Hydrological study of the Tana-Beles sub-basins”, hydrological monitoring network, p.1-44

[2] Kebede, S., Travi, Y., Alemayehu, T., and Marc V. 2006. “Water balance of Lake Tana and its sensitivity to fluctuations in rainfall, Blue Nile Basin, Ethiopia” Journal of Hydrology, 316, 233-247

[3] Nyssen, J., Poesen, J., Moeyersons, J., Deckers, J., Mitiku Haile, Lang, A., 2004. “ Human impact on the environment in the Ethiopian and Eritrean Highlands” – a state of the art. Earth Science Reviews, 64/3-4: 273-320

[4] A.Wale, T. H. M. Rientjes, A. S. M. Gieske, and H. A. Getachew. “Ungauged catchment contributions to Lake Tana’s water balance”. Hydrol. Process. 23, 3682–3693 (2009)

[5] BCEOM (1999c). “Abay River Integrated Development Master Plan Project”. Phase 2, Section II, Volume IX: Semi-detailed soils survey. Addis Ababa: Ministry of Water Resources.

[6] World Meteorological Organization, 1994” Guide to Hydrological Practices; Data acquisition and processing, analysis, forecasting and other applications” WMO–No. 168.

[7] Chow.V.T., 1959 “Open Channel Hydraulics” New york, McGraw-Hill Book Co., 680p.

[8] Henderson, F.M., 1966. ” Open Channel Flow”, New York, MacMillan Publishing Co., Inc., 522p.

[9] Streeter, V.L., 1971 “Fluid mechanics” New York, McGraw-Hill Book Co., 5th ed.705p.

[10] Vanmaercke , M,. Zenebe, A., Poesen , J., Nyssen, J., Verstraeten, G., Deckers, J. (2010) “Sediment dynamics and the role of flash floods in sediment export from medium-sized catchments: a case study from the semi-arid tropical highlands in northern Ethiopia”, Journal of soils and sediments, 10: p 611–627

[11] Klein, M., “Anti clockwise hysteresis in suspended sediment concentration during individual storms”: Holbeck catchment; Yorkshire, England, CATENA, Volume 11, Issues 2–3, August 1984, p. 251-257

Numerical groundwater flow modeling of the Meki River catchment, central Ethiopia

251

NCSTI-2012

Numerical Groundwater Flow Modeling Of the Meki River Catchment, Central Ethiopia

Dereje Birhanu1, Tenalem Ayenew2

1 Institute of Technology, Hawassa University, Hawassa, P.O.Box 790, Ethiopia, email: [email protected] 2 Addis Ababa University, Department of Earth Sciences, A.A, P.O.Box 1176, Ethiopia, email: [email protected]

Abstract — A three dimensional steady-state finite difference groundwater flow model is used to quantify the groundwater fluxes and analyze the subsurface hydrodynamics in the Meki river catchment by giving emphasis to the well field that supplies water to the community. The area is characterized by Quaternary volcanic covered with lacustrine, alluvial, talus, and pyroclastic deposits. The model is calibrated using head observations from 95 wells. The simulation is made in a one layer unconfined aquifer with spatially variable recharge and hydraulic conductivities under well-defined boundary conditions. The calibrated model is used to forecast groundwater flow pattern, the interaction of groundwater and surface water, and evaluate the behavior of the groundwater system under possible future utilization scenarios. A sensitivity analysis conducted indicates that the model is more sensitive to decrease in recharge and increase in hydraulic conductivity but less sensitive to increment or decrement of pumpage. The simulation result indicates that the groundwater flows from western escarpment to east directions finally join Lake Ziway. Lakes and rivers play important role in recharging the aquifer. Simulations made under different possible future utilization scenarios including increase in pumping rate results in substantial regional groundwater level decline, which will lead to the drying of springs, and shallow hand dug wells. It has also implications of reversal of flow from contaminated rivers in to productive aquifers close to main river courses; decrease in recharge caused more inflow from Lakes as well as increase stream flow but decrease drains, and disappearance of Lake Tuffa results in increased recharge and groundwater outflow through springs. The sensitivity and scenario analysis provided important information on the data gaps and the specific sites to be selected for monitoring that may be of great help for transient model development. This study has laid the foundation for developing detailed predictive groundwater model, which can be readily used for groundwater management practices.

Keywords- Central Ethiopian, Meki, Modeling, Modflow, Volcanic aquifer, groundwater management

I. INTRODUCTION

Numerical groundwater flow models are important tools in hydrogeological studies in different parts of the world. Groundwater models have played an increasingly important role in the evaluation of alternative approaches to groundwater development and management. In recent times groundwater simulation models have received attention in Ethiopia for hydrogeological system analysis [1]. However, due to limitations in pertinent data it has been difficult to develop robust three-dimensional transient models that can be used as groundwater management tool.

Meki river catchment has an abundance surface and groundwater resources; however, due to climatic change, industrialization, high population growth, the amount of water available is decreasing and its quality is degrading. In a country like Ethiopia, where rain fed agriculture is the main source of economy and ensures the wellbeing of many people, water resources are essential. Nevertheless, if the water resources are not utilized properly in an integrated planning manner, its sustainability and its support to food security to the country will become endanger. Therefore proper planning of water resources development as well as utilization is very essential.

Groundwater plays significant role in the region and presently used for almost all town and village

water supply of the sub basin. The integral approach considers both surface and sub-surface water as the major resource of the region. It is observed that there are a lot of boreholes, shallow wells, dug wells, wind pumps and springs over the region which serve as community water supply.

Several conventional investigations have been carried out in the area over the last two decades. The most important groundwater resource assessment has been carried out by [2]. General hydrology and hydrogeology of Ziway–Shalla basin and numerical groundwater flow modeling of the central main Ethiopian rift lake basin have also been addressed [3]-[4].These studies provided initial conceptual view of the quaternary volcanic aquifer systems, hydraulic parameters and the movement and occurrence of groundwater.

Groundwater is pumped by different industries, institutions. Often groundwater abstractions are carried out without the basic understanding of the groundwater recharge, lateral and vertical extent of aquifers and the available groundwater reserve. One important issue to be addressed is the mechanism of groundwater flow, occurrence and assessing the response of the system to different abstraction and recharge rates. In this regard groundwater flow models play important role. This work tries to describe the movement and occurrence of groundwater using steady-state groundwater model. Model simulation is made under different stress scenarios (variable pumping rates) with distributed


252

NCSTI-2012model input parameters such as recharge and hydraulic conductivity.

II. SITE DESCRIPTION The Meki river catchment is located in the central

main Ethiopian rift valley. The total surface area of the catchment is 2318.58 km2. It is bounded between 7051’ to 8027’ N latitude and 38015’ to 38051’ E longitude (Fig. 1). The high plateaux of the Gurage highlands such as Zebidar Mountain of the area range (3611 m.a.s.l) forming the main recharge area. The lowest elevation is to the rift floor around Lake Ziway close to the Meki River (1636 m.a.s.l).

Fig.1. Location map of Meki river catchment with A-B schematic section

The western escarpment or highlands of Gurage Mountain comprises highly welded ignimbrites, tuff, rhyolite and trachyte without visible large faults. The upper weathered rock and soils are permeable; however, the underlying volcanic

sequences are massive. Butajira-Pediment, Kuntane-Inseno-Kela-Plain, Tora-Koshe-Dugda-Ridge and North Eastern area is characterized by ignimbrite, fracturing and weathering grade. The study area consists of recent basalts and highly fractured ignimbrites; and scoria cones along the major fault east of Butajira. Ziway-Plain covers large area around the Lake Ziway. The lithological groups found in this area are ignimbrites overlain by lacustrine sediments such as: clay, diatomite, shale beds and reworked pumice. Major perennial rivers start from the western escarpment and drain to the east. The main ones are the weja, Irinzaf and Meki rivers draining to the Ziway Lake.

The Climate consists of three ecological zones: humid to dry humid, dry sub-humid or semi-arid and semiarid or arid lands [5]. Accordingly, highland areas west of Butajira are categorized under humid to dry sub-humid land. The areas east of Butajira around Lake Abaya are dry sub-humid lands. The rest of the area which is around the lake is in semiarid or arid zone. Rainfall and temperature in the area show strong altitudinal variations. The average annual rainfall varies spatially and ranges from around 715 mm/year in the rift floor to more than 1100 mm/year at extreme highland areas. The average annual prevailing mean temperature ranges from about 110C in the highlands to around 26 0C in the rift.

III. METHODOLOGY The study focuses more on groundwater flow

system analysis, rather than developing a well calibrated model that can readily be used as a management tool. Because the available data does not allow to establish a robust three-dimensional transient flow model. However, the model developed in this study lays the foundation for attaining versatile transient three-dimensional flow models that can play important role for sustainable groundwater management including assessment of contaminant transport.

The catchment was modelled using the widely used United States Geological Survey groundwater flow model called Modflow, under steady-state conditions. Modflow is a modular three-dimensional finite difference groundwater flow code [6], which simulates saturated porous media under steady-state and transient conditions. Groundwater flow models have been used to solve practical problems in a wide variety of hydrogeological environments in different parts of the world [7]-[8]-[9]. The steady-state groundwater flow is simulated based on the following governing differential equation under two-dimensional aerial view [10].

0)()(

R

yhT

yxhT

x yx (1)

Where, h is hydraulic head and R is a sink/source term. Tx and Ty represent the principal components


253

NCSTI-2012of the transmissivity tensor in the x and y directions respectively.

Well logs and pumping test data were collected mainly from previous studies [11] to develop the conceptual hydrogeological model. The pumping tests were made by different organizations using different methods to estimate the aquifer parameters. Both constant discharge and recovery tests were carried out, with very few step-drawdown tests. The most important methods applied include [12]-[13]. In few cases [14] method has been applied. The duration for the constant discharge tests range from 24 to 72 hours, and for recovery tests it ranges from 2 to 12 hours. The estimated transmissivity and hydraulic conductivity values obtained are quite similar to [3].

