ETHIOPIAN INSTITUTE OF WATER RESOURCE

ENGINEERING AND MANAGMENT

Research
On
RAINFALL-RUN-OFF SIMULATION AND MODELING
(THE CASE OF DECHATU WATERSHED)

Submitted in Partial Fulfilments of the Requirement for the Award of

Master of Science (MSc) in Water Resources Engineering and
Management

By

Kedir Hassen

January, 2020Addis Ababa

i
 APPROVAL SHEET

ADDIS ABABA UNIVERSITY

RAINFALL-RUNOFF SIMULATION AND MODELING

(THE CASE OF DECHATU WATERSHED)

Submitted by:

KEDIR HASSEN MOHAMMED ......................... .................

Signature Date

Approved by:

1. Dr. Tena Alamirew................................. ......................

(Major Advisor) Signature Date

2. Dr. Dangenet Sultan ................................. ......................

(Co- Advisor) SignatureDate

3. School Chairman .............................. ......................

Signature Date

4. Dean of School............................ .......................

Signature Date

ii
ACKNOWLEDGEMENTS

First and foremost, I want to thank the Almighty God for giving me the chance, strength
and courage to continue my study.

Then, I would like to express my sincere appreciation and gratitude to my advisor Dr.Tena
Alamirewand for his guidance, encouragement and enthusiastic support and many
kindnesses extended during the seemingly interminable period of this research. My sincere
appreciation also goes to all my friends, instructors and class mates for sharing their
experience, materials and valuable cooperation throughout my study.

Finally, I would like to express my thankfulness to all my families and friends for their
encouragements and support. KEDIR HASSEN.

iii
ABSTRACT

This paper presents an application of surface rainfall run-off simulation using the
Hydrologic Modelling System (HEC-HMS). A case study was carried out for the Dechatu
watershed, a Rainfall is classified as “bi-modal” with the first significant rains from March
to May and the second rains from is July to Septembertypical semi-arid and sub-humid
geo-climatic region in eastern Ethiopia Awash River basin. The modeling schemes using
mechanism of models provided by HEC-HMS for runoff volume. The modeling results
were compared with historical observation data.

Several River Basins (Watershed) across the world have been simulated using hydrological
models to understand hydrological processes and the availability of water resource. Some
of these basins are ungauged. In this study in order to evaluate the hydrological process of
Dechatu Watershed for sustainable management; HEC-HMS 4.2 hydrologic model (with
CN) is used to simulate its runoff (Watershed Modelling) was used as an interface to
delineate the watershed and generate some input (basins parameters). The CNparameters
are computed in HEC-GeoHMS using land use and soil type data.Using the Chi-Squared
test ranking, Dengego rainfall data fits Lognormal 3 parameter distribution.

Annual Maximum Daily Rainfall (mm) for selected return periods50years were used for
the meteorological inputs. The results showed a total volume of runoff 150.20MM3 during
the fifty years of the simulation. The peak discharge was found to be 2,528.7m 3 /s and the
model evaluation has showed, indicating that the results of the simulation are satisfactory.

iv
 List of Figure
FIGURE 1 LOCATION AND MAP OF DECHATU WATERSHED DIRE DAWA ADMINISTRATE ........................................ 6

FIGURE 2. SLOPE MAP OF DECHATU CATCHMENT .................................................................................................... 7

FIGURE 3. REPRESENTATION OF RAINFALL RUNOFF PROCESS...................................................................................... 21

FIGURE 4 MODELING APPROACH RAINFALL RUNOFF MODELING ............................................................................... 30

FIGURE 5. THE HYDROLOGICAL MODEL IMPLEMENTED IN HEC-HMS. ....................................................................... 34

FIGURE 6 MEAN MONTHLY RAINFALL (MM) DENGEGO STATION RAIN FALL............................................................... 34

FIGURE 7. DEM OF DECHATU RIVER WATERSHED AREA ............................................................................................ 36

FIGURE 8. SOIL MAP OF DECHATU RIVER WATERSHED .............................................................................................. 37

FIGURE 9. LAND USE\LAND COVER MAP 1985. (LANDSAT MSS 1985) ....................................................................... 40

FIGURE 10 LAND USE\LAND COVER MARCH (LANDSAT TM 2000) ............................................................................. 41

FIGURE 11 LAND USE\LAND COVER (LANDSAT ETM + 2014) ................................................................................... 42

FIGURE 12 FREQUENCY ANNUAL DAILY MAXIMUM RF VALUES USING GUMBEL (EV1).............................................. 45

FIGURE 13. GRAIN SIZE FREQUENCY DISTRIBUTION AND MATERIAL AT THE FLOW GAUGE REACH .............................. 48

FIGURE 14 DECHATU RIVER FLOW ............................................................................................................................. 49

FIGURE 15. LOGNORMAL, GAMBEL, LOG PEARSON III ................................................................................................ 58

FIGURE 16 GRAPH OF DECHATU PEAK FLOOD OF 6 HOUR ........................................................................................................ 59

FIGURE 17 PEAK INFLOW & PEAK DISCHARGE........................................................................................................................ 60

FIGURE 18. TIME SERIOUS RESULT ..................................................................................................................................... 60

FIGURE 19. DISCHARGE VS. RETURN PERIOD.......................................................................................................... 64

LIST OF TABLESPage
FIGURE 1 LOCATION AND MAP OF DECHATU WATERSHED DIRE DAWA ADMINISTRATE ........................................ 6

FIGURE 2. SLOPE MAP OF DECHATU CATCHMENT .................................................................................................... 7

FIGURE 3. REPRESENTATION OF RAINFALL RUNOFF PROCESS...................................................................................... 21

FIGURE 4 MODELING APPROACH RAINFALL RUNOFF MODELING ............................................................................... 30

FIGURE 5. THE HYDROLOGICAL MODEL IMPLEMENTED IN HEC-HMS. ....................................................................... 34

FIGURE 6 MEAN MONTHLY RAINFALL (MM) DENGEGO STATION RAIN FALL............................................................... 34

FIGURE 7. DEM OF DECHATU RIVER WATERSHED AREA ............................................................................................ 36

FIGURE 8. SOIL MAP OF DECHATU RIVER WATERSHED .............................................................................................. 37

v
FIGURE 9. LAND USE\LAND COVER MAP 1985. (LANDSAT MSS 1985) ....................................................................... 40

FIGURE 10 LAND USE\LAND COVER MARCH (LANDSAT TM 2000) ............................................................................. 41

FIGURE 11 LAND USE\LAND COVER (LANDSAT ETM + 2014) ................................................................................... 42

FIGURE 12 FREQUENCY ANNUAL DAILY MAXIMUM RF VALUES USING GUMBEL (EV1).............................................. 45

FIGURE 13. GRAIN SIZE FREQUENCY DISTRIBUTION AND MATERIAL AT THE FLOW GAUGE REACH .............................. 48

FIGURE 14 DECHATU RIVER FLOW ............................................................................................................................. 49

FIGURE 15. LOGNORMAL, GAMBEL, LOG PEARSON III ................................................................................................ 58

FIGURE 16 GRAPH OF DECHATU PEAK FLOOD OF 6 HOUR ........................................................................................................ 59

FIGURE 17 PEAK INFLOW & PEAK DISCHARGE........................................................................................................................ 60

FIGURE 18. TIME SERIOUS RESULT ..................................................................................................................................... 60

FIGURE 19. DISCHARGE VS. RETURN PERIOD ............................................................................................................. 64

Table of Contents Page

CHAPTER ONE.................................................................................................................................. 1
1. INTRODUCTION ........................................................................................................................... 1
1.1 General Background ................................................................................................................... 1
1.2 Scope of the problem ................................................................................................................. 2
1.2. Statement of the Problem ......................................................................................................... 3
1.3. Objectives................................................................................................................................. 4
1.3.1. Specific objectives of the studies are....................................................................................... 4
1.4. Research questions ................................................................................................................. 5
1.5. Significance of the Study ........................................................................................................... 5
CHAPTER TWO................................................................................................................................. 6
2. DESCRIPTION OF THE STUDY AREA ............................................................................................... 6
2.1. Location.................................................................................................................................... 6
2.2. Physiography ............................................................................................................................ 7
2.3. Climate and rainfall ................................................................................................................... 8
2.4. Regional hydrology ................................................................................................................... 9
2.5. Geology .................................................................................................................................. 11
2.6 Selection of hydrological model ................................................................................................ 11
2.7. Reason for selection of (HEC-HMS) model ................................................................................ 12
2.8. Meteorological Data Analysis .................................................................................................. 13
2.9. Filling Missed Meteorological data’s ........................................................................................ 13
2.10. Data Quality Assessment ....................................................................................................... 14
2.11. Test for Absence of trend ...................................................................................................... 14
2.12. Test for stationery ................................................................................................................. 15
2.13 Test for homogeneity ............................................................................................................. 15
2.14. Test for Consistency .............................................................................................................. 16
2.15. Data Available....................................................................................................................... 16
CHAPTER THREE ............................................................................................................................ 18

vi
3.0 LITERATURE REVIEW ................................................................................................................ 18
3.1. General................................................................................................................................... 18
3.1.1 Model and software description ............................................................................................ 18
3.1.2 ArcGIS: HEC-GeoHMS ............................................................................................................ 18
3.1.3. Hydrological Modeling ......................................................................................................... 19
3.1.4 Rainfall Runoff Model: HEC-HMS ........................................................................................... 20
3.1.5. Basin Model ......................................................................................................................... 22
2.1.6 Flood frequency analysis ....................................................................................................... 24
3.1.7. Previous studies of the area ................................................................................................. 25
4.5. Runoff Estimation ................................................................................................................... 28
4.0 METHODOLOGY ....................................................................................................................... 30
4.1 Hydrological Model Development ............................................................................................ 30
4.2. Terrain Pre-processing ............................................................................................................ 31
4.3. Developing hydrological parameters ..................................................................................... 31
4.4 Developing HEC-HMS model files:............................................................................................. 33
4.6. Model Inputs .......................................................................................................................... 35
4.7. Digital Elevation Model (DEM) ................................................................................................. 35
4.8. Soil Data and Maps (Hydrologic Soil Groups) ............................................................................ 36
4.9. GIS based land use/cover change assessment .......................................................................... 39
4.10 Rainfall Frequency.................................................................................................................. 44
4.11 Rainfall Frequency Analysis .................................................................................................... 44
4.12 Rainfall Intensity .................................................................................................................... 44
4.13 Main rivers and run-off characteristics .................................................................................... 45
4.17. Peak Flow Analysis ................................................................................................................ 50
4.18. Extreme value I distribution .................................................................................................. 50
4.19. EVI–Gumbel distribution by the method of moments ............................................................. 53
4.20 Log Pearson type III distribution............................................................................................ 54
5.0 Results and Discussion ............................................................................................................. 57
5.1. The Runoff-Volume Models ..................................................................................................... 57
5.2. Flood Flows............................................................................................................................. 63
5.3. Land use-land cover and changes in Dechatu watershed......................................................... 65
6. Discussion.................................................................................................................................. 67
7.0 Conclusion ............................................................................................................................... 69
Reference ...................................................................................................................................... 71