The groundwater recharge was taken from previous studies [11] estimated using a soil-water balance model and chloride mass balance approach.The recharge estimation processes accounted all pertinent hydrometerological data, soil types, slope and land use aspects. Static groundwater levels were recorded from 95 wells for calibration. Water levels were measured in limited wells periodically since the year 2000. A systematic hydrogeological field survey of stream courses was made to determine the inputs for the river package such as riverbed hydraulic conductance, thickness of riverbed sediments, length and width of rivers at different reaches. The river bed hydraulic conductance was adapted from similar areas in the Ethiopian rift system [3]. The discharge data of the Meki River (MoWR, 2010) was used to check the aquifer-river relations locally.

Groundwater modeling often involves large geospatial data sets. A Geographic Information System (GIS) provides an integrated platform to manage, analyze, and display spatial data. It also greatly facilitates modeling efforts in data compilation, calibration and presentation. Furthermore, it can be used to generate information for decision making through spatial overlay and processing of model simulation results. GIS has also been used to develop the digital elevation model (DEM) of the catchment which in turn helped to define the top of the aquifer. The DEM was derived from the Shuttle Radar Topography Mission (SRTM) satellite data at a resolution of 90 by 90 meters, which was later converted to the resolution of the model grid. For this study one of the most widely used windows-based GIS software ArcGIS, surfer 10, global mapper 12 and 3D master is used.

IV. MODEL CALIBRATION Calibration checks that the simulation is

reproducing field measured heads and flows [10]. It involves adjustment and refinement of parameter structure and values to provide the best match between measured and simulated hydraulic heads and flows. Steady-state calibration was made using static water level observations of 95 wells. In the course of calibration adjustments on aquifer

thickness, hydraulic conductivity and recharge were made within reasonable ranges based on field hydrogeological observations and pumping test data.

Calibration can be achieved in two ways; forward and inverse problem solutions. In an inverse solution method one determines values for a given parameter structure and hydrologic stress using a mathematical technique, such as non-linear regression from information about head distribution [10]. This technique is sometimes called parameter estimation and it finds the set of parameter values that minimize the difference between simulated and measured quantities such as hydraulic heads and flows. The forward problem calibrates parameters, such as hydraulic conductivity and hydrologic stresses, are specified and the model calculates the head distribution. In this study, the forward solution is used by a conventional trial and error method in which model parameters were adjusted manually within reasonable limits of the existing data and field hydrogeological observations to achieve the best fit. The effectiveness of calibration is evaluated by comparing measured heads with simulated heads for all observation wells used. Two calibration criteria were used; visual matching of simulated contours to those of observed contours was set and matching simulated hydraulic heads at 90 % of the points to within 5 m of the observed hydraulic heads. The model was assumed calibrated when the fit between observed and calibrated heads was within this criteria and simulated groundwater contours (Fig. 2). The correlation coefficient is the found to be 0.996.

Fig.2 Scatter plot of simulated and observed groundwater heads in meter

Fig. 2 Scatter plot of simulated and observed groundwater heads in meter

V. RESULT AND DISCUSSIONS The steady-state model provided valuable

information on the groundwater balance, the response of the catchment for different stresses, the groundwater-surface water interactions and flow patterns. It has also shown the specific areas to be monitored and additional data gathered in the course of future detailed transient model development.


254

NCSTI-2012

A. Groundwater balance and the role of surface waters

Groundwater balance refers to the quantification of the inflow and outflow from part or the entire model domain. In a steady-state model simulation, inflows into and outflows from aquifers should be quantified. In this case groundwater inflow includes aerial recharge from precipitation and fluxes from rivers and lakes. An outflow includes base flow of rivers, well withdrawals and spring discharge.Table I summarizes the steady-state groundwater

Table. I. Groundwater balance and effect of each scenarios

balance of the entire model domain. The simulation was made under three different scenarios; with the aquifer system response to increased pumpage by 50% (scenario 1), decreased recharge by 25% (scenario 2), and complete disappearance of Lake Tuffa(scenario 3).

Inflow Outflow DifferenceScenarios Water balance components

(m3/day) (m3/day) (m3/day)

Chang with respect to the calibrated Value (%)

Calibrated value Constant head 98868 262635

Wells - 22570

Drains - 1158

River leakage 151919 664285

Recharge 699698 -

Scenario 1 Constant head 99203 261798 837 0

(Increasing pumping by 50%) Wells - 33855 -11285 -50

Drains - 1139 19 2

River leakage 153109 655381 8904 1

Recharge 699698 - - -

Scenario 2 Constant head 106989 240549 -22086 -8

(Decreased recharge by 25%) Wells - 22570 0 0

Drains - 855 -303 -26

River leakage 173487 541396 -122889 -18

Recharge 524960 - - -

Scenario 3 Constant head 20586 166359 -96276 -37

(Disappearance of Lake Tuffa) Wells - 22570 0 0

Drains - 1250 92 8

River leakage 150298 683849 19564 3

Recharge 702979 - - -


255

NCSTI-2012

The rationale of selecting the different scenarios is to assess the changes in the various water balance components under variable recharge rates. Changes in climatic conditions from time to time affect groundwater resources as a result of changing recharge rates. Recent studies clearly show large water resources variability and environmental changes in Ethiopia dominantly governed by climatic changes [15]. The study made by the United Nations Food and Agricultural Organization (FAO) in East African countries show wide variability of rainfall in the region over the last half a century[16]. B. Groundwater flow pattern

The model estimated groundwater head distribution reasonably agrees with the regional groundwater contour map reconstructed from limited wellhead measurements. Unfortunately piezometric data in the west, northwest and southwest is scarce making comparison difficult in these areas. The simulated head distribution shows the groundwater flow from western escarpment to east directions finally join Lake Ziway. The groundwater level is generally flat to gentle slope except at Tora-Koshe-Dugda-Ridge and the Cinder Cone areas. In these areas the groundwater contour shows steep slope probably due to the nature of the rocks or the fault systems separating these zones. The groundwater has a slope of 0.1% at Ziway-Plain; Tora-Koshe-Dugda ridge has 1.4%, Kuntane Inseno-Kela-Plain 0.3% and Cinder-Cone and Basaltic areas have 3.7%.The groundwater level

drops from 2000 m.a.s.l in Butajira-Crescent to 1800 m.a.s.l in Kuntane-Inseno area.

Fig.3. Model simulated groundwater contour

C. Sensitivity analysisSensitivity analysis was made to understand the

uncertainty in the calibrated model caused by limitation in the estimates of aquifer parameters and stresses. Groundwater models are sensitive to different model input parameters and parameters for which the model is most sensitive; small changes in those parameters will result in large differences in simulated heads or fluxes. The response of the calibrated numerical model to changes in model parameters like hydraulic conductivity and recharge was examined. During simulation, the effect of one parameter was being tested, the other parameters were kept to the steady-state calibrated value and each parameter was changed uniformly over the whole model domain. The magnitude of changes in heads or fluxes from the calibrated solution was used as a measure of the sensitivity of the model to that particular parameter.

In the model, simulated water levels were more sensitive to the decrease in the recharge values, mainly away from 40 percent. But it is more sensitive to the increase of hydraulic conductivity values, especially above 30 percent. Compared to recharge and hydraulic conductivity the model is less sensitive to the decrease or increase of pumpage. Model runs have been made by changing the hydraulic conductivity and recharge by the specified percent and the respective mean absolute error head changes in percent from the calibrated value are shown in Figure 3.

VI. LIMITATIONS AND RECOMMENDATIONS At this stage the model has provided valuable

hydrogeological information of the catchment. However, the steady-state model may not be readily used for detailed groundwater management purposes owing to data limitations on hydraulic conductivity, wellheads and aquifer thickness information in some areas of the catchment. The overall accuracy of the results depends on how these


256

NCSTI-2012input parameters are quantified and temporal variations are treated under transient conditions.

The results obtained here should not be interpreted as a perfect simulation, rather as system response projections within a fairly realistic model input parameters. Hence, the results should be interpreted and applied by considering all the limitations and drawbacks associated to the data gaps.

The authors believe that this study has laid the foundation for future detailed predictive transient model development. However, achieving a well-calibrated transient groundwater flow model requires filling the data gaps and collecting multi-temporal data at higher resolution. In this regard the following recommendations can be forwarded.

Great effort is needed to develop a 3D dimensional conceptual hydrogeological model. The well lithologic logs in some areas show different volcanic layers representing different hydraulic characteristics. These issues have to be seen in a detailed manner in the future.

The big data gaps on the hydraulic conductivity have to be filled by collecting more pumping test data. The pumping test data will also provide data for transient simulation such as specific yield, storage coefficient, etc.

A fair estimate has been done on the groundwater recharge, although more refinements can be made in the future according to the ever changing land use and expansion of the city with in the catchment which reduces the groundwater recharge. It is also essential to determine distributed multi-temporal recharge for transient models by accounting all influencing hydro-climatic factors.

Very little is known about the aquifer layers below the depth of 250 meters. It is essential to study this system by drilling test wells and using proper geophysical methods.

VII. CONCLUSION Both the conventional field hydrogeological

evidences and model simulations indicate that groundwater converges towards southeast (Lake Ziway).

The comparison made between the groundwater and topographic contours and the model sensitivity analysis indicate the dominance of topographically-driven flow system under water table conditions, although limited lithological differences and large faults form local discrete flows in fractured and weathered zones.

A shallow flow system mainly controlled by steep topography and sharp hydraulic contrast between geologic units is observed in the western escarpment or Gurage mountainous areas in the foot hills where a number of small and intermediate discharges springs emerge.

The model-based groundwater balance and flow system analysis indicates that surface water bodies (Lakes and rivers) are intimately linked to the productive aquifers systems. This has serious water

quality implications, if rivers continue to be polluted and large-scale groundwater pumping continues. This is more apparent in the rift floor, Butajira pediment field and Lake Tuffa. The scenario analysis reveals a real danger of over-pumping of the groundwater in the future. Increase of the existing pumping rate by over 50% will certainly bring about visible groundwater decline with in the catchment areas. This will result in the drying of low discharge springs, reduction in base flow and reversal of flow from contaminated rivers to the surrounding shallow aquifers. It is clear that with rapidly growing water demand of the cities with in the catchment, the existing wells will not be enough. Decrease in recharge by 25% will result in reduction of low discharge springs and increase stream flow but simulation without Lake Tuffa will results in increased recharge and groundwater outflow from drains.