vii
ABBREVIATIONS
Dire Dawa Administrative
DDAC
Council
DD Dire Dawa
DEM Digital Elevation Model
E East
EQN Equation
ERA Ethiopian Roads Authority
Federal Urban Planning
FUPI
Institute
Geo Geographical
Geographical Information
GIS
Systems
GUI Graphical User Interface
Hydrologic Engineering
HEC
Centre
HMS Hydrologic Modelling System
Km Kilometer
M Million
m Meter
masl meter above sea level
mm millimeter
MoWR Ministry of Water Resources
N North
PET Potential Evapo-Transpiration
RAS River Analysis System
SCS Soil Conservation Service
Sec Second
U.S. United States
UH Unit Hydrograph
Yr year

viii
CHAPTER ONE

1. INTRODUCTION

1.1 General Background

Water is the most vital resource to support all forms of life on earth. It will remain essential
for mankind survival and the future development of the world. Water is not evenly
distributed over the world by season or location, i.e. global fresh water distribution is
neither uniform in space nor in time. Some parts of the world are prone to drought making
water scarce and precious commodity, while in other parts of the world it appears in raging
torrents causing floods and loss of life and property.

Ethiopia, often referred to as the water tower of East Africa, is dominated by mountainous
topography, and the rainfall-runoff processes on the mountainous slopes are the source of
the surface water for much of Ethiopia (Derib, 2009), and thus, understanding the rainfall-
runoff processes is critical to controlling erosion and enhancing agricultural productivity.
The majority of the sedimentation of rivers in the basin occurs during the early period of
the rainy season and peaks of sediment are consistently measured before peaks of discharge
for a given rainy season (Steenhuis, 2009). In fact, while the primary cause of flooding is
abnormally high rainfall, there are many human-induced contributory causes such as land
degradation, deforestation of catchment areas, increased population density along
riverbanks, poor land use planning, zoning and control of flood plain development,
inadequate drainage, particularly in cities, and inadequate management of discharges from
river reservoirs (Mulugeta, 2007). In Ethiopia, a research in Awash River basin by (Terefe,
2006) indicates that human factor plays a crucial role in causing frequent flood disasters in
upper, middle, and lower Awash. Moreover, projections of flood changes at river basin
scale are scarce in the scientific with the alteration of natural environment due to human
interventions together with the effects of global climate change, recent years have seen
floods to occur more frequently and unpredictably across the globe. Urbanization and
changing of demographic features within the river flood plain has led to increased exposure
of communities to flood hazard. A model simulation through calibration and validation
with measured values (Sultan, 2017). In the absence of extensive field studies and runoff
measurements, models have been used to estimate site specific information (Hadadin
,2013)Several approaches can be used to estimate runoff, from simple empirical rainfall-

1
runoff models to conceptual and highly parameterized process-based models (Haregeweyn,
2016). The common method for predicting event-based surface runoff volume from small
watersheds is the Soil Conservation Service Curve Number (SCS-CN) method (SCS 1972).

A rainfall runoff modelling (hydrological) anticipate evaluating the runoff from

precipitation in a catchment and hydraulic modelling aims to evaluate magnitude of floods
and the area inundated by them. Combination of both will result in runoff simulation and
flood inundation levels. The flood at times of unusually high rainy days over top the
normal flood ways and create a lot of calamity to the residents of Dire Dawa city. Several
flash floods events occurred pass through the city every year, and caused a lot of distraction
on properties and people. a major flood swept through Dire Dawa city sequence of floods
that occurred in 1981, 1994, 2004, and 2005 (Demessie, 2007;Alemu 2009; DDAEPA
2011) that caused significant fatalities and damages to property, on August 6, 2006, the city
of Dire Dawa experienced one of the largest and the most devastating flood ever. The area
inundated during the 2006 flood was about 1 km2 and 86 % of this area was covered by 1–
2 m of water (Alemu, 2009). Although, on the night of 5/6 August 2006, resulting in over
300 fatalities and significant damage to the flood defenses, public infrastructure, housing
and livelihoods. Often causing loss of life and damage to property and infrastructure
presently in the existing situation like this, it is important to consider major factors that
contributed most in the past flood hazard calamities. It has of paramount importance to take
notice of these factors to arrive at wise and comprehensive solution towards mitigating the
challenge (that is flooding) which is erratic and unpredictable. Such regions are exposed to
the hazard that the available fresh water resources fail to meet the water demand in the
domestic, agricultural and industrial sectors.

1.2 Scope of the problem

In Africa floods are among the most devastating natural hazards, whereas flash floods are
among the greatest hazards arising from tropical cyclones and severe storms. Floods and
flash floods cause loss of life, damage to property, and promote the spread of diseases such
as malaria, dengue fever, and cholera. From 1,900 to 2006, floods in Africa killed
nearly 20 000 people and affected nearly 40 million more, and caused damage estimated at
about US$4 billion (ICSUAfrica, 2007). Water related disasters increased more than
four folds in Africa (Adikari, 2008). According to World Bank (2003), in most developing

2
 countries flood disasters still claim tens of thousands of lives each year and destroy
livelihoods in an instant

Ethiopia’s topographic characteristics has made the country pretty vulnerable to floods and
resulting destruction and loss to life, economic, livelihoods, infrastructure, services and
health system (DPPC, 2007).

Several factors could be mention as causes of flooding by different writers. Deforestation

can impact hydrological processes, leading to localized declines in rainfall, and more
rapid runoff of precipitation, causing flooding and soil erosion, a common phenomenon in
most parts of Ethiopia (Dagnachew, 2003). On the other hand, the high infiltration rates
under natural forests serve to reduce surface runoff and flood response. Certain types of
plantation forests may also serve to increase infiltration rates through providing preferential
flow pathways down both live and dead root channels. From the theoretical considerations
it would be expected that interception of rainfall by forests would reduce floods by
removing the proportion of the storm rainfall and by allowing the build-up of the soil
moisture deficits (Calder, 1999).

According to (Dagnachewet, 2003), Land-use change due to the expansion of urban areas
also affects the ground infiltration rate which in turn gives the way flooding to occur.
Land-cover change has one of the causes of flooding phenomenon of the awash basin,
which resulted in millions worth of resources lost nearly every main rainy season.
Lowlevel vegetative cover could also affect infiltration and could lead to reduced
groundwater levels and the base flow of streams. It is obvious that land-cover can affect
both thedegree of infiltration and increases runoff following rainfall events (Dagnachewet,
2003)

1.2. Statement of the Problem

High flood, which is normally due to the intensive rainfall in the up lands of the watershed,
sparse vegetation cover, steep slopes and low infiltration capacity of the ground surface,
which makes the watershed susceptible to soil erosion, sedimentation and run-off
contribution to the dawn stream area. (Wale and Mengistu, 2009). In developing countries,
studies on ephemeral streams flash floods are uncommon and very few data are available to
design appropriate risk mitigation countermeasures and warning systems. Various studies
have indicated that it is not unusual to have high flow in the upper catchments and low or

3
 zero flow at the downstream gauges (Renard et al. 1966; Cordery et al. 1983; Walters
1990; Al-Qurashi et al. 2008). Hughes and Sami 1992 studied two events in Cape Province,
South Africa, and found that 75 and 22% of the flow volumes were lost to the alluvium and
sand channel beds, respectively Ethiopia’s topographic and climatic characteristics have
made the country vulnerable to high floods that resulted in destruction, casualties and
damages to economic, livelihoods, infrastructure, services, and health systems. In Ethiopia,
flood disasters and the toll paid in terms of human lives and property damage show an
increasing trend (Billi, Alemu and Ciampalini, 2015). Flash floods are formed from excess
rain falling on upstream watersheds, flow down-stream with massive concentration, high
speed, and typically occur suddenly (Lin, 1999). Heavy downpours in mountainous
highlands can lead to surges of water that turn dry river beds or flood plains into raging
torrents in minutes. Therefore, the damages of such floods become particularly pronounced
and devastating when they pass across or along human settlements and infrastructures. In
2006, the town of Dire Dawa experienced a typical flash flood that, following the heavy
rain on the upland areas of eastern Harerge highlands, within a few hours turned the dry
bed of the Dechatu River into a swelling and devastating river that caused several
casualties and property damage for millions of Euros. The problem of flash floods in the
semiarid area of Dire Dawa is not new, but their frequency has significantly increased in
the last decades (Alemu, 2009; DDAEPA, 2011). This flood affected more than 117,000
people (i.e., one-third of the town population) and officially resulted in the loss of 256
human lives and 244 missing. It caused also the worst property damages to housing and
infrastructures in the town history with an estimated total damage of 10 million USD
(Alemu, 2009).

1.3. Objectives

The main objective of this research to estimate the runoff catchment .This study seeks to
develop hydrological and hydraulic models for Dechatu River. First objective is to estimate
the Run-off modeling of Dire Dawa (Dechatu catchment) develop a physical based
hydrological model with HEC-HMS software and simulate the effects of rainfall on surface
runoff and flood discharges.