ACKNOWLEDGEMENTS

The author is grateful to Pro. Tenalem Ayenew. Thanks go to the Geological Survey of Ethiopia, Ethiopian Meteorological Services Agency, Ministry of Water Resources, Ethiopian Mapping Authority and AG Consult for providing relevant data.

REFERENCES

[1] AAWSA. 2000. Addis Ababa water Supply project. Stage-IIIA Groundwater-Phase II, Modelling of Akaki Well field, V1, main report, Addis Ababa Water and Sewerage Authority, Addis Ababa, Ethiopia, 54p.

[2] Halcrow, Sir, & Sons, 1989. Rift valley lakes integrated natural resources development master plan. Ethiopian Valleys Development Studies Authority. Unpublished Technical Report of Sir Halcrow& Sons, Addis Ababa, Ethiopia.

[3] Ayenew, T. 1998.The hydrogeological system of the Lake District basin, Central Main Ethiopian Rift. PhD Thesis, Free University of Amsterdam. 259 pp.

[4] Ayenew, T. 2001. Numerical groundwater flow modelling of the Central Main Ethiopian Rift lakes basin, SINET: Ethiopian Journal of Science 24(2):167-184.

[5] Makin MJ, Kingham TJ, Waddams AE, Birchall CR, Eavis BW., 1976. Prospects for irrigation development around Zwai, Ethiopia. Land Resource Study 26, Land Resources Division, Surbiton, UK.

[6] MacDonald, M.G. and Harbaugh, A.W. 1988. A modular three-dimensional finite difference groundwater flow model, Book 6. Modelling techniques, techniques of water resources investigation of the United States Geological Survey.

[7] Winter, T.C. 1978. Numerical simulation of steady-state three dimensional groundwater flow near lakes. Water Resources Research. 14:245-254.

[8] Rouleau, A., 1988. A numerical simulator for flow and transport in stochastic discrete fracture network. NHRI paper 39. National Hydrology Research Institute, Saskatoon, Saskatchewan. 204 pp.


257

NCSTI-2012

[9] Ayenew, T., Kebede, S., Alemyahu, T., 2007b. Environmental isotopes and hydrochemical study as applied to surface water and groundwater interaction in the Awash river basin. Hydrological processes (published online DOI: 10.1007/s00254-007-0914-4).

[10] Anderson, M.P. and Woessner, W.W. 1992. Applied Ground Water Modeling. Simulation of Flow and Advective Transport, Academic Press. Florida.

[11] AG Consult. 2008. Hydrogeology and groundwater modeling, Ziway and Butajira areas. Ministry of Water Resources (MoWR), Addis Ababa, Ethiopia.

[12] Jacob, C.E., 1944. Notes on determining permeability by pumping test under water table conditions. U.S.G.S open file report.

[13] Neuman, S.P., 1975. Analysis of pumping test data from anisotropic unconfined aquifers considering delayed yield, Water Resources Research, vol. 11, no. 2, pp. 329-342.

[14] Theis, C.V., 1935. The relation between the lowering of the piezometric surface and the rate and duration of discharge of a well using groundwater storage, Am. Geophys. Union Trans., vol. 16, pp. 519-524.

[15] Ayenew, T. and Legesse, D., 2007. The changing face of the Ethiopian rift lakes: Call of the time. Lakes and reservoirs: research and management. 12 (3): 149–165.

[16] FAO, 1995. FAOCLIM 1.2. A CD with Agroclimatic data for Inter-Governmental Authority on Drought and Development (IGAAD).ow near lakes. Water Resources Research. 14:245-254.


258

NCSTI-2012


Getachew Tegegne1, Dereje Hailu2

1Department of Civil Engineering, Adama Science and Technology University, Adama, P.o.box-1888, Ethiopia, email: [email protected]

2Department of Civil Engineering, Addis Ababa University, Addis Ababa,P.o.box- 32490,Ethiopia

Abstract - This paper presents simulation of Lake Tana reservoir future water use under emerging scenario with and without climate change impacts. Two different development and climate change scenarios were developed to simulate Lake Tana water level i.e., i) Base line scenario (1991-2000) ,ii) Future development scenario on short term periods(2031-2040) , and ii) Future development scenario on long term periods (2091-2100). River head flow estimated by Soil and Water Assessment Tool (SWAT) was used as an input to Water Evaluation And Planning (WEAP) model to simulate the Lake level for each scenario. Based on WEAP model simulation results, demand coverage and reliability of 100% was observed in all scenarios for Tana-Beles hydropower project. For scenarios without climate change impacts, there are longer periods of time when mean monthly lake levels are below 1785 masl (i.e., the minimum lake level required for shipping). Under natural conditions (lake level without project), they exceed this level in 100%.under current conditions (Base line scenario, BLS), they exceed this level in 89% of the months. In the full development scenario (FDSCʹ), this will decrease to 83%. For all scenarios with climate change impacts, Lake water Level will not significantly be affected by climate change impacts. Keywords: SWAT, WEAP, Lake Level, Lake Tana, Climate Change, Reliability.

I.INTRODUCTION The increased demand of water for agriculture,

industries, domestic, and power generation in Lake Tana sub-basin requires proper planning and management of water resources in the basin. The basin has more than 40 rivers inflow in to Lake Tana and about 93% of the inflow is coming from the four major rivers Gilgel Abbay, Gummera, Rib and Megech [3].

The purpose of this study is therefore applying

a physically based semi distributed model called Soil and water assessment tool (SWAT), to understand the hydrology of the basin, to know the water resource potential as a whole from gauged and un-gauged catchments as well as water evaluation and planning (WEAP) model[6] were used to asses upstream catchment development and climate change impact on Lake Tana water level and to assess the sustainability of Tana –belles Hydropower plant on the basis of

adjusting the operation rule of Lake Tana reservoir.

The analysis presented in this paper is the first of its kind which was done using actual data in Lake Tana sub-basin. The study addresses; The assessment of water resources potential of Lake Tana basin, Assessment of Impact of upstream irrigation development on Lake Tana water level and Tana-belles Hydropower plant, Assessment of impact of climate change on Lake Tana and Beles hydro power plant with and without emerging upstream irrigation project, and assessment of the sustainability of Tana-Beles development on the basis of adjusting the operation rule of the Lake.

The output of this study can be used as an input

for decision support for water resources planning, development, and management of water resources in the basin.

II.STUDY AREA DESCRIPTION Lake Tana Basin is part of the Blue Nile basin,

which lies in a natural drainage basin of about 15114 Km2 as per this research work using

SWAT delineation. Among which about 20.47% is covered by the Lake Tana. Lake Tana basin is found in North West part of Ethiopia and it


259

NCSTI-2012

extends between 10.950N to 12.78oN latitude and from 36.89oE to 38.25oE longitude (highlighted in fig.1).

Topography is generally uniform and quite well adapted to irrigation development surrounding Lake Tana [5, 9]. The elevation ranges between 914 m to 4096 m +MSL, which is extracted from DEM (90*90m) resolution. There are two seasons rainy and dry. The rainy season has two periods, the little rains, during April and May, and the big rains, which last from mid- June to mid-September. The rainfall distribution in the basin is found to be a mono-modal pattern i.e. one peak value observed during rainy season especially in July, and August. Considering the rainfall stations in the basin for a period of 1996-2006 the mean annual rainfall amount ranges between 813 mm in Yifag and 2328 mm in Enjibara. Similarly the mean annual

minimum and maximum temperature ranges between 9.3 oC in Dangila and 29.6 oC in Gorgora respectively.

Land use of study area was classified based on Abay river master plan study conducted by BCEOM, in 1996-1999[2], about 51.37 % of the watershed area was covered by Agriculture, 21.94 % by Agro-pastoral, 20.41 % by Lake Tana, 0.39 % by Agro-Sylvicultural, 0.13 % by wetland, 5.47 % by Pastoral, 0.15 % by Sylvicultural, 0.03 % by sylvo-pastoral and 0.11 % by Urban.

The soil classification for the study area is also adopted from Abay river master plan study in 1996-1999 conducted by BCEOM [2]. Based on the classification Halpic luvisol which covers about 20.68 % of the watershed area is considered to be the major dominant soil in the study area.

Figure 1: Location Map of Lake Tana Basin


260

NCSTI-2012

III.METHODOLOGY

The following flow chart indicates the overall framework of the methodology to be followed throughout the study.

Meteorological Data

Hydrological Data

GIS Data

Figure 2: General frame work adopted for the study

DEM (90*90m)

Land use Map

Soil Map

SWAT Model Application on gauged

Catchments Calibration

Model parameters at gauged catchments

Parameters transferred to un-gauged catchments based on similar HRUs

Simulated Stream flow at un-gauged

& gauged catchments

Reservoir physical & operation data

Meteorological, lake evaporation, & areal

rainfall data

Water use at demand

sites

WEAP model configuration & scenario development

Runoff contribution from gauged catchments

Runoff contribution from un-gauged

catchments

Lake Tana water level for various scenarios

U/s catchments development and climate change

impact on lake level

Validation

Water balance components & river head flow

Legend

Inputs

Process

Result


261

NCSTI-2012

IV.SCENARIO DEVELOPMENT In this study the model was set-up to simulate

two scenarios based on development plans of the basin. The model is first configured to simulate a base line scenario, for which the water availability and demands can be confidently determined. It is then used to simulate for future development scenarios to assess the impact of development and climate change on the hydrology and water resources.

a) Scenario BLS: Base line scenario(1991-2000), b) Scenario FDSCʹ: Future development scenario (future water demand without climate change); c) Scenario FDSC1: Future development scenario under climate change on short term basis (2031-2040),and d) Scenario FDSC2: Future development scenario under climate change on long term basis (2091-2100).

Interventions in the water sector in Abbay basin fall in to three main areas: irrigation, hydropower, and water supply. However, water supply requirements are small relative to those for irrigation and hydropower [2]. Projects, like water supply and sanitation, that do not significantly influence the results of water availability in the basin was not also be considered. A. Scenario BLS

This Scenario represents the existing development in the basin. Relative to the basin water resource potential; development activities achieved yet is insignificant. Koga irrigation, Tis Issat fall and Tana Beles Hydropower schemes was considered for the base line scenario. When Tana Beles transfer become operational Tiss Abbay I and II hydropower stations will be used as standby stations, only to operate as a backup system when problems in the National grid may require [2,7]. Hence they are not considered in this scenario.

Figure 3: Schematic of the model configuration for the Current situation Scenario

B. Scenario FDSCʹ This Scenario represents full development

activities in the basin which are expected to be operational in future period of time. The analysis includes projects which are currently operational, ongoing development, and likely development activities. The scenario not considers the impact of climate change on hydrology and water resources in the basin. C. Scenario FDSC1 & FDSC2

This Scenario represents full development activities which are expected to be operational in the long period of time in the future. Scenario FDSC1 represents future development scenario with climate change for time periods of 2031-2040 and scenario FDSC2 represents future development scenario with climate change for time periods of 2091-2100.The analysis includes projects which are currently operational, ongoing development, and likely development activities. The scenario considers the impact of climate change on hydrology and water resources in the basin. i.e., the climate variables are under the influence of climate change in the future.


262

NCSTI-2012

Figure 4: Schematic of the model configuration for the future development Scenario.

V.RESULTS AND DISCUSION A. Evaporation and rainfall over the lake

From the Thissen polygon analysis, the annual average areal rainfall over the Lake for the simulation periods of (1991-2000), (2031-2040), and (2091-2100) found to be 1291mm/year, 1737.693mm/year, and 1690.104 mm/year respectively. From CROPWAT model the average annual evaporation over the Lake for the simulation periods of (1991-2000), (2031-2040), and (2091-2100) found to be 1618 mm/year, 1767 mm/year, and 1909 mm/year respectively.

Figure 5: Monthly average rainfall and evaporation over Lake Tana.

B.Modeling of Gauged Catchments Historical observed stream flow of Gilgel Abay

at Merawi, Gummera at Bahirdar, Rib at Addiszemen and Megech at Azezo were calibrated from a period of 1996-2002 and validated from a period of 2003- 2005 . Table 1: Calibration & validation statistics of observed and simulated stream flow Average monthly flow

(m3/sec.) Observed Simulated

R2

NSE

Gilgel Abay River (Calibration period 1996-2002) 57.26 58.21 0.91 0.91 Gilgel Abay River (Validation period 2003-2005) 34.64 32.95 0.93 0.92 Gummera River (Calibration period 1996-2002)

37.74 38.8 0.70 0.70 Gummera River (Validation period 2003-2005) 34.06 27.69 0.91 0.90 Rib River (Calibration period 1996-2002) 14.93 15.82 0.82 0.82 Rib River (Validation period 2003-2005) 13.88 13.28 0.84 0.83 Megech River (Calibration period 1996-2002) 7.18 7.04 0.8 0.76 Megech River (Validation period 2003-2005) 8.06 4.53 0.92 0.91


263

NCSTI-2012

C.The inflow hydrograph from gauged and ungauged catchments

Once the model is calibrated and verified at the gauged location the model output during that period were quantified and taken as simulated inflow series. Later this inflow series will be used for water balance analysis.

Similarly, the inflow series for ungauged catchments were done by transferring calibrated parameters having the same HRUs as gauged catchments. The total inflow in to the Lake mouth was determined after having the inflow from gauged catchments and inflow from ungauged catchments separately and later the total inflow was taken as the aggregate of inflow series from gauged and ungauged catchments.

From the model result total inflow from gauged catchments was found 2850.727MCM, 3595.137MCM, and 3311.873MCM for time period (1991-2000), (2031-2040), and (2091-2100) respectively. Total inflow from ungauged catchments was found 3759.228MCM, 5382.034MCM, and5006.76MCM for time periods of (1991-2000), (2031-2040), and (2091-2100) respectively.

Figure 6: Total Inflow Hydrograph from gauged and ungauged catchment for all scenarios.

D.Elevation Area Volume Relation ship The elevation Area Volume relation ship of Lake

Tana reservoir was calculated for the calibration period (1996-2002).The polynomial fitted

bathymetry by Pietrangeli and Abeyou [1, 8] used in this research work is as follows: Elevation-Volume-Area relation ship as per Pietrangeli and Abeyou Table 2: Elevation Volume Area relation ship Pietrangeli

E = 1.08*10-9(V)2+3.88*10-

4(V)+1775.58

A = 6.20*10-8(V)2+1.72*10-

2(V)+2516.3

Abeyou

E = 1.21*10-13(V)3-1.02*10-

8(V)2+6.20*10-4(V)+1774.63,

A = 7.93*10-11(V)3-5.81*10-

6(V)2+1.65*10-1(V)+1147.51

Where E= Lake level elevation, m. +MSL A= Surface area of the Lake, Km2 V= Lake volume, MCM The basic equation used in the water balance:

outintt GGtEtOtPtISS )()()()(1 ∆s (1)

Where: t Lake storage volume at the end of current

month, S

1t Lake storage volume at the end of previous month, S

I(t) =Simulated inflow volume from gauged and un-gauged catchments at current month, O(t) = Outflow volume at the Lake outlet, P(t)=Areal rainfall volume on the Lake surface, E(t)=Evaporation volume on the Lake surface,

in (t) = Ground water inflow in to the Lake at the end of current month, G

out (t)= Ground water outflow from the Lake at the end of current month. G

∆s = other losses. The water balance terms were computed using EXCEL spread sheet model and the monthly water balance result obtained by using the relation ship developed by Abeyou, (2008) has been best fitted than Pietrangeli, (1990).


264

NCSTI-2012

(lake level without project), they exceed this level in 100%.under current conditions (Base line scenario, BLS), they exceed this level in 89% of the months. In the full development scenario (FDSCʹ), this will decrease to 83%.

Figure 7: Observed and Simulated Lake Level with out project for the period 1996-2002. Table 3: Lake Tana Annual water balance components simulated from 1996-2002

Water balance components mm/year

Lake areal rainfall +1291

Gauged River inflow +822 Figure 8: Comparison of simulated lake level with project and without project (1991-2000) Un-gauged river inflow +1297 Lake Evaporation -1618

River outflow -1725 F.Upstream Catchment Development Impact on Tana-Beles Hydropower Plant Change in storage

67

E.Upstream Catchment Development Impact on Lake Tana Water Level

Figure 8 presents a comparison of the time series of simulated lake levels with project and without project for all scenarios. The results indicate the decline in mean annual lake levels, and consequently lake area, as water resources development in the catchment increases. As water resources development increases there are longer periods of time when mean monthly lake levels are

Table 4 and 5 presents annual unmet demand in million cubic meter for base line scenario and full development scenario respectively. From the result, 100 percent reliability of Tana-Beles hydropower project was observed for base line scenario and full development scenario.

below 1785 masl (i.e., the minimum lake level required for shipping) [4]. Under natural conditions

Table 4: Yearly unmet demands (MCM) for BLS scenario (1991-2000)

Scheme 1991* 1992* 1993 1994 1995 1996 1997 1998 1999 2000 Sum Tana Beles Hydropower

0 0 0 0 0 0 0 0 0 0 0

Koga Irrigation

47.4446 0.00407 0 0 0 0 0 0 0 0 47.44

Sum 47.4446 0.00407 0 0 0 0 0 0 0 0 47.44 1991* 1992* shows the "warm up'' period for reservoirs filling and not considered for analysis


265

NCSTI-2012

Table 5: Yearly unmet demands (MCM) for FDSCʹ scenario (1991-2000)

Scheme 1991* 1992* 1993* 1994 1995 1996 1997 1998 1999 2000 Sum Tana Beles Hydropower

0 0 0 0 0 0 0 0 0 0 0

Gilgel Abay Irrigation

91.37 46.72 34.31 0 0 0 0 0 0 172.40

Gumera Irrigation

26.01 0 0 0 0 0 0 0 0 0 26.01

Koga Irrigation

48.300 33.21 0 0 0 0 0 0 0 0 81.51

Megech Irrigation

39.639 29.15 13.61 0 0 0 0 0 0 0 82.41

Rib Irrigation 166.94 42.64 19.86 0 0 0 0 0 0 229.44 Tana Pump Irrigation

0 0 0 0 0 0 0 0 0 0 0

Sum 372.26 151.73 67.79 0 0 0 0 0 0 0 591.79 1991* 1992* 1993* shows the "warm up'' period for reservoirs filling and not considered for analysis.

G.Climate Change Impacts on Lake Tana Water Level

Currently, there is great uncertainty about the likely impacts of climate change in the Abay Basin. Results from Global climate models (GCMs) are contradictory; some show increases in rainfall whilst others show decreases. A recent study of 17 GCMs indicated that precipitation changes between -15% and +14% which, compounded by the high climatic sensitivity of the basin [4]. Generally there is an increasing trend in both precipitation and runoff in the basin for the time

period of 2031-2040 and 2091-2100. PET and reservoir evaporation shows an increasing trend for all future scenarios. The cumulative impacts of these hydrologic parameters on Lake water Level were checked using WEAP simulation model.

The results of the WEAP model simulation

shows that Lake water Level will not significantly be affected by climate change impacts for all future scenarios. Figure 9 and 10 presents a comparison of the time series of simulated lake levels for all future scenarios with and without project.