The main objective of this research proposal is to estimate the runoff catchment and
intervention scenarios by using, Arc-Hydro, HEC-Geo-HMS, HEC- HMS.

4
 1.3.1. Specific objectives of the studies are

1. To run the applicability of physical based model (HEC- HMS)& to estimate rainfall
run-off yield from the study catchment.

2. To characterize the Runoff catchment yield.

3. To assess and evaluate the spatial variability of runoff yield in the watershed and
identify priority hot spot areas.

4. To assess the impact of different catchment management interventions on runoff

yield.

1.4. Research questions

1. How much Runoff yield estimated from Catchment?

2. Can Watershed management structures reduce the run-off yield in upstream

catchment?

3. What is the systematic approach implement for Rain water retention

andmanagement?

4. What is the impact of catchment management interventions on runoff yield?

1.5. Significance of the Study

• To understand hydrological processes and the availability ofwater resource.

• Reduced soil erosion from catchment, increased crop yields, improved management
of community lands, improved water retention and supply, increased flood
concentration time, improved access.

• Flood Protection Works and associated re-settlement and flood zoning.

• Early Warning System and Catchment Monitoring.

• Soil Conservation, Land Management and Rural Infrastructure reduced soil erosion
from catchment, increased on-farm crop yields, improved management of
community lands, improved water retention and supply, increased flood
concentration time, improved access.

5
CHAPTER TWO

2. DESCRIPTION OF THE STUDY AREA

2.1. Location

Figure 1 Location and map of Dechatu Watershed Dire Dawa Administrate

The Dire Dawa Administration is located in the eastern part of Ethiopia between
geographic coordinates of 9 27’N to 9 50 N Latitude and 41 38 to 42 20 E
Longitude. The Administration is bordered in the South by the East Haraghe Zone of the
Oromia National Regional State, and in the East, West and North by the Shinile Zone of
the Somali National Regional State.Dire Dawa city is accessible by airplane, train and cars,
and is about 515kms road distance to the east of Addis Ababa and 31 l kms to the west of
Djibouti port. The total area of the region is about 128,802ha; out of this urban accounts for
2,684ha (2%) and the balance 98% is for rural (IDP, 2006). The total Dire Dawa area can

6
 be divided in to three major areas; the south and south-eastern part of the city is
characterized by a chain of mountains and upland at the foot of the mountain chain
covering 45%; and low lying flat land (Ephrem, 2006).

2.2. Physiography

The Dire Dawa Administrative Council is generally situated within the altitude range of
950-2,260masl. The study area is characterized by very diverse spatial variation of
topographic features. It ranges from very steep high mountains to flat plains with the
general topography direction being from south towards the north.

Figure 2.Slope Map of Dechatu catchment

The physiography of The DDAC can be classified in to four major

The mountain ranges: - These mountainous areas are located at the southern,
southwestern and southeastern parts of the Council and reach an altitude of some
2,321masl with slopes range of 30-56% and more. The major big waterways start from
these mountainous area and most of the cultivable lands are located on the foot-slopes of
these mountains and inter-ridge valleys on terraced lands.

7
The hills: - These are found scattered all over the watershed, with slope ranging between
16 and 30% with very shallow soils and composed mainly of stones and rock out-crops.

The valley bottoms and river terraces: - These are mainly found at the foot slopes of the
mountain ranges and the river banks with relatively fertile and deep soils. The slope ranges
from 0-5%. These are the units where the rain fed as well as the irrigated crop cultivation
are concentrated.

The flat plains: - These are mostly concentrated in the northern, northeastern and
northwestern part of the study area at an altitude of some 950 – 1000masl with slope
ranging between 0-3 % and mainly used as grazing and browsing of the pastoral livestock
types. These areas comprise of severely dissected undulating to rolling areas caused by
erosion, especially along wadis (slope 2-8%) in the western plain area of shinile area and
moderate to many low to high lying relief hills and rock outcrops in the eastern part
especially in Kelead area.

2.3. Climate and rainfall

The climate is warm and dry with relatively low, erratic precipitation which exhibits spatial
and temporal variability. The main rainfall mechanism is the Inter-Tropical Convergence
Zone (ITCZ). The ITCZ is a low pressure converging zone of the dry tropical easterlies and
moist equatorial westerlies. The position of ITCZ as it migrates north-south determines the
season of the area. The advance of ITCZ to the north brings the spring rains in March. The
Indian Ocean is the moisture source for these spring rains and the main summer rains are
derived from the high pressure system from the Gulf of Guinea and Congo basin. The
ITCZ returns southward in August and in October restoring the dry season. Rainfall data
for the Dire Dawa station cover a 65 year period (1952-2016) and there are few missing
data. The Dengego station rainfall record covers 1981 to 2016 with data missing for just 3
years (1992-94). The Kersa station rainfall record is the shortest of the three stations with
just 35 years of data available from 1996 to 2016. Useful data are also available for two
stations just outside the catchment. The annual average rainfall ranges in Kersa from 317 to
1,038 mm depending on the location and altitude (CRS/Ethiopia: Projects with ECC-
SDCOH-Activity Report, November, 2006). The rainfall in Dire Dawa is highly variable
from year to year and the mean annual rainfall is between 410 and 800mm (DDA, 2003,

8
Integrated Resource Development Master Plan study Project ,Volume III ,Water
Resources)

The highest maximum daily rainfall recorded at any of the above three stations was at
Kersa where 166 mm was measured on 15 April 2001. Very little or no rainfall was
measured at the other two stations on the same date. 113 mm was recorded at Dire Dawa
on 7 December 2003 in the dry season; again no or little rainfall was registered in the other
two stations for the same date. This clearly indicates that extreme rainfall events generally
have limited aerial coverage and are relatively unrelated to the amount of annual rainfall. In
addition they do not necessarily occur during the main rainy season. Lower (in terms of the
depth at any particular point) rainfall events can have much larger areal coverage. Floods
generated from such storms will also have longer durations. Mean annual temperature
ranges from 20-30C and as a result potential evapo-transpiration is relatively high at 3,255
mm on average. Relative humidity at 36% and 40% at elevations of 1,200 masl and 1,800
masl respectively is relatively low. Traditionally Dire Dawa is classified between the Kolla
and the Dry Weyna Dega agro-climatic zones. Kolla lies within the altitude range 500 to
1500 masl and has an annual rainfall of less than 900 mm. This is primarily found on the
gently sloping and plain areas to the foothills and valley bottoms. The Dry Weyna Dega
lies in the altitude range 1500 to 2300 masl and has an annual rainfall of 900 to 1400 mm.
This zone is found on the Harar Plateau and parts of the escarpment (DDA, 2003,
Integrated Resource Development Master Plan study Project ,Volume III ,Water
Resources)

2.4. Regional hydrology

The entire Dechatu River watershed area lies in the Eastern Escarpment of the Awash
River basin, which is one of the dryer regions of the basin (Figure 1). Since the
administration found close to the watershed divide with Wabi-Shebele basin it has only
covered small area. The study area is political boundary, the watersheds of the ephemeral
streams cross the administration, and in some cases most of their watershed areas are
outside the boundary of the administration. There is large topographic elevation difference
in a short distance between the highlands and the rift floor. This favors high groundwater
recharge in the areas where there are large faults.

9
The streams draining from the western Hararghe highlands disappear in the rift margin
faults. This is clearly visible from satellite images and remote sensing. The presence of
thick and wide sand deposits along the stream course may favor indirect recharge (channel
losses) from rivers. Groundwater movement in the study area is mainly towards north
similar to the surface flow. The main sources of the recharge to aquifer (groundwater
replenishment) in basin are subsurface recharge from the escarpment zone, direct recharge
from precipitation and recharge through river beds (wadi-gravel) during times of flow and
rain events.

The climate of the study region is dominated by various inter factors, but the main factors
are the near equator altitude. The year is divided into (Kiremt) from July to mid-October to
February, and a "small rainy" season (April. The small rain, originate from the Indian
Ocean and are brought by Northwest winds; w the Atlantic Ocean with north 9e northern
parts most of location and the three seasons: a main rainy season-September, a dry season
(Belg) in hile the heavy rains in the wet season come from-east winds. The southern parts
of Council get, on the average, about 800 mm of rainfall per annum, while the northern and
north-eastern parts receive about 400 mm annual rainfall. Which flow throughout the year
as that of the regions of the country. Only few intermittent and perennial streams pre
dominate the natural water flow system of the region. According to the study made by the
agricultural development office of the DDAC in the year 1992 EC, the region has over 130
springs with different water discharging capacity and over 44 perennial and intermittent
streams. The most important intermittent and perennial streams that drain the Dire Dawa
region are Dechatu, Butiji, Lega Hare, Dube, Goro and Elbah WWDSE (2004).

There is large topographic elevation difference in a short distance between the highlands
and the rift floor the area to drop its elevation from more than 2,200m at Dhangago to
below 1000m at the north part of Shinile. Groundwater movement in the study area is
mainly towards north similar to the surface flow. The main sources of the recharge to
aquifer (groundwater replenishment) in basin are subsurface recharge from the escarpment
zone, direct recharge from precipitation and recharge through riverbeds during times of
flow and rain events.

10
2.5. Geology

The geology of the Administration has been studied or included in the studies conducted
during different time periods with different scales and well documented. Some of the major
studies include those conducted by the Geological Survey of Ethiopia and mapped at a
scale of 1:2000, 000 (Mengesha et al., 1996; Kazmin, 1973), and the National Research
Council of Italy done for Ethiopia and Somalia (Merla et al., 1973). These studies, owing
to their coarse scale, only provided a broad view of geology of the area. These studies
identified three main rock units formed during different periods.

Ethiopian Geological Survey also carried out geological mapping of the Dire Dawa Sheet
(NC 37-12) at a scale of 1:250,000, which covered the DDAC up to 42o00’ E longitude
(Seife, 1985: cited in WWDSE, 2003). Greitze (1961, 1970) prepared a 1:40,000 scale
geological map for the central part of the Administration. More recently, the Water Works
Design and Supervision Enterprise conducted a study on geology and
mineral resources of the Administration. General observation of geology and mineral
resources was made, for these are substrates for the formation of soils. In areas where
moisture is limiting and the dominant type of weathering is the physical one, the
behavior of the local soils is very much affected by the parent material from which it was
formed. According to this team, the major rocks identified in the Administration are
dominantly granite, limestone, basalt, some unidentified metamorphic looking rocks, and
unconsolidated sediments. The granite occupies the higher elevation areas around
Dengego, while the limestone occupies extensive area in the lowlands.