Figure 9: Comparison of simulated Lake Levels with and without project (2031-2040)

Figure 10: Comparison of simulated Lake Levels with and without project (2091-2100)


266

NCSTI-2012

H.Climate Change Impacts on Tana-Beles Hydropower Project

The result shows climate change impacts are not significant in the basin and hence, there is no problem of water shortage for proposed

development activities due to climate change impacts. By giving higher priority for hydropower schemes than irrigation schemes, the demand coverage and reliability of 100 percent was observed for Tana– Beles hydropower project

Table 6: Yearly unmet demands (MCM) for FDSC1 scenario (2031-2040

Scheme 2031* 2032* 2033 2034 2035 2036 2037 2038 2039 2040 Sum Tana Beles Hydropower

0 0 0 0 0 0 0 0 0 0 0


97.63 0.002 0 0 0 0 0 0 0 0 97.63

Gummera Irrigation

25.42 0 0 0 0 0 0 0 0 0 25.42

Koga Irrigation 48.14 0.004 0 0 0 0 0 0 0 0 48.14 Megech Irrigation

39.44 0.01 0 0 0 0 0 0 0 0 39.45

Rib Irrigation 164.12 0.99 0 0 0 0 0 0 0 0 165.11 Tana Pump Irrigation

0 0 0 0 0 0 0 0 0 0 0

Sum 374.77 1.01 0 0 0 0 0 0 0 0 377.7 2031* 2032* shows the "warm up'' period for reservoirs filling and not considered for analysis

Table 7: Yearly unmet demands (MCM) for FDSC2 scenario (2091-2100)

Scheme 2091* 2092* 2093* 2094 2095 2096 2097 2098 2099 2100 Sum Tana Beles Hydropower

0 0 0 0 0 0 0 0 0 0 0


106.71 0.006 0 0 0 0 0 0 0 0 106.71

Gummera Irrigation

26.14 0.01 0 0 0 0 0 0 0 0 26.15

Koga Irrigation

48.30 0 0 0 0 0 0 0 0 0 48.31

Megech Irrigation

39.64 30.41 15.74 0 0 0 0 0 0 0 85.80

Rib Irrigation

167.93 19.27 0 0 0 0 0 0 0 187.20

Tana Pump Irrigation

0 0 0 0 0 0 0 0 0 0 0

Sum 388.72 49.70 15.74 0 0 0 0 0 0 0 454.18 2091* 2092* 2093* shows the "warm up'' period for reservoirs filling and not considered for analysis

VI.SUMMARY AND CONCLUSION A cascade of two models was used in this study.

The SWAT model was setup from January 1985 - December 2006.Calibration and validation was done for seven years monthly time step (1996-2002) and three years monthly time step (2003-2005) respectively. After modeling the gauged

watershed, calibrated parameters were transferred to ungauged watershed by lumping the parameters having the same hydrologic response unit (HRUs). The model output indicates that, the total annual inflow volume from gauged and ungauged catchments estimated to be 6229.115 MCM for calibration period. From the Thissen polygon analysis, the annual average areal rainfall over the Lake for the simulation periods of (1991-2000),

Trend Analysis of Runoff and Sediment fluxes using Statistical and Physically based models: ...

267

NCSTI-2012

n

Trend Analysis of Runoff and Sediment Fluxes Using Statistical and Physically Based Models: Upper Blue Nile Basin

T.Gebretsadkan1, Y.A. Mohamed1,2,3, G.D. Betrie4, P. Van der Zaag1, 2, E.Teferi1,5

1UNESCO-IHE Institute for Water Education, P.O. Box 3015, 2601DA Delft, The Netherlands 2Delft University of Technology, P.O. Box 5048, 2600 GA Delft, The Netherlands 3Hydraulic Research Station, P.O.Box 318, Wad Medani, Sudan 4School of Engineering, University of British Colombia Kelowna, BC, Canada V1V 1V7. 5College of Development Studies, Addis Ababa University, P.O. Box 2176, Ethiopia

Abstract — The landuse/cover changes in the Ethiopian highlands have significantly increased the variability of runoff and sediment fluxes of the Blue Nile River during the last few decades. The objectives of this study were to understand the long-term variations of those fluxes using statistical models, verify the statistical results using a physically-based hydrological model. The Mann-Kendall and Pettit tests were used to test the trends of Blue Nile discharge (1970 to 2009) and sediment load (1980 to 2009) at the outlet of Upper Blue Nile, at El Diem Station. These tests showed significant increasing trend of annual stream flow, wet season stream flows and sediment load respectively. The dry season flow showed a significant decreasing trend. However, during the same period the annual rainfall over the basin showed no significant trends. The results of the statistical tests found to be sensitive to the time domain. The Soil and Water Assessment tool (SWAT) model was used to simulate the runoff and sediment fluxes in the early 1970s and at the end of the time series (2000s) in order to detect the physical causes of the trends. A comparison of model parameters values between the 1970s against 2000s shows significant change, which could explain catchment response changes over the 28 years of record. The results of the statistical test and SWAT model have resulted in significant change of runoff and sediment load from the Upper Blue Nile. This is an important finding to guide optimal water resources development in the whole basin, both upstream in the Ethiopian highlands, and further downstream in the plains of Sudan and Egypt.

Keywords: Upper Blue Nile, SWAT, Trend analysis, Stream flow, Sediment transport

1. IntroductioThe Upper Blue Nile River basin contributes more

than 60% of the Nile water. It is crucial for the socio-economic development and environmental stability of the three riparian countries, Ethiopia, Sudan and Egypt. However, landuse/cover and climate changes have affected the value of the Blue Nile’s water, through increasing inter-annual and inter-decadal variability of runoff and sediment fluxes [10], [25]. These changes result in negative impact for both upstream and downstream countries. In the Ethiopian highlands, landuse change has led to severe soil erosion which reduced soil moisture capacity and challenged food production [23], [43]. The downstream countries (Sudan and Egypt), have experienced problems on their storage reservoirs and irrigation canals due to excessive sediment loads [14]. The literature showed an increase of sediment yield at the Upper Blue Nile outlet (El Diem gauging station) from 111 to 140*106 ty-1 [5], [11], [18], [33].

A number of local and basin level studies about the Blue Nile have been reported in the literature, e.g., the long term trend analysis of runoff by [6], [16], [26], [28]. [41].They have shown no consensus on conclusions of the flow trends. [9] and [28] showed an increasing trend of the Blue Nile annual flow, while [6], [16] indicated a decreasing trend. Analyzing 40 years of data (1964 to 2003), [41] showed no change of annual flow from the Upper Blue Nile basin, but with an increase of flood season flow.

There are a number of studies estimated the annual sediment load from the Upper Blue Nile basin, (e.g., [5], [11], [41], though showed a disparity among different researchers. To our knowledge, no study was done to analyse the long term trend of the sediment load. The disagreements on the hydrological trends and on the amount of annual sediment load show a limited understanding of the underlying causes.

Therefore, the objectives of this study were (i) assessing the long-term variability of runoff and sediment fluxes of the Upper Blue Nile using


268

NCSTI-2012

ea

statistical methods; (ii) infer the causes of changes in runoff and sediment load as a function of landuse change derived from changes of physically based model parameters (SWAT).

The remaining part of this paper is structured as follows: we give a brief description of the study site in Section 2, followed by introducing and brief discussion of material and methodology in Section 3. Section 4 presents results and discussion, and finally Section 5 presents the conclusion and main findings.

2. Description of study arThe Upper Blue Nile basin (locally called Abay) is

located in the north western part of Ethiopia (Fig.1). The topography of the basin is composed of highlands, hilly in the north-eastern part and dominated by valleys in the southern and western parts. The elevation varies from 480m near the Sudan/Ethiopian border, to over 4000m near the center. The climate of the basin is tropical highland monsoonal with the majority of the rain falling between June and October. The average rainfall of the basin varies from 1000 mm/year in the north-east to above 2000 mm/year in south-east of the basin. Over 80% of the annual flow occurs from July to October flowing directly to the downstream countries [40].

The basin is composed of mainly volcanic rocks and preCambrian basement rocks with small areas of sedimentary rock [8]. The dominant soil types are latosol and Alisols 21%, Nitosols 16%, Vertisols 15% and cambisols 9% (Betrie et al., 2011). The dominant land cover of the basin are Savannah, dry land crop and pastures, grassland, crop and woodland, water body and sparsely vegetated [2].

Figure 1: Location map of the Upper Blue Nile basin

3. Material and Methodology3.1. Input data

The datasets used in this study include: soil, climatic (e.g., rainfall and temperature), runoff, sediment, and landuse/cover maps. Long term records of monthly data on rainfall, runoff and sediment load were used for the statistical analysis. Daily climate, runoff, and sediment load data were used for the SWAT modelling. Satellite images (Landsat) of 1973 and 2000 were used to detect long-term landuse change.

The annual rainfall data from 1970 to 2005 of nine gauging stations were used to detect long term trends of rainfall over the Upper Blue Nile basin. Because of limited spatial distribution, rainfall stations closer to headwater of the tributaries were selected to ensure better coverage. The monthly river discharge data from 1970 to 2009 and suspended sediment loads from 1980 to 2009 at El Deim station were used to assess seasonal and annual trends of flow and sediment, respectively. The observed daily data of discharge and sediment load were used for SWAT model simulations. The available data shows that there is a lack of adequate sediment data in Upper Blue Nile. Therefore, we assumed equal sediment concentrations at El Diem and Roseires (110 km apart) as both stations were linearly correlated. The data from the latter were used to fill in missing gaps at El Diem. The good correlation (R2=0.88) between sediment concentration of the two stations indicates realistic assumption.

The sediment concentration in the Blue Nile is measured only during the rainy season, from June to October, and assumed negligible during the remaining months [5], [11]. The observed sediment data were complete for simulation of the model in year 2000 to 2005, but with many missing values during 1970 to 1976 simulation period, with measurements only available for 1972 to 1973. Therefore, the model in the 2000s was calibrated from observed data whereas for 1970 to 1976 simulation period, sediment data derived from the NBCBN’s rating curve has been used. The NBCBN derived a sediment rating curve based on observed data during the early 1970s, [33]. However, the available measurements for period 1972 to 1973 have been used to verify data derived from the rating curves, and showed high correlation with coefficient of R2=0.80.

The observed daily data of precipitation and minimum and maximum temperature were obtained from 27 stations for daily rainfall and 19 stations for daily temperature. The daily data were used to run the SWAT model for two simulation periods (i) 1970


269

NCSTI-2012

to 1976, and (ii) 2000 to 2005. A weather generator model based on statistical summaries of long-term monthly means was used to generate the relative humidity, solar radiation and wind speed data for 18 stations within the basin. 3.2. Experimental approach

3.2.1 Man-Kendall and Pettitt tests

The long term trends of the hydrological and sediment fluxes were estimated using the non parametric statistical tests of Mann-Kendall (MK) and Pettit [10], [25]. Mann-Kendall test is a rank based method for trend analysis of time series data [7], [41]. First the presence of monotonic increasing/decreasing trend was tested using Mann-Kendall test. Secondly, the Pettit test was applied to investigate the difference between cumulative distribution functions before and after a time instant. The significance of any trend in the data set is provided in 'no trend', 'an increasing or a decreasing trend' designation based on defined confidence level [29]). Man-Kendall calculates Kendall's statistics (s), the sum of difference between data points and a measure of associations between two samples (Kendall's tau) to indicate increasing or decreasing trend. Positive values of those parameters indicate a general tendency towards increasing trend while negative values show a decreasing trend. The Pettit test is a non parametric test that requires no assumption about the distribution of the data and is used to identify if there is a point change in the data series.

However, the presence of serial correlation can increase the expected number of false positive outcomes for those statistical tests [46], [48], [50]. The existence of positive and negative serial correlation results an overestimation and underestimation of the probaplity of trends, respectively. Therefore, it is essential to eliminate or reduce the influence of serial correlation on the tests and in this study, the trend-free pre-whitening (TFPW) method developed by [48] was used. This approach involves estimating a monotonic trend for the series and then removing the trend prior to pre-whitening the series [7].

The gradual trend test (Man-Kendall), and abrupt change test (Pettit) have been employed on seasonal and annual discharge series of 1970-2009 and sediment load series of 1980-2009 at the outlet of the basin (El Diem Station). The rainfall data series of nine stations were also tested against long term trends.

3.2.2 The SWAT model description

The SWAT model was used to interpret results of the statistical tests, and infer if the long-term trends are attributed to landuse change. The SWAT model describes the relationship between inputs (e.g. rainfall), the system condition (e.g. landuse/cover) and the outputs (e.g. stream flow and sediment load). To this effect, two independent SWAT simulations were performed from 1970 to 1976 and from 2000 to 2009. The difference between model parameters values of the two simulations could explain reasons for the envisaged trends of runoff and sediment fluxes [42].

The SWAT model is a conceptual, GIS interface tool that operates on daily time step to envisage the impact of landuse and climate change on water, sediment and agricultural yields from large watershed with varying soil, landuse and management practices over a certain period of time [4], [34]. The model divides the basin into subbasins and further into hydrological response units (HRUs) with homogenous soil type, slope, landuse and management practices.

The SWAT model computes surface runoff with two methods, the soil conservation service (SCS) curve number (CN) method [44] and the Green-Ampt infiltration method, [19]. The CN method was used in this study because of its capability to use daily input data [4], [5], [34], [37], [43].

The SWAT model simulates the hydrology into land phase and the routing phase. In the land phase, the amount of water, sediment and other non-point loads are calculated from each HRU and summed up to the level of sub-basins. Each sub-basin also controls and guides the loads towards the basin outlet. The routing phase defines the flow of water, sediment, and other non-point sources of pollution through the channel network to the outlet of the whole catchment [34]. SWAT computes the soil erosion at an HRU level using the modified Universal Soil Loss Equation MUSLE [47]. This constitutes the sediment yield from each sub-basin and is routed to the basin outlet. Detailed steps for computing of hydrological and sediment components are available in the literature (e.g. [34]). Although SWAT provides three methods for estimating potential evapotranspiration: Penman-Monteith [30], [36] and Hargreaves methods [20], the Hargreaves method was used in this study since it is most suitable where data are limited.


270

NCSTI-2012

n

3.2.3 The SWAT model setup

The SWAT model was built for two simulation periods 1970 to 1976 and 2000 to 2005. The Digital Elevation model (DEM) obtained from Global US geological survey) ite, soil map and landuse maps were used by SWAT to delineate the HRUs. The Landuse/cover maps were prepared from Landsat MSS and ETM+ imageries [17].

The soil map obtained from global soil map of the Food and Agriculture Organization [13]. It includes more than 500 soil types, at a spatial resolution of 10 km. The soil physical properties (e.g. bulk density, available water capacity, hydraulic conductivity, saturation hydraulic conductivity, particle-size distribution were taken from [5].

The landuse, soil and topography maps were overlaid to create a total number of 1553 HRUs over the Upper Blue Nile. The HRUs were selected by ignoring landuse, soil and slope areas covering less than 5% of the total sub-basin area. This was necessary to reduce computation time of the model.

The simulation period of the first model was done from January 1, 1970 to December 31, 1976. The first year was used for model warm-up, the years 1971 to 1973 were used for model calibration, and the years 1974-1976 were used for model validation. The simulation of the second model was performed from January 1, 2000 to December 31, 2005 with the first three years (2000-2002) for model calibration and the last three years (2003-2005) for model validation. The two periods were selected to detect landuse change over a relatively long period. It includes the period of high landuse changes of the 1980s [49].

Sensitivity analysis was done to identify the most sensitive SWAT model parameters for model calibration. In this study, the automatic sensitivity analysis extension of SWAT, Latin-hypercube one factor at a time (LH-OAT) algorithm developed by [45] which is part of SWAT model was used to identify the most sensitive parameters. Next, the identified most sensitive parameters were automatically calibrated using sequential uncertainty fitting algorithm SUFI-2 [1]. The validation was done by running the same model for different simulation periods. The performance of the model for both runoff and sediment load was then evaluated using statistical indices [31], Nash-Sutcliffe coefficient of efficiency (E) and Coefficient of determination (R2). Graphical comparisons of the daily simulated and observed time series and water balance checks have been used in the validation as well.

After obtaining best fitting parameters for flow and sediment simulations from the early 1970s and late 2000s models, two different approaches were used to detect causes of runoff and sediment. First, we compared parameter values for the two periods assuming those values are not the same if landuse has changed in the basin. Next water balances and annual average sediment yields were compared.

4. Result and discussioThe changes in catchment response as manifested

in a modified pattern of stream flow and sediment load could be attributed to climate variation (mainly rainfall) or landuse change. Often landuse changes are attributed to anthropogenic impacts, for instance, large scale deforestation, urbanization, and/or agricultural expansions. To understand catchment response behaviour, it is important first to quantify the changes occurred in hydrological fluxes of runoff and sediment load.

3.3. Trend analysis

The Mann-Kendall and Pettitt tests were applied to the annual rainfall pattern at nine stations in Upper Blue Nile. The results showed no change of annual rainfall for the last 36 years (1970-2005). This result well agrees with earlier studies in the basin (e.g., [8]. [12], [42] This is an important conclusion, implies that interannual rainfall pattern is not the major driver for the trend changes of runoff and sediment fluxes in the Upper Blue Nile.

The trend analysis of the seasonal and annual stream flows as computed by Man-Kendall and Pettitt tests are summarized in Table 1 and Fig.2(a-d). The Man-Kendall results were given at 5% significance level.

The results in Table 1 show a significant increasing trend of stream flow during the wet season, short rainy season, and annual time period and a decreasing trend of stream flow during the dry season. These results were supported by Pettitt test, which shows a significant abrupt upward change of stream flow. Most of these changes occurred in the early 1990s (Fig. 2a, and 2b). A significant abrupt downward change of the dry season stream flow occurred in 1979 (Fig.2c). Figure 2d shows the increasing trend of annual stream flow from the basin. To further validate the findings, the trend of annual flows at three key locations of the basin (Bahirdar, kessi and Dedesa) were analyzed. The change point at Bahir Dar and Kesssie is consistent, in 1991-92, while for Dedessa occurred 6 years later, and no reason to reject it. These results of dry and


271

NCSTI-2012

wet seasons well agree with [42] but the results of short rainy season and annual flows do not agree. [42] reported that short rainy season and annual flows are constant for the analysis period of 1964-2003, whereas our study (from 1970 to 2009) showed an increasing trend in both cases. However, we obtained similar results to [42] for the same period of analysis (1964 to 2003). Hence, it is clear that the period of analysis is a crucial factor to determine the given trends. Most likely that the last six years (2004 to 2009) showed relatively higher discharges.

As the rainfall over the basin has remained constant, the increasing trend of runoff from the Upper Blue Nile could be attributed to landuse change within the basin. A decreasing trend of the dry season flow (base flow), and an increasing trend of wet season flow (peak flow), while no change of the rainfall suggests modifications of catchment response that led to an enhanced surface runoff from the Upper Blue Nile basin. The increase of the sediment load by 61% during the past 30 years could be attributed to modified runoff process associated

with large scale landuse change that boosted soil erosion in the basin. Direct runoff is the only flow responsible for soil erosion and sediment transport in the stream [39]. The annual flow showed a significant increasing trend (Table 1), as both wet and spring season flows increased more than that the base flow was reduced.

The trend of the sediment load at basin outlet was examined using the same statistical tests, and the results indicated an increasing trend of sediment transport between 1980 and 2009, see last row of Table 1. The Man-Kendall test shows that the sediment load was significantly increased at 5% significance level. Similarly, the Pettitt test revealed an increasing sediment transport from 91*106 in 1980-1992 to 147 *106 ton/year in 1993-2009. To further confirm the result, silt concentration in the 1970's and 2000's were also inspected and the comparison indicated that the concentration has increased significantly, implies increasing of sediment load.

Table 1: Man-Kendall/Statistical summary of seasonal and annual flow and sediment load at El Diem Station for 1970 to 2005

Season Kendall's tau S p-value trend

Wet (Jun to Sep) 0.34 237 0.003 Significantly increasing

Dry (Oct to Feb) -0.37 -259 0.001 Significantly decreasing Short rain (Mar to May) 0.41 285 0.001 Significantly increasing Flow

Annual 0.25 175 0.028 Significantly increasing

Sediment load Annual 0.7 200 < 0.0001 Significantly increasing

3.4. SWAT model results

The most sensitive parameters with their calibrated optimum values are presented in Table 3.

The 17 most sensitive parameters were used for model calibration of stream flow and sediment load also given in

Table 2. Initial parameter estimates were taken from default lower and upper bound values of the SWAT’s model database and from previous studies in the basin. Calibration parameters were derived for two periods, 1971 to 1973, and 2000 to 2002.