2.6 Selection of hydrological model

Each model type serves a purpose, and a particular model type may not categorically be
considered more appropriate than others in all situations. Choice of a suitable model
structure relies heavily on the function that the model needs to serve.

There are various criteria which can be used for choosing the right hydrological model for
a specific problem. These criteria are always project dependent, since every project has its
own specific requirements and needs. Further, some criteria are also user-depended (and
therefore subjective). Among the various project-dependent selection criteria, there are four
common, fundamental ones that must be always answered (Cunderlik):-

11
 I. Required model outputs important to the project and therefore to be estimated by the
model (Does the model predict the variables required by the project?).

II. Hydrologic processes that need to be modelled to estimate the desired outputs
adequately (Is the model capable of simulating single-event or continuous Processes?)

III. Availability of input data (Can all the inputs required by the model be provided within
the time and cost constraints of the project?).

IV. Price (Does the investment appear to be worthwhile for the objectives of the project?).

The selection of model were by considering the above criteria’s with inclusive of
availability of data, level of application, purpose, required accuracy, space and time scale,
catchment area, simplicity, previous trends(studies) in the surrounding area & Ethiopia as a
whole. Considering all the criteria’s set above data driven model (Artificial Neural
network) and physical based models (SWAT) were adopted for this study.

2.7. Reason for selection of (HEC-HMS) model

HEC-HMS model was selected due to the following reasons:-

i. It has been indicated and tested by different researcher and journal paper that the model
is calibrated and simulated with satisfactory results on Dechatu catchment.

ii. It has been tested that the model has obvious advantage as a hydrological modeling tool
that includes modularity, computational efficiency, ability to predict long-term impacts as a
continuous model, and ability to use readily available global datasets, availability of a
reliable user and developer support has contributed to its acceptance as one of the most
widely adopted and applied hydrological models worldwide.

iii. The model simulates the major hydrological process in the watersheds, less demanding
on input data, and it is readily and freely available.

iv. Specifically the model was tested for prediction of runoff yield in Dechatu watershed
with satisfying results, since HEC-HMS can be applied for the same land use /land cover
and topographical conditions it will have good performance on watershed too. Moreover,
projections of flood changes at river basin scale are scarce in the scientific

12
literature. These same limitations and uncertainties are presently effective also in Ethiopia.
In the last decade, an increasing occurrence of floods is reported for Africa in general,

2.8. Meteorological Data Analysis

Meteorological data sets are the key inputs for hydrological modeling purpose, but
the selection representative meteorological gauging station depends on the data availability
(including existence of enough length of record and distance from the area of interest).
Gauging stations were selected based on their relevance for the study. Nevertheless, daily
metrological d for only seven gauging stations were obtained from the Hydrology
Department of the Ministry of Water Resources. Locations of these seven gauging stations
are as in the Figure. Even if the Model (HEC-HMS) us the nearest station from the centroid
of each sub basin, the other unused station in the model are used for t purpose of filling in
missed data and computation of areal rainfall.

Adequate number of gauging stations might have been established in the Blue Nile Basin.
However, long records are not available at most of the stations and the reliability of
available data is also questionable. The number of operational stream flow gauging stations
has increased in recent years (MoWR, 2019). This might have been recorded as good
change which may only be increase in number rather than quality oriented improvement.
The Hydrological Department under the Ministry of Water Resources is responsible for
hydrological and sediment data collection, processing and distribution to the data users.

2.9. Filling Missed Meteorological data’s

Complete record of meteorological data is required for hydrological modeling and analysis.
There exists small and large duration of missed records in the selected meteorological
stations especially on rainfall and temperature. Data for the period of those missing data
could be filled using estimation technique. Arithmetic mean, Normal ratio, Regression and
distance power method are the most commonly used methods for estimation of missing
rainfall & temperature data sets. Depending on the simplicity and length of missing data,
regression and distance power method are selected for estimation of missing temperature &
rainfall data.

13
Distance Power Method: The method is easy to use for fill in missing precipitation data
(Hubbard, K.G, 1994). It is reliable if only the weather stations are situated inside of 100
Km radius (Tronci N, Molteni F, Bozzini, M, 1986).

The rainfall at a station is estimated as a weighted average of the observed rainfall at the
neighboring stations. The weights are equal to the reciprocal of the distance or some power
of the reciprocal of the distance of the estimator stations from the estimated stations. Let Di
be the distance of the estimator station from the estimated station. If the weights are an
inverse square of distance, the estimated rainfall at point of interest (missed) is: Where Pi
and PA are precipitation at neighboring stations and at the target station respectively Di is
the distance between the target station and the neighboring stations, estimated as: Where X
and Y are the coordinates of the station whose data is estimated and Xi and Yi are the co-
ordinates of stations whose data are used in estimation.

Regression Method: This method of data infilling is applied for gaps having short duration
of missing value for temperature and precipitation by considering strong correlation from
the existing stations.

2.10. Data Quality Assessment

Engineering studies of water resources development and management depends on

hydrological and meteorological data. These data should be stationary, consistent, and
homogeneous when they are used for frequency analyses or to simulate a hydrological
system. To determine whether the data meet these criteria, the engineer needs a simple but
efficient screening procedure statistical variability. Accordingly, in this study, the data
quality assessment goes through the following key tests. Rough screening of the data and
compute or verify the totals for the hydrological year test for Absence of trend Test for
stationery of time series. Homogeneity Test Consistency test.

2.11. Test for Absence of trend

After plotting a time series, one must be sure that there is no correlation between the order
in which the data have been collected and the increase (or decrease) in magnitude of those
data. It is common practice to test the whole time series for absence of trend.

14
Accordingly test for absence trend is checked by applying spearman correlation method in
this study as it was simple, distribution free and power full for both linear & nonlinear
trend. Where n is the total number of sample data’s, D is the difference and Rsp is spear
man correlation coefficient. Here is the rank of the variable and K is chronological
transformed series for observation y. The Null hypothesis is finally checked for the
acceptance with t-test statics which is described by

2.12. Test for stationery

A time series of hydrological data is strictly stationary if its statistical properties (e.g. its
mean, variance, and higher-order moments) are unaffected by the choice of time origin.
(By ‘unaffected’, we mean that estimates of these properties agree within the range of
expected. Stationery of time series was checked based upon split-record tests for stability
of the variance (F-test) and stability of mean (t-test). The tests for stability of variance and
mean verify not only the stationary of a time series, but also its consistency and
homogeneity. In the basic data-screening procedure, these two tests are reinforced by a
third one, for absence of trend. Although stability of these two properties indicates only a
weak form of stationary, this is enough to identify a non-stationary time series or to select
those parts of a time series that are acceptable for use. A time series is first divided in to
subsets and computation of standard deviation for each subset will be attained first, then
variance of the two subsets is checked for the acceptable range for stability of variance.

The t-test for stability of mean involves computing for the two subsets and then comparing
the means of two or three non-overlapping sub-sets of the time series (the same subsets
from the F-test for stability of variance) A suitable statistic for testing the null hypothesis,
Ho: XI = X, against the alternate hypothesis

2.13 Test for homogeneity

The data qualities with regard to possible temporal and spatial variations or errors should
have to be investigated by checking homogeneity and consistency of selected stations.
Non- homogeneity is a change in the statistical properties of the time series. Its causes can
either natural or man-made. These include alterations to land use, relocation of the
observation station, and implementation of flow diversions. Rainbow and non-dimensional
plot are the widely used methods for checking homogeneity of time series data’s of rainfall.

15
In this study Absolute homogeneity is checked by rainbow software and relative
homogeneity is checked by non-dimensional plot.

In RAINBOW the test for absolute homogeneity is based on the cumulative deviation from
the mean and clearly shows the probability of rejecting homogeneity.

There exist two types of homogeneity called absolute homogeneity and relative
homogeneity; once the absolute homogeneity of each station alone is checked their relative
homogeneity is then checked (by using Non Dimensional Plot) in this study.

2.14. Test for Consistency

A time series of hydro meteorological data is relatively consistent if the periodic data are
proportional to an appropriate simultaneous time series.

Double mass curve is a simple, visual and practical method, and it is widely used in the
study of the consistency hydro-meteorological data and it was a commonly used data
analysis approach for investigating the behavior of records made of hydrological or
meteorological data at a number of locations. It is used to determine whether there is a need
for corrections to the data to account for changes in data collection procedures or other
local conditions. Such changes may result from a variety of things including changes in
instrumentation, changes in observation procedures, or changes in gauge location or
surrounding conditions. Double mass analysis for checking consistency of a hydrological
or meteorological record is considered to be an essential tool before taking it for analysis
purpose. This method is based on the hypothesis that each item of the recorded data of a
population is consistent. Double mass curve method is adopted for checking consistency of
both meteorological and hydrological data.

2.15. Data Available

There are three important rainfall stations within the catchment area, Dire Dawa, Kersa,
and Dengego (there are additional rain gauges in adjacent catchments that are relevant to
some aspects of the hydrological analysis – see below, Section 3.6). Dire Dawa city is the
center of the flood problem at the downstream end of catchment. Dengego is located in the

16
upper part of the Dechatu catchment near the watershed and Kersa likewise in the Chirecha
(Goro) catchment.