Parameters such as SCS curve number (CN2), Base

Figure 2: Pettitt homogeneity test of seasonal and annual flows: (a) wet season flow, (b) short rainy season flow, (c) dry season flow and (d) annual flow


272

NCSTI-2012

flow alpha factor (ALPHA_BF), Soil evaporation

compensation factor (ESCO), Threshold water depth in the shallow aquifer (GWQMN), Channel effective hydraulic conductivity (CH_K2), Ground water ''revap'' coefficient (GW_REVAP), surface runoff lag time (SURLAG), deep aquifer percolation fraction (RCHRG_DP), available water capacity (SOL_AWC), soil depth (SOL_Z), and Ground water delay

(GW_DELAY) were the most sensitive parameter for flow predictions from the basin. Parameters including linear re-entrainment parameter for channel sediment routing (SPCON), USLE support practice (USLE_P), Channel effective hydraulic conductivity (CH_K2) were among the most sensitive parameters for sediment prediction only. SWAT uses those parameters to estimate the amount of flow and sediment yield from the catchment.

Figure 3: Calibration results of daily discharge values for the two periods: (a) 1971-1973, (b) 2000-2002

Table2: SWAT sensitive model parameters and their (final) calibrated values for 1971 to 1973 and 2000 to 2002 runs

Best Fitted values Parameter Description 1971 to

1973 2000 to

2002 Change (%)

CN2* ALPHA_BF**

ESCO** CH_K2**

GWQMN***

GW_REVAP** SURLAG**

RCHRG_DP** SOL_AWC* CANMX**

GW_DELAY** SPCON*

USLE_P** SPEXP**

HRU_SLP** SLSUBBSN*

SOL_Z*

curve number Base flow alpha factor

Soil evaporation compensation factor Channel Effective hydraulic conductivity

Thresh hold water depth in shallow aquifer Ground water "revap" coefficient

Surface runoff lag time Deep aquifer percolation factor Available water capacity of soil

Maximum canopy storage Ground water delay

Linear re-entrainment parameter for channel sediment routing

USLE support practice Exponentiation re-entrainment parameter

for channel sediment routing

Average slope steepness Average slope length

Soil depth

-0.17 0.21 0.72

16.32 1002.25

0.12 6.35 0.56 0.62 4.18

78.16 0.01

0.58 1.2

0.08 -0.35 0.22

-0.03 0.15 0.43 17.54

823.54

0.17 4.68 0.38 0.48 3.21 72.96 0.01

0.83 1.32

0.08 -0.27 0.21

14 -28.6 -67.3

7.5 -21.8

41.7

-26.3 -32.1 -22.6 -23.2 -6.7

0

43.1 10

0 8 1

* Relative change in the parameters where value from SWAT database is multiplied by 1 plus a given range, ** Replace the initial parameter by the given value, *** Adding the given value to initial parameter value


273

NCSTI-2012

Figure 3 shows the calibration results of daily discharge hydrographs for two simulations periods. The model captured the daily runoff hydrographs both for low and high flows. The obtained models performances for calibration were E= 0.8, and R2= 0.89 for 1971-1973 and E=0.84 and R2=0.92 for 2000-2002 simulation periods. For the validation period E=0.78 and R2 =0.84 for 1974-1976 period, and E=0.82 and R2=0.88 for 2003-2005. The performance of both models found to be satisfactory and comparable to previous studies in the basin. For instance, [11] reported E=0.87 and R2=0.92 for calibration of daily flow, while [42] showed E=0.68 for daily calibration at El Diem station.

The last column of Table 2 gives the percentage change of best fitted model parameters for period 2000 to 2002 relative to 1971 to 1973, i.e., before and after landuse changes. As can be seen, higher percentage change was obtained for some of the parameters, which indicates significant changes of the catchment response behaviour he Surface runoff response parameters (e.g. CN2, ESCO and SOL_AWC) showed a higher change. An increase of CN2 value indicates that a higher amount of surface runoff was generated in the 2000s compared to the 1970s. The decrease of ESCO explains that more water was extracted from lower soils to meet evaporative demand, indicating a significant reduction of soil water. The available soil water capacity (SOL_AWC) was also significantly decreased for the past 35 years suggesting a shallower soil profile. Lower SOL_AWC implies the retention capacity of soil is reduced and as a result surface runoff generation is increased.

Similarly, there was a clear change of subsurface response parameters (ALPHA_BF); Threshold water depth in the shallow aquifer (GWQMN), Ground water ''revap'' coefficient (GW_REVAP), deep aquifer percolation fraction (RCHRG_DP), Ground water delay (GW_DELAY) between the analysis periods. All changes indicated a faster response towards surface runoff generation. ALPHA_BF is a direct index of ground water flow response to ground water recharge, and its lower value implies a smaller contribution of base flow to the river discharge. The reduction of the GWQMN parameter means a decrease of the threshold value, implying faster surface flow response. The deep aquifer percolation coefficient (RCHRG_DP) that controls the movement of water to the lowest depth of the soil profile, showed a higher reduction. This indicates that less water percolates to the deep aquifer as compared to 1970s. Conversely, the Ground water "revap" coefficient that controls movement of water between the soil profile and shallow aquifer was increased. This may indicate that water from shallow aquifer moves back to the overlying dry material

(unsaturated zone) during dry period. As water is evaporated from capillary fringes, it is substituted by water from underlying aquifer.

Similar to the flow simulation, the SWAT model was used to simulate the sediment load from the basin. The calibration results for the daily sediment yields at El Diem station are displayed in Fig.4.

As can be seen from the Fig. 4a and 4b, the magnitude and temporal variation of simulated sediment yield closely matches observations. The performance of daily sediment load simulations results showed that E=0.76 and R2 =0.78 for the calibration period (1971 to 1973) and E=0.73 and R2=0.75 for the validation period (1974 to 1976). Similarly, the performance of the second model shows acceptable model performance of E=0.78 and R2 =0.75 during calibration (2000 to 2002) and E=0.8 and R2=0.72 during validation (2003 to 2005). These results are comparable with model performances of recent studies by [5], [11], who obtained E=0.74 and E=0.88, respectively for sediment simulation using SWAT model.

Next, model results were checked using annual water and sediment balance. The validation period annual water balance components were presented as shown in Table 4. In which, ET=Evapotranspiration; Qsurf=Surface runoff; Qlat= Lateral flow; GWQ= Ground water flow; Water yield is the total water yield (Qsurf+QLat+GWQ-Transmission Losses); SW=soil water; PERC=percolation (ground water recharge).

The average annual water balance of the basin shows that the surface runoff (Qsurf) contribution to the total river discharge has increased by 75%, while the subsurface flow (Qlat) and ground water (GWQ) flow has decreased by 25% and 50%, respectively (Table 3). Even with negligible change of rainfall between the two periods (1.3%), the total water yield at the outlet has increased by 25%. This clearly depicts a modification of catchment response and thus a change of physical characteristics between 1970s and 2000s. The simulations results of the three major components of the water balance (Rainfall, ET and water yield) resemble observed values of the basin at basin outlet, EI Deim. It seems that the model has unrealistically over-predicted the deep aquifer recharge PERC, 22% compared to total yield of 18.7%. SWAT considers PERC as a loss from the system, and doesn’t contribute to the total yield from the basin [4], [34]. This may not be realistic and literature shows similar difficulties of estimating ground water flow and deep ground water recharge using SWAT model [39]. However, the uncertainty of the model on deep water recharge may have negligible effect in the conclusion of this study, assuming errors in both models can offset each other.


274

NCSTI-2012The annual average suspended sediment load from the

basin were 4.46 t/ha and 6.8 t/ha during validation periods of 1974-1976 and 2003-2005, respectively. These results show that the total sediment yield from the basin has

increased by 53% in the past 28 years. This could be due to high sediment production and soil erosion rates from the basin.

Figure 4: Calibration result of Upper Blue Nile daily Sediment load at el Diem station for:

(a) 1971-1973, (b) 2000-2002

Table 3: Annual water balance of the basin for 1974-1976 and 2003 -2005 validation periods

Simulation Period

units Rain Fall ET Qsurf

Qlat

GWQ

Water yield

SW

PERC

TLosses

mm/year 1426 758 145 97 24 267 77 315 9 1974-1976 % age 100 53 10.3 6.8 2.4 18.7 5.4 22.1 0.64

mm/year 1445 774 254 73 12 332 97 220 12 2003-2005 % age 100 53.6 17.6 5.05 0.83 23.7 6.72 15.2 0.83

Therefore, the results of the SWAT simulations

confirmed the earlier conclusions derived from the statistical tests, in that both runoff and sediment load from Upper Blue Nile basin has shown an increasing trend during the last 28 years. Moreover, the comparisons of the SWAT model parameters showed that the likely reasons for changes are attributed to increased surface runoff compared to ground water flow component.

The observed increasing trends of surface runoff and sediment load from Upper Blue Nile basin could be attributed to landuse change over large areas of the basin. Specifically, changes of natural vegetation cover into agricultural crop land.

5. Conclusio

nThe objectives of this study were to understand the

long-term variations of hydrology and sediment fluxes of the Upper Blue Nile Basin using statistical models, and to

verify the results using a physically-based hydrological model. The Man-Kendall and Pettitt tests showed that no change of annual rainfall over the Upper Blue basin between 1970s and 2000s. However, both tests showed a significant increasing trend of stream flow during the long wet season of from June to September and the short season from March to May, and a decreasing trend of dry season flow from October to February. The annual stream flow and sediment load from the basin were also increased significantly for the past 39 years (1971-2009). The Pettit test showed that most of these changes appeared in the early 1990s, and that a significant abrupt downward change of dry season stream flow occurred around 1979.

The SWAT daily model was used to predict runoff and sediment load at the basin outlet (El Deim station), which is at the Ethiopia-Sudan border. The null hypothesis was that optimal (calibrated) model parameters will remain unchanged. The modelling results showed that model parameters, specifically surface runoff and groundwater


275

NCSTI-2012

parameters, were significantly changed between 1970s and 2000s simulation periods. It is most probable that these changes are attributed to modification of catchment physical characteristics.

The results from the two methods of analysis based on statistical tests, and a physically based hydrological model, confirmed increasing trends of runoff and sediment load from the Upper Blue Nile. These increasing trends were attributed to change of natural land cover over large areas into agricultural crops, which modified runoff generation processes. These findings are very important for basin wide water resources management of the Blue Nile basin by providing insights on catchment behaviour, which is a big challenge for decision makers of both watershed management and downstream irrigation managers. Moreover, it can give a better understanding of embedded interdependencies between upstream and downstream areas.

Reference 1 Abbaspour, K., Yang, J., Maximov, I., Siber, R.,

Bogner, K., Mieleitner, J., Zobrist, J., Srinivasan, R., 2007. Modelling hydrology and water quality in the pre-alpine/alpine Thur watershed using SWAT, J. Hydrol., 333,413-430.

2 Ahmed, S., 2010. Analysis of the impact of landuse change and Climate change on the flows in the Blue Nile River Using SWAT, Katholieke Universiteit Leuven, Belgium, Leuven.

3 Amsalu, A. L., Stroosnijder and de Graaf, J., 2007. Long-term dynamics in land resource use and the driving forces in the Beressa watershed, highlands of Ethiopia. Journal of Environmental Management 83 (4): 448-459.

4 Arnold,J., Srinivasan, R., Muttiah, R., Williams. J.,1998. Large area hydrologic modeling and assessment part i: model development1, JAWRA , 34, 73-89.

5 Betrie, G., Mohamed, Y., van Griensven, A., Srinivasan, R., Mynett, A.,2011. Sediment management modelling in Blue Nile Basin using SWAT model, Hydro. Earth Syst. Sci., 7, 5497-5524.

6 Bewket, W., 2004. Towards integrated watershed management in highland Ethiopia: the Chemoga watershed case study, Tropical resources management papers, 44, 2003.

7 Burn, D., Cunderlik, J., Pietroniro, A: Hydrological trends and variability in the Liard River basin, Hydrol. Sci. J., 49,53-68.

8 Conway, D., 2000. The climate and hydrology of the Upper Blue Nile River, Geog. J., 166, 49-62.

9 Conway, D., Hulme, M., 1993. Recent fluctuations in precipitation and runoff over the Nile sub-basins and

their impact on main Nile discharge, Climatic Change, 25,127-151.

10 Conway,D., Mould, C., Bewket, W., 2004. Over one century of rainfall and temperature observations in Addis Ababa, Ethiopia, Int. J. Climatology, 24,77-91.

11 Easton, Z., Fuka, D., White, E., Collick, A., Asharge, B., McCartney, M., Awulachew, S., Ahmed A., Steenhuis, T., 2010. A multibasin SWAT model analysis of runoff and sedimentation in the Blue Nile, Ethiopia, Hydrol.Earth. Syst. Sci., 7, 3837-3878.

12 Elshamy M., Wheater, H., 2009. Performance assessment of a GCM land surface scheme using a fine-scale calibrated hydrological model: an evaluation of MOSES for the Nile Basin. Hydrol. Processes 23:1548-1564.

13 FAO, 1995. Digital Soil Map of the World and Derived Soil Properties, Food and Agriculture Organization of the United Nations , Rome, Italy.

14 Garzanti, E., And, S., Vezzoli, G., Ali Abdel Megid, A., El Kammar, A., 2006. Petrology of Nile River sands (Ethiopia and Sudan): sediment budgets and erosion patterns, Earth and Planetary Science Letters, 252,327-341.

15 Gassman,P.W., Reyes, M.R., Green,C.H., Arnold,J.G., 2007. The soil and water assessment tool: historical development, applications, and future research directions, American society of Agricultural and biological Engineers, 50, 1211-1250.

16 Gebrehiwot, S., Taye, A., Bishop, K., 2010. Forest Cover and Stream Flow in a Headwater of the Blue Nile: Complementing Observational Data Analysis with Community Perception, AMBIO: A J. Human Environment, 39, 284-294.

17 Global Land Cover Characterization (GLCC), 2010. http://edcsns17.cr.usgs.gov/glcc/glcc.html, last access: 12 December 2010.

18 Gordon, N., 2004. Stream hydrology: an introduction for ecologists, John Wiley & Sons Inc.

19 Green, W., Ampt, G., 1911. Studies on soil physics, the flow of air and water through soils, J. Agric. Sci., 4, 1-24.

20 Hargreaves, G., Hargreaves, G., and Riley, J., 1985. Agricultural benefits for Senegal River basin, J. Irrig. Drain. E-ASCE, 111, 113–124.

21 Herweg, K., Stillhardt., 1999. The variability of soil erosion in the highlands of Ethiopia and Eritrea, Research report, 42.

22 Huang, C.,Wylie, B., Yang, L., Homer, C., and Zylstra, G., 2002. Derivation of a Tassled Cap transformation based on Landsat and at-satellite reectance, Int. J. Remote Sens., 23, 1741–1748.


276

NCSTI-2012

23 Hurni, H., 1993a. Land degradation, famine, and land resource scenarios in Ethiopia:World siol erosiion and Conservation, edited by: Pimentel, D, Cambridge University Press, Cambridge, UK.

24 Hurni, H., 1993b. Land degradation, famine, and land resource scenarios in Ethiopia Cambridge University Press, Cambridge, UK.

25 Hurni, H., Tato, K., Zeleke, G., 2005. The implications of changes in population, landuse, and land management for surface runoff in the upper Nile Basin Area of Ethiopia, Mountain research and development, 25, 147-154.

26 Kebede, E.,2009. Hydrological responses to land cover changes in Gilgel Abbay catchment of Ethiopia, ITC, The Netherlands.

27 Kendall, M., 1975. Rank Correlation Methods, Charles Griffin, London.

28 Legesse, D., Vallet-Coulomb, C., Gasse, F., 2003. Hydrological response of a catchment to climate and landuse changes in Tropical Africa: Case study South Central Ethiopia. J. Hydrol., 275, 67-85.

29 Lu, X., 2005. Spatial variability and temporal change of water discharge and sediment flux in the Lower Jinsha tributary: impact of environmental changes, River Research and Applications, 21, 229-243.

30 Monteith, J. L., 1965. Evaporation and environment, Symp. Soc. Exp. Biol., 19 205–234.

31 Moriasi D., Arnold J., Van Liew M., Bingner R., Harmel R., Veith T., 2007. Model evaluation guidelines for systematic quantification of accuracy in watershed simulations. American society of Agricultural and biological Engineers, 50(3): 885−90.

32 Nash J, Sutcliffe, J., 1970. River flow forecasting through conceptual models part I--A discussion of principles. Journal of Hydrology 10: 282-290

33 NBCBN, 2005. Survey of litrature and data inventery in watershed erosion and sediment transport, Nile basin Capacity Buliding Network, Cairo.

34 Neitsch, S.L., Arnold, J.G., Kiniky, J.R., Williams, J.R., 2005. Soil and Water Assesement Tool Technical Docummemntation: Version 2005, United States Department of Agriculture, Agricultural Research Service.

35 Pettitt, A., 1979. A non-parametric approach to the change-point problem, Applied Statistics, 28, 126-135.

36 Priestley, C. and Taylor, R., 1972. On the assessment of surface heat flux and evaporation using large-scale parameters, Mon.Weather Rev., 100, 81–92.

37 Setegn, S., 2010. Modelling Hydrological and Hydrodynamic Processes in Lake Tana Basin, Ethiopia, Ph.D. thesis, TRITA-LWR, KTH., Sweden.

38 Setegn, S., Srinivasan, R., Dargahi, B., 2008. Hydrological modelling in the Lake Tana Basin, Ethiopia using SWAT model, The Open Hydrol. J., 2, 49-62.

39 Steenhuis, T.S., Collick, A.S., Easton, Z.M., Leggesse, E.S., Bayabil, H.K., White, E.D., Awulachew, S.B., Adgo E., Ahmed, A.A., 2009. Predicting discharge and sediment for the Abay (Blue Nile) with a simple model, Hydrol. Processes, 23, 3728-3737.

40 Sutcliffe, J., Parks, Y.,1999. The hydrology of the Nile, IHAS special publication no.5, Int. Hydrol. Sci. Wallingford, UK.

41 Tesemma, Z., 2009. Long term Hydrological trends in the Nile Basin, Integrated Watershed managemnt and hydrology, Cornell University, Bahirdar, Ethiopia.

42 Tesemma, Z.K., Mohamed, Y.A., Steenhuis, T.S., 2010. Trends in rainfall and runoff in the Blue Nile Basin: 1964–2003, Hydrol. Processes, DOI: 10.1002/hyp.7893.

43 Tibebe, D., Bewket, W., 2010. Surface runoff and soil erosion estimation using the SWAT model in the Keleta Watershed, Ethiopia. Land Degradation & Development, DOI: 10.1002/ldr.1034.

44 USDA, 1972. USDA Soil Conservation Service, National Engineering Handbook, Section 4.

45 Van Griensven A., Meixner, T., Grunwald, S., Bishop, T., Diluzio, M., Srinivasan, R., 2006. A global sensitivity analysis tool for the parameters of multi-variable catchment models, J. Hydrol. 324. 10-23.

46 Von Storch, H., 1995. Misuses of statistical analysis in climate research. In Analysis of Climate Variability: Applications of Statistical Techniques, eds. by H.V. Storch and A. Navarra, Springer–Verlag Berlin, 11–26

47 Wischmeier, W.H., Smith, D.D., 1978. Predicting rainfall erosion losses: A guide to conservation planning, Agricultural Handbook No. 537, US Department of Agriculture, Washington, DC, 1-58.

48 Yue S., Pilon P., Phinney B., and Cavadias, G., 2002. The influence of autocorrelation on the ability to detect trend in hydrological series. Hydrol Process, 16, 1807–1829.

49 Zeleke, G., Hurni, H., 2001. Implications of landuse and land cover dynamics for mountain resource degradation in the northwestern Ethiopian highlands, Mountain research and development, 21, 184-191.

50 Zhang, S., Lu, X.X., Higgitt, D.L., Chen, C.-T.A., Han, J., Sun, H., 2008. Recent changes of water discharge and sediment load in the Zhujiang (Pearl River) Basin, China. Globa and Planetary Change, 60, 365-380.

Documents

Proceeding for Conference(Bahir Dar University).pdf