17
CHAPTER THREE

3.0 LITERATURE REVIEW

3.1. General

Simple modeling approaches with a fewer model parameters are generally regarded as an
accepted strategy in rainfall-runoff modeling (Nash and Sutcliffe, 1970; Beven 1989, 1993;
Jakeman and Hornberger 1993; Jakeman et al. 1990; Young and Beven 1994; McIntyre and
Al-Qurashi, 2009). The value of making the model more complex in light of the general
data restrictions in arid and semiarid regions is questionable (Al-Qurashi, 2008). Major
research priorities toward improved hydrological models for arid regions include improved
methods for rainfall observation and modeling. The Soil Conservation Service SCS Soil
Conservation Service 1971 curve number CN methodology has been widely used because
it represents reliable procedures that have been implemented for many years in different
parts of the world. It is computationally efficient and its inputs are generally available. SCS
methodology relates runoff to soil type, land use, and management practices. There are
many classification schemes of hydrologic models, based on the method of representation
of the hydrologic cycle or a component of the hydrologic cycle (Cunderlik, 2003).
Hydrologic simulation models use mathematical equations to calculate results like runoff
volume or peak flow.

3.1.1 Model and software description

This chapter provides an insight on the theoretical and mathematical background involved
behind the software describing the data processing and modelling procedures. Mainly three
open source software are used for the project: ArcGIS, Arc Hydro,HEC-GeoHMS and.
HECHMS is used for hydrologic modelling. While ArcGIS is used as a platform for
generating physical basin models for HEC-HMS and geometric model of river using
interfacing hydrological extensions HEC-GeoHMS.

3.1.2 ArcGIS: HEC-GeoHMS

HEC-GeoHMS is an interfacing tools between GIS and HEC-HMS. HEC-GeoHMS is a

geospatial hydrology toolkit in ArcGIS to create hydrologic inputs that can be directly used
with HEC-HMS. It allows to visualize spatial information, extract watershed physical
characteristics from DEM (Digital Elevation Model) and GIS data, perform spatial

18
analysis, and delineate sub basins and streams to develop hydrologic parameters as well as
construct inputs to hydrologic models. (Fleming et. al., 2013) HEC-GeoHMS 10.4.2
version was used during this project.

3.1.3. Hydrological Modeling

Hydrological models are characterizations of the real world system. Modeling of the
rainfall runoff processes of hydrology is needed for many different reasons the main
reasons being limited range of hydrological measurement techniques and limited range of
measurements in space and time Anderson, (M.J. & Burt, T.P). A watershed model
simulates hydrologic processes in a more holistic approach compared to many other models
which primarily focus on individual processes or multiple processes at relatively small-or
field-scale without full incorporation of a watershed.

A watershed model simulates hydrologic processes in a more holistic approach compared

to many other models which primarily focus on individual processes or multiple processes
at relatively small-or field-scale without full incorporation of a watershed area. Watershed-
scale modeling has emerged as an important scientific research and management tool,
particularly in efforts to understand and control water pollution. A model is physical or
mathematical description of a physical system, including the interaction with its outside
world, which can be used to simulate the effects of changes in the system itself or the
changes in the condition imposed upon it. The primary features for distinguishing
watershed-scale modeling approaches include the nature of the employed algorithms
(empirical, conceptual, or physically-based), whether a stochastic or deterministic approach
is used for model input or parameter specification, and whether the spatial representation is
lumped or distributed

Watershed models can also be categorized as deterministic or stochastic depending on the

techniques involved in the modeling process. Deterministic models are mathematical
models in which outcomes are obtained through known relationships among states and
events. Stochastic models will have most, if not all, of their inputs or parameters
represented by statistical distributions which determine a range of outputs. Even though
most models are deterministic in nature, stochastic models provide two important
advantages. First, their conceptually simple framework makes it possible to describe
heterogeneity when there are limited spatial or temporal details. Second, they provide

19
decision makers with the ability to determine uncertainty-associated with prediction.
Empirical models consist of functions used to approximate or fit available data. Such
models span arranges of complexity, from simple regression models to hydro informatics-
based models which utilize Artificial Neural Net-works (ANNs), Fuzzy Logic, Genetic,
and other algorithms

3.1.4 Rainfall Runoff Model: HEC-HMS

Hydrologic Modelling System (HEC-HMS) is an open source computer software

developed by U.S. Army Corps of Engineering´s Hydrologic Engineering Center that helps
in simulating the hydrologic cycle (precipitation, evapotranspiration, infiltration, surface
runoff and base flow) of a catchment by describing its physical and meteorological
properties. A simple schematic representation of rainfall runoff process replicated in HEC-
HMS is shown in figure 3. Wide options of mathematical models for all the hydrological
components that conceptually represent watershed behavior are incorporated in this
program. The program uses separate model to represent each component of the runoff
process like model to compute runoff volume, model of direct runoff/base flow/ channel
flow as well as alternative models to account for the cumulative losses for e.g.: SCS CN
loss model. Then, it computes runoff volume by subtracting losses (infiltration, storage,
interception, evaporation etc.) from precipitation. HEC-HMS 4.2.1 was used during this
research. (Fleming and Brauer, 2016)

20
Figure 3. Representation of rainfall runoff process

21
The HMS is designed to simulate the precipitation-runoff functions of watershed systems.
HEC-HMS provides a variety of options for simulating precipitation-runoff processes. In
addition to UH and hydrologic routing options, the model implements a distributed runoff
transformation that can be applied to gridded e.g., radar rainfall data, and a “moisture
depletion” option that can be used for continuous simulations. It is designed to be
applicable for a wide range of geographic areas and for solving various possible ranges of
problems. These include large river basin water supply and flood hydrology as well as
small urban or natural watershed runoff. The resulted hydrographs can be used for studying
water availability, urban drainage, flow forecasting, future urbanization impact, reservoir
spillway design, flood damage reduction, floodplain regulation, and others. Prior to
program execution, three input data sets are required. The first, labeled basin model,
contains parameters and connectivity data for hydrologic elements sub basin, routing reach,
junction, reservoir, source, sink, and diversion. The second set, precipitation model,
contains meteorological data. The final input data set, control specifications, specifies
simulation control parameters.

3.1.5. Basin Model

The program uses separate model to represent each component of the runoff process like
model to compute runoff volume, model of direct runoff/base flow/ channel flow as well as
alternative models to account for the cumulative losses for e.g.: SCS CN loss model. Then,
it computes runoff volume by subtracting losses (infiltration, storage, interception,
evaporation etc.) The resulted hydrographs can be used for studying water availability,
urban drainage, flow forecasting, future urbanization impact, reservoir spillway design,
flood damage reduction, floodplain regulation, and others. Prior to program execution,
three input data sets are required. The first, labeled basin model, contains parameters and
connectivity data for hydrologic elements sub basin, routing reach, junction, reservoir,
source, sink, and diversion. The second set, precipitation model, contains meteorological
data. The final input data set, control specifications, specifies simulation control
parameters.

22
•Meteorological Models:

Information regarding meteorological components such as temperature, precipitation

evapotranspiration, sunshine, humidity, and snowmelt is defined in meteorological model.
HEC-HMS provides variety of options to define each meteorological element.

• Control Specification:

Starting date and time, ending date and time and computational time step for the simulation
are defines in control specification.

• Time series Data:

Real time series data for all the meteorological elements defined in meteorological model
are fed in this part. Apart from above mentioned meteorological element, discharge data
can also be supplied for calibration and simulation of the developed model.

• Paired Data:

Meteorological data in tabular/graphical form are supplied as paired data. (Scharffenberger,

2016) In HEC-HMS, the hydrological procedure of changing rainfall into runoff has been
represented by four processes: loss, transform, baseflow and transform. These processes are
described in following section: HEC-HMS 4.2.1 was used during this studies (Fleming and
Brauer, 2016)

• Transform method

Transform methods is an approach for computing direct runoff at the outlet of watershed
area from the excess precipitation falling over it and this is done based on principles of unit
hydrograph. Unit hydrograph can be defined as the runoff hydrograph produced from
excess rainfall of unit depth occurring over the watershed. The theories of unit hydrograph
are:

(i) Excess precipitation and runoff produced are directly proportional to each other,

(ii) Excess precipitation is distributed uniformly with respect to time and space over the
watershed area and

23
(iii) Runoff produced from given excess rainfall is independent of time of occurrence and
precedent moisture content (Subramanya, 2008).

The transformation method used for this study was SCS Unit Hydrograph. The resulting
runoff hydrograph from this model is described by properties of unit hydrograph using one
or more equations of the parameters involved. The peak of unit hydrograph and its time of
peak is given by following equations.

Where, Up = Peak of unit hydrograph, A = Area of watershed, Tp = Time of peak, Δt =

Excess precipitation duration and tlag = Basin lag (Feldman, 2000)

Basin lag can be defined as the time difference between the peak of unit hydrograph and
centroid of the associated excess rainfall hyetograph which is depicted in the figure 3.3
below?

3.1.6 Flood frequency analysis

Flood frequency analysis is an estimation of how often a certain amount of flow is

reoccurring. Such estimation is pre-requisite for carrying out hydraulic computation of
river and developing flood inundation map. The analysis is done by fitting a probability
model to the sample of annual extreme flood values recorded over a long period of time,
for a catchment. The model parameters established can then be used to predict the extreme
events of large recurrence interval (Pegram and Parak, 2004). For this project, Gumbel
Distribution method has been selected for flood frequency analysis.

Gumbel’s distribution is a statistical method commonly used for predicting extreme

hydrological events such as floods (Haan, 1977). According to Gumbel, the probability (P)
of occurrence of any extreme event is given by the following equation.

Qsurf = 1.285
(x − x0 ) + 0.577
( x ) 2.1

24
P = ( X  x0 )1 − e− ey
2.2

(YT − 0.577 )
K = YT + 0.577
1.2825
2.3

And

YT = ln ln 
T
T − 1 (2.5)

(2.4)

Recurrence interval T is calculated

1
T=
P (Subramanya, 2008)

3.1.7. Previous studies of the area

In the past, flood studies for Dire Dawa town have been done by different organizations
before and after the most severe flood event occurred on the fifth day of August 2006.

Dire Dawa administrative council integrated resources development master plan study
project (January 2004) this study was conducted before the most severe flood event. it is
intensive and consistent study participating Dire Dawa administrative council waters,
mines & energy office as a client and water works design & supervision enterprise as a
consultant. According to this study, dechatu catchment flood estimation summary is

25
Table 1. Dechatu catchment flood estimation summery

Peak Flood (m3/sec) at Various Return Periods

Catchment
50 100 200 500 1000
Rational 863 1082 1143 1330 1936
Boldakov 792 865 913 985 1009
Dehatu SCS 96 105 141 152 156
Creager 629 715 755 834 1032

Selected Rational 863 1082 1143 1330 1936

peak

The design of flood protection works and spillways, the average values of the Rational,
Boldakov and Creager methods are considered. Also the study states, as per the interview
of the older people, a flood mark was fixed on the Dechatu River and its flood was
estimated as 1170m3.The general comment is that results from different methods shall be
compared, not averaged. Usual practice is to use the discharge that best reflects local
conditions.

Awash River Basin Flood Control and Watershed Management Study Project (February
2008).”The general comment is that results from different methods shall be compared, not
averaged. Usual practice is to use the discharge that best reflects local conditions. This
study was conducted after the most severe flood event. The client and consultant are
Ministry of Water Resources and Halcrow respectively. This report is not consistent; the
data are not commensurate/corresponding/matching and tailored to the methodology. The
given data shows that average data for the whole catchment considered as one (lumped).
However, the catchment is divided into three sub catchments (distributed) to use HEC-
HMS.

➢ How can we run uniform rainfall data (distributed) across the whole area of three
sub catchments without having their specific data (lumped data)?
➢ The options used for each step to estimate the maximum flood discharge is not clear,
Which Metrologic station data used?

26
 ➢ Does the Runoff formed by Localized rain storms? Flooding formed by localized
rainstorms: the runoff was mainly generated by infiltration-excess when the flooding
had a rapid rise and recession with a single peak value but the total flood volume
was small.

➢ Does the runoff formed by uniform rainfall across the whole area? How much
volume of run-off was estimated?

➢ Flooding formed by mixed rainfalls: the runoff was due to both infiltration- and
saturation-excess when the flooding possessed the two previous characteristics.

Even if these questions are not clear in the report, according to this study, Dechatu
catchment flood estimation summary is given as follows. The following major points are
not clear in the report:

Table 2. Summary of Awash River Basin Flood Control and Watershed Management Study
Dechatu Catchment
Return Period(years) Estimated Peak Discharge
(m3/Sec)

10 901
25 1166
50 1368
100 1582
200 1750

27
4.5. Runoff Estimation

The Curve Number method (SCS, 1972), also known as the Hydrologic Soil Cover
Complex Method, is a versatile and widely used procedure for runoff estimation. In this
method, runoff producing capability is expressed by a numerical value varying between 0 -
100. In the past 30 years, the SCS method has been used by a few researchers because it
gives consistently usable results (Rao et. al., 1996; Sharma et al. 2001; Chandramohan and
Durbude, 2001; Sharma and Kumar, 2002 for runoff estimation.

The SCS curve number equation is (SCS, 1972):

Qsurf =
(R day − Ia )
2

(R day − Ia + S)
(2)

Where Qsurf is the accumulated runoff or rainfall excess (mm), Rday is the rainfall depth
for the day (mm), Ia is the initial abstractions (mm), which includes surface storage,
interception and infiltration prior to runoff, and S is the retention parameter (mm). Runoff
will only occur when Rday > Ia. The retention parameter varies spatially due to changes in
soils, land use, management and slope and temporally due to changes in soil water content.
The retention parameter, S is defined as:

 1000 
S = 25 .4 − 10 
 CN  (3)

Where CN is the curve number for the day. The initial abstractions, Ia, is commonly
approximated as 0.2S and equation (2) becomes

Qsurf =
(R day − 0.2S )
2

(Rday + 0.8S )
(4)

Where CN (called the SCS curve number) is used to represent the combined effects of the
primary characteristics of the catchment area, including soil type, land use and the previous
moisture condition. It takes a value in the range of 30–98.

28
29
 4.0 METHODOLOGY

4.1 Hydrological Model Development

Rainfall runoff modelling was carried out with the help of HEC-HMS and HEC-GeoHMS
a hydrological extension in ArcGIS. Detailed description regarding these software has been
done in Section 3.1.2.

An overview of working mechanism of rainfall runoff model is shown with the help of
schematic diagram below in figure 4.

The methodology used for carrying out Rainfall Runoff Modelling can be described by
categorizing them into two sections, which are as follows:

• Creating Basin Model

• Developing Hydrological Parameters

• Hydrological Modelling

• Creating Basin Model

Basin model was created with the help of HEC-GeoHMS, a hydrological toolkit in
ArcGIS.

Figure 4 Modeling approach Rainfall Runoff Modeling

30
4.2. Terrain Pre-processing

Before carrying out terrain pre-processing, the input terrain data DEM was refined using
DEM reconditioning. After this process, the DEM was pre-processed in HEC-GeoHMS to
derive sub-basins and drainage network of the catchment. The steps included were fill
sinks, flow direction, flow accumulation, stream definition, stream segmentation,
catchment grid delineation, catchment polygon processing, and drainage line processing
and ad joint catchment processing. After terrain pre-processing, HEC-HMS project was
created. At first, a project point was defined at the downstream end of the watershed based
on which the software delineated the project area. The resulting project area for Dechatu
river watershed was 327.3381 km2

Some major characteristics of the basin model for Dechatu are tabulated in Table 4.2
below:

Table 3: Basin Model characteristics

Characteristics Dechatu Watershed

Basin area (km2) 327.3381

Number of sub-basin after terrain pre-

8
processing
Number of batch points 1

4.3. Developing hydrological parameters

This step parameterizes the values of different hydrological processes involved in

modelling. The hydrological parameters were estimated by using the land and soil use data
for each sub basin. Different steps involved for developing hydrological parameters are as
follows: Select HMS processes: HMS processes for modelling transform, were selected. In
this project, same HMS processes were selected for study area which are listed below in
Table 4.3.

The precipitation data from the three stations were weighted at the center of each sub-basin
using Tyson polygon method. The historical precipitation and discharge data show that
rainfall and flooding in the Dechatu catchment area present three main characteristics, i.e.:

31
• Flooding formed by localized rainstorms: the runoff was mainly generated by infiltration-
excess when the flooding had a rapid rise and recession with a single peak value but the
total flood volume was small.

• Flooding formed by uniform rainfall across the whole area: the runoff was dominated by
saturation-excess when the flooding changed little over the period of time and the total
flood volume was large.

• Flooding formed by mixed rainfalls: the runoff was due to both infiltration- and saturation-
excess when the flooding possessed the two previous characteristics

Table 3Selected methods for HMS

HMS Processes Method

Transform SCS Unit Hydrograph
Routing Muskingum

CN Lag: This function calculated lag time for transform method based on CN grid. CN
grid was generated in ArcGIS using land use and soil cover layers. CN values adopted for
different land use type for Dechatu watershed are given in Table 4.4 and Table 4.5
below.

Table 5: CN values adopted for Dechatu

Description A B C D
100 100 100 100
Developed areas 61 75 83 87
Pasture 68 79 86 89
Cropland 71 80 87 90
Description A B C D
Water bodies 100 100 100 100
Developed areas 61 75 83 87
Forest 68 79 86 89
Cropland 71 80 87 90
Barren land 76 85 90 93

32
For each of the sub-basins and river, physical characteristics were computed based on the
refined DEM. The computed characteristics for river included river length and river slope
and for basin included basin slope, longest flow path to the basin, basin centroid, centroid
elevation and centroidal longest flow path. To calculate basin slope, watershed slope was
required which was calculated using Arc Hydro tool. Batch point was onein the Dechatu
watershed.

4.4Developing HEC-HMS model files:

In this step, model files such as background- map file, basin model file and meteorological
model file required for HEC-HMS were generated. At first, all the physical characteristic
values of reaches and sub-basins were converted to user defined unitary system. In this
case, SI system was used. After this HMS basin schematics and legends were added to the
basin map. HMS basin schematics included HMS links that represented river and
HMSnodes that represented sub-basins and junctions. By adding HMS legend, HMS nodes
representing sub-basins and junctions were replaced with HMS legend. Further,
coordinates were added to the features in HMS nodes and HMS links.

Finally, background-map file and basin file were created for exporting them to HMS.
Gauge
weight method was chosen for creating meteorological model file for basin. For using
this method, Thiessen polygon for the available precipitation stations within or in the
periphery of the basin area was created in ArcGIS.

4.5.Hydrological Modelling

After completion of building model framework in HEC-GeoHMS, modelling was

performed in HEC-HMS by importing files from ArcGIS which were:

• Background- map and river

• Basin file with extension ‘Watershed’
• Meteorological files with extensions ‘Met’ and ‘Gage’

33
 .

Figure 5. The hydrological model implemented in HEC-HMS.

MEAN MONTHLY RF AND ETO

300 DENGEGO STATION 30

250 25
(MM)

200 RF (mm) ETo (mm) 20

150 15

100 10

50 5

0 0
Jan Feb Mar Apr May MONTHS
Jun Jul Aug Sept Oct Nov Dec

Figure 6mean monthly Rainfall (mm) Dengego Station Rain fall

After the meteorological components were defined, real time series data defining them
were entered. Description of all the time-series data entered for both models are given in
Section 3.1.2. All time-series data were entered manually for both models. Finally,

34
simulation run time and computational time step was set in control specification. Control
specifications for calibration and validation of bath basins are tabulated in Table 4.7.

4.6. Model Inputs

This chapter gives a detailed overview on data required for the model development and
stepwise procedures for the simulation and modelling.

4.7. Digital Elevation Model (DEM)

Topography is defined by a Digital Elevation Model (DEM), which describes the elevation
of any point in given area at a specific spatial resolution as a digital file. A digital elevation
model is needed for raster-bas hydrological analysis in a GIS. The DEM used in the study
was a (30*30 m) made resolution elevation data which was taken from Minister of water
GIS department. The DEM was used to generate percent slope values, to automatically
delineate watershed boundary, stream networks, and identify gage outlets. It is
characterized by chain of mountains that stretch along the south-eastern and western parts
bordering Oromia region,

35
Figure 7. DEM of Dechatu River watershed area

Isolated hills in the low lands and extensive plains surrounded by hilly areas towards the
north-western part. DEM (Digital Elevation Model) represented the topographic feature of
the study area. For Dechatu, characterized by diverse topographic features 30*30 m
resolution clip from an open source, Awash River basin.

4.8. Soil Data and Maps (Hydrologic Soil Groups)

Soil properties influence the relationship between rainfall and runoff by affecting the rate
of infiltration. The SCS has divided soils into four hydrologic soil groups based on
infiltration rates (Groups A, B, C, and D). These groups were previously described for the
Rational Formula (see Section 5.7, Table 5-7).

Consideration shall be given to the effects of urbanization on the natural hydrologic soil
group. If heavy equipment can be expected to compact the soil during construction or if
grading will mix the surface and subsurface soils, appropriate changes shall be made in the
soil group selected. Also runoff curve numbers vary with the antecedent soil moisture
conditions, defined as the amount of rainfall occurring in a selected period preceding a
given storm. In general, the greater the antecedent rainfall, the more direct runoff there is
from a given storm. A five-day period is used as the minimum for estimating antecedent
moisture conditions.

Leptosols (15% of area) are mostly found on the mountain and hilly areas and are
characterized by very shallow depth. These soils are seldom cultivated, being droughty
even with soil conservation measures. They are largely left as open access communal
grazing land. The excessively drained Fluvisol (22%), Heptlic Xerosol (55%) and
Solonchak (4%) soil types are found on flat, alluvial plains and in valley bottoms and
depressions. They are deep with medium to coarse texture. Very small areas of Cambisols
are found on the undulating and rolling plains and have moderately deep, medium texture.
Antecedent soil moisture conditions also vary during a storm; heavy rain falling on a dry
soil can change the soil moisture condition from dry to average to wet during the storm
period

36
 Figure 8. Soil Map of Dechatu River Watershed

Table 6 Soil and area coverage

Sub-Basin Hydrologic Soil
Name
Soil Type Group
Calcaric
B
W160 Fluvisols
W200 Dystric nitisols B
Gypsic
D
W150 yermosols
W180 Haplic Xerosols B
W190 Leptosols D
Orthic
B
W210 solonchaks
Gypsic
B
W220 Yermosols
W140

37
38
4.9. GIS based land use/cover change assessment

Runoff is rainfall excess or effective rainfall - the amount by which rainfall exceeds the
capability of the land to infiltrate or otherwise retain the rainwater. The principal physical
catchment area characteristics affecting the relationship between rainfall and runoff are
land use, land treatment, soil types, and land slope.

Land use is the catchment area cover, and it includes both agricultural and nonagricultural
uses. Items such as type of vegetation, water surfaces, roads, roofs, etc. are all part of the
land use. Land treatment applies mainly to agricultural land use, and it includes mechanical
practices such as contouring or terracing and management practices such as rotation of
crops.

The SCS uses a combination of soil conditions and land-use (ground cover) to assign a
runoff factor to an area. These runoff factors, called runoff curve numbers (CN), indicate
the runoff potential of an area. The higher the CN, the higher is the runoff potential.

To assess the land use/cover change, Multi-temporal (Landsat MSS 1985, Landsat TM
2000, and Landsat ETM + 2014) remote sensing data of the area was used. Image
enhancement, rectification and classification were also applied on the raw images. The land
cover condition of two different periods was spatially compared (1985 to 2000 and 2000 to
2014) and the rate and quantity of change was calculated. Image classification was only
used for the extraction of distinct classes or land use/land cover categories from satellite
imagery. Ground survey was also made to cross check some of the features identified from
satellite image. Detail ground surveywas conducted to assess tree species diversity and the
status of forests. Based on knowledge of the study area, visual interpretations of imagery
and detailed reconnaissance field survey, different land use/land cover categories were
distinguished.

39
Figure 9. Land use\Land cover Map 1985. (Landsat MSS 1985)

40
Figure 10 Land use\Land cover March (Landsat TM 2000)

41
Figure 11 Land use\Land cover (Landsat ETM + 2014)

42
Table 7. Land use and area coverage

Land use and area coverage (m2)

Sub-Basin Name Land use Area (m2)

1465426.089
Bare land 13614529.814

300001
Cultivated land 2375623.222
Urban and / or
built up area vegetation 1465426.089
8289619.470
300002 Urban and / or built up area 8289619.470
Bare land
10937772.287
Bare land 10937772.287
Cultivated land
15457133.938
Cultivated land 15457133.938
300003
vegetation 8289619.470
954918.964
Physiognomic 3618165.897
R290W290 Physiognomic
vegetation
181575.748
Bare land
3618165.897 Bare land 15996685.663
2785613.103
Physiognomic
300004
8289619.470
R310W310 Cultivated land 181575.748
Bare land Physiognomic
1705976.569 vegetation
18197477.950
1718837.683
Cultivated land 1179814.008
300005
Physiognomic
vegetation
181575.748

Bare land
15996685.663

18197477.950

1179814.008

Bare land
13465828.591

19759198.658

Physiognomic
vegetation
1566783.216

Bare land
9314750.141 43
d area coverage
4.10 Rainfall Frequency

Rainfall frequency analysis was used as a key input to the calculation of flood frequency
and magnitude. .

The rainfall frequency analysis was carried out using the available 24 hour rainfall data for
the rain gauge stations situated in the headwater area. Such data is limited and it was
necessary to combine data from different stations in order to arrive at an acceptable
representative record. The probability of non-excedence of a given amount of precipitation
in a 24 hour period is presented in Figure 4.

Though rainfalls are registered over 24 hours, rainfall the actual duration of the storm will
normally be much less than 24 hours and will include all or nearly all of the recorded 24
hour rainfall amount. For this reason it has been assumed that the maximum rainfall
duration of large storms is 6 hours.

4.11 Rainfall Frequency Analysis

4.12 Rainfall Intensity

The highest maximum daily rainfall recorded at any of the above stations was at Kersa
where an amount of 166 mm was measured on April 15, 2001. Very little or no rainfall was
measured at the other two stations on the same date. 113 mm was recorded at Dire Dawa
on December 7, 2003 during the “dry” season and again no or little rainfall is registered at
the other two stations at the same date. This clearly indicates that extreme rainfall events of
this type will generally have limited aerial coverage and are relatively unrelated to the
amount of annual rainfall. In addition, they do not necessarily occur during the main part of
the rainy season which clearly complicates the issue of flood preparedness. Lower (in terms
of the depth at any particular point) rainfall events can have much larger areal

44
 0.2 35

30
R² = 0.9425 25

Return Period
0.1
20

15
0
10

R² = 0.9119 5

-0.1 0
0 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 160
RF (mm)

Figure 12 Frequency annual daily maximum RF values using Gumbel (EV1)

and Log Pearson III distributions.

4.13 Main rivers and run-off characteristics

There was no flow in the ephemeral rivers and most of the riverbeds were observed that
they were completely dry. It was evident from observation of their courses and interviews
with the local people that run-off in these streams mostly occurs when there is rainfall in
the highlands in the respective catchment. The spate hydrographs characteristically
indicate a sharp rise to the peak discharge within a short duration. The duration of run-off
is nearly equal to that of rainfall and the run-off hydrograph has very steep rising and
falling limbs with a very pronounced peak. In general, run-off occurs vigorously and very
quickly and ceases shortly after a while. The catchments of the study area are drained to
the Awash River.

45
The Dechatu River has been gauged at the location of 09034'73" North and 41052'58" East
with river width of 80 meters with estimated slope of 0.8% close to the proposed dam site
for a few years, but only water level data are available from March 2003 to September
2010 and no rating curve has ever been constructed. In order to reconstruct the flood
history and to calculate the peak discharge of the main floods occurred in such interval,
field measurements of cross section, streambed gradient and bed material grain size were
made in the reach with the flow gauge in order to introduce these data into a simple
uniform flow equation such as the Chezyequation.

The river bed at the flow gauge site is rather regular as it is straight, has a rectangular
cross-sectional geometry and is confined between bedrock slope sides (only on the right
banks there is a small alluvial accumulation). Given the reach geomorphology, the
variability through time of the monitoring site cross-section width can be considered as
negligible. As regards the variation of the streambed elevation, very scarce information is
reported in the literature, especially for ephemeral streams with a sand bed and hyper
concentrated flow. Billi (2011), in his study on field measurement of bedload transport of
the Gereb Oda, a sand bed ephemeral stream in Tigray, Ethiopia, found that in case of
hyperconcentrated flow, streambed scouring is limited. This finding is supported also by
Powell et al. (2005) in their study using scour chains in Arizona. These latter authors
conclude that the streambeds experience little, if any, bed activity during an event.
Following these considerations, the streambed elevation at the Dechatu flow gauge site was
considered as stable.

The cross-sectional and streambed gradient were measured by a theodolite, whereas the
grain size frequency curve of bed material (Fig. 5) was obtained by the transect line
frequency by number sampling method (Leopold 1970). Since bed material includes also a
non-negligible proportion of sand, the size of the sandy particles was identified by means
of a visual comparator with specimens of all the sand fractions, arranged on phi scale, stuck
on a wooden tablet. The modal class grain of the sand in a 1 x 1 cm area near the meter
dent is considered and visually compared to the reference sieve specimens to assign it to a
specific phi class (Billi, unpublished).

Mean flow velocity (v) was calculated by the Chezy uniform flow equation

46
 V = C (RS) 0.5 (1)

In which R is the hydraulic radius, S the streambed gradient, and C the roughness
Coefficient.

C=8g 0.5 / f 0.5 (2)

In which g is gravity and 1/f0.5 is the Darcy–Weisbach friction factor.

To calculate 1/f0.5, the following equations were used: Leopold and Wolman (1957)

1/f 0.5= 1+2 * log (h/ D84) (3)

In which h, is mean depth and D84 is the grain size for which 84 % of the distribution is
finer, Limerinos (1970) and Knighton (1998), respectively:

1.16+2* log (R/D84) (4)

0.28 * ln (R/D84) (5)

Equations (3) and (4) were selected because they are among the most used in the literature
and were found to be suitable for sandy gravel rivers. Equation (5) was derived by
Knighton (1998) from a very large set of field data measured on fine gravel to sandy rivers
and reported in the literature.

In addition, the regime theory equation of Lacey (1946) was used to calculate Velocity as:

47
Figure 13. Grain size frequency distribution and material at the flow gauge reach

V=10.8 R 2/3 S 1/3(6)

And was derived for fine-grained mobile bed rivers. Finally, Grant’s equation (1997) was
selected as well since it was developed on the base of critical flow condition considerations
for mainly sand bed streams as:

V= (ghS) 0.5 4.8 (h/ D50)0.11 (7)

In which D50 is the grain size for which 50 % of the distribution is finer. No specific
equation developed for fine-grained, ephemeral streams is available in the literature;
however, Eqs. (3) to (7) that result from different approaches are based on a rather large set
of field data and proved to work satisfactorily for gravelly sand rivers; hence, they
represent the best alternative to be used on a dry land river such as the Dachau.

The river discharge was calculated for a wide range of flow levels recorded by the flow
gauge using all the criteria described above and the results averaged. By these results, a

48
stage/discharge rating curve was constructed (R2 = 0.95) (Billi, Alemu and Ciampalini,
2015)

Chart Title
2,500.00

2,000.00

1,500.00

1,000.00

500.00

Figure 14 Dechatu River Flow

49
4.17. Peak Flow Analysis

From outlier value peak flow for Dechatu river is calculated using:-

4.18. Extreme value I distribution

Extreme values are selected maximum or minimum values from a set of data. Extreme
value distribution called type I, II and III. The properties of the three limiting forms were
further developed by:-

❖ Gumbel - For Extreme value Type I distribution

❖ Frechet - For Extreme value Type II
❖ Weibull - For Extreme value Type III distribution

Extreme value I distribution is one of the distribution among EVI, i.e include Gumbel,
exponential, X2, and EVI itself and etc.

50
 Extreme value I distribution
Table 4 EVI distribution using probability of weighted moment (PWM)

YEAR Discharge Qi (m3/s) RANK (i) Qi Fi Qi Fi2

2003 2,157.60 1.00 0.02 43.83 0.89

2004 1,682.00 2.00 0.05 86.73 4.47
2005 1,276.00 3.00 0.08 105.67 8.75
2006 1,160.00 4.00 0.11 132.31 15.09
2007 1,044.00 5.00 0.15 151.71 22.04
2008 933.80 6.00 0.18 164.87 29.11
2009 800.40 7.00 0.21 166.33 34.57
2010 783.00 8.00 0.24 187.19 44.75
2011 696.00 9.00 0.27 188.14 50.86
2012 638.00 10.00 0.30 192.40 58.02
2013 539.40 11.00 0.33 179.52 59.75
2014 464.00 12.00 0.36 168.93 61.50
2015 266.80 13.00 0.40 105.47 41.69
957.00 144.08 431.49

 Qi | N = 957  Qi Fi | N =  Qi Fi 2
|N =
144.08 431.49

STD  = 519 .3457

 Qi | N =
M100 =957

 Qi Fi | N =
M110 = 144.08

 Qi Fi 2
| N =
M120 = 431.49
Where i = the rank and non – negative integer
M = moment of the sample
The peak flow is calculated as:-
   1  
X =  +  − ln  − ln 1 −  
   T   

2 Mˆ 110 − Mˆ 100
 =
Where ln 2

 = Mˆ 100 −  (0.5772 )

51
  and  are derived from sample estimate and from the theoretical one, i.e.

M̂ 100 =  +  
 = Euler number = 0.5772
T = Return period and taken as 100 years.
2 * 144 .08 − 957
  = = 965
ln 2
 = 957 – 965 (0.5772) = 400

   1  
 − ln  − ln  1 −  
   100   
= 4.6001
X100 = 400+ 295.695 * 4.6001 = 1,760.197m3/s

Standard Error

St =
s
(1 + 1.14 k + 1.1k 2 )
N

24 .0124
St = = 6.67
13
For 95 % confidence interval and N = 13
t  |2 , V = 13
Xˆ  t | 2 , V * S t (Xˆ )
Xˆ  2.037 * 6.67

Xˆ  13 .586
V = Degree of freedom
X1 = 1,760.197 + 13.586 =1,774.4
X2 =1,760.81 - 13.586 = 1,747.224

The peak flow 1,760.197 found between x1 and x2 with 95 % probability.

52
4.19. EVI–Gumbel distribution by the method of moments

Extreme value Type I by the method of moments was introduced by Gumbel. Gumbel
defined a flood as the largest of the 365 daily flows and the annual series of flood flows
constitute a series of largest values of flows. Gumbel equation is f function of recurrence
interval “T”.
The peak flow using Gumbel method calculated as:-
x = x + kx 

YT − Yn
Where K = S n = frequency factor
In which:-
YT = reduced variant , a function of T and given by
   1 
YT = −  ln  − ln 1 −   
   T 

Yn =
reduced mean, a function of sample size N and is given in table.
Sn = reduced standard deviation, a function of sample size N is given in table.
   1  
 YT =  − ln  − ln 1 −   
   T  
= 4.60
Y190 = 0.5220

S190 = 1.0565
4.60 − 0.5220
K = = 1.875
2.175
x = 965

 x −1
= 519.34
x100 = 965 + 1.875 * 519.34
x100 = 1,899.812 m3/s

Standard Error

S =
S
(1 + 1 . 14 k + 1 . 1k 2 )
N

=
24 .03205
(1 + 1.3 * )
1.875 + 1.1 * 1.785 2 = 25.476
13

53
 For 5 % of level of significant and N = 32
V = 32 – 1 = 31
t |2 , V  2 .101

xˆ  t | 2 , v * S (xˆ )
xˆ  2.037 * 25 .476 = 51 .89

X1 = 1,989.547 + 51.89 = 1,951.702

X2= 1,989.547 - 51.89 = 1,847.922
The peak flow 1,899.812 m3/s found between x1 and x2 with 95 % of probability.

4.20 Log Pearson type III distribution

This distribution is the standard distribution for frequency analysis of annual maximum
floods in the United States. It was developed as a method of fitting a curve to data, and also
justified by the fact that it has been found to yield good results in many application,
particularly for flood peak data. The mean y standard deviation Sy, and coefficient of
skew ness Cs are calculated from the logarithms of the data. The frequency factor kT
depends on return period and coefficient of skew ness. When coefficient of skew ness
given calculated the standard and error is given in the table

4.21 Log Pearson type III

Peak flow using log person type III calculated with the formula:-

Z = Z + kz  z

Where, kz is a function of skew ness (Cs) and return period and found from table.

 (Z − Z )
3
N 13 * 0 . 039831
Cs = =
( )
=0.5178
(N − 1) (N − 2)  3
Z 12 * 11 * 0.2 3

Cs = Coefficient of skew ness

N = Sample size

54
 Cs = 0.5178

KZ = F (CS, T), T = 100

ZT = Z +δZ*CS

= 2.92+0.2*0.5178=3.02356

XT = Antilog ZT = 103.02356= ➔2,864

The standard error is read from table

Table 5 skew ness

Skewness/Return 5 10 25 50 100 200
Period
1,709.62 1,851.4 2,518.26 3,126.80 3,844.15 4,691.37

Peak flow using log person type III calculated with the formula:-

Z = Z + k z  z 2.686

Where, kz is a function of skew ness (Cs) and return period and found from table.

From the table for coefficient of skew ness = by interpolation the standard error is 2.72.

For 5 % level of significance (  ) and N = 31

t | 2 V = 31
Xˆ  2.040 * 2.72
Xˆ  5.55

X1= 1,125.58 + 5.715 = 1,131.295 m 3/s

X2 = 1,125.58 - 5.715 = 1,119.865 m 3/s

Peak discharge 1,125.58m3/sfound between x1 and x2 with 95 % probability.

55
SUMMERY OF THE DISTRUBUTONS
Table 6. Summary of the peak flow for T = 100 years with 95 % of probability
Types of Peak flow (m3/s) Standard Error Maximum flow
Distribution (m3/s)
EVI 1,896.81 8.65 1,888.16

Gumbel 1,989.547 51.89 2,041.437

Log person 2,864 5.715 2,869.715

Log person type III distribution is selected. It is due to less standard error, it is widely
applicable particularly for flood peak data. Therefore, the peak flow is 1,989.547 m3/s with
an error of 5.715 and confident interval of 95 %.

56
5.0 Results and Discussion

Table 7 and Table 8 list the modeling results of the flood volume, peak discharge, their
relative errors with respect to the observation data, the difference between the modeled and
the observed time of peak appearance, and the Nash efficiency coefficients for the two
modeling schemes. Figure 3, Figure 4, Figure 5 and Figure 6 show the modeling and
observed variation in the flow

In the modeling, the parameters fc and N showed a high sensitivity in Scheme I, while the
CN showed a high sensitivity in Scheme II. Meanwhile, the Muskingum parameter K
showed a high sensitivity in both schemes. These results demonstrate that the soil types,
land use, basin topography and the river length play a major role in the flood run-off
processes.

5.1. The Runoff-Volume Models

Schemes I and II employ the Initial and Constant-rate model and the SCS-CN model,
respectively, to simulate runoff generation. According to the result of the flood volume
relative error, REv, it can be seen that the error of Scheme I is relatively small for the four
flood events, 13 June 1990, 11 July 1990, 16 August 1991, and 3 July 1993, with absolute
values less than 10%, but the error increases significantly for the events of 27 July 1982, 14
July 1987, 28 July 1995, and 08 July 1998, with absolute values greater than 15%.
However, the error of Scheme II has a small variation and a relatively small average
absolute value for most of these events.

This can be explained by the following: due to the semi-arid and sub-humid attributes, the
flood runoff is dominated by infiltration-excess under the condition of short but intensive
rainfall, or by combined infiltration- and saturation-excess under the condition of long-
lasting rainfall of various intensities. In Figure 3, Figure 4, Figure 5 and Figure 6, it can be
seen that the floods on 13 June 1990, 11 July 1990 and 3 July 1993 were caused by rainfall
of short duration, while the flood on 16 August 1991 was caused by rainfall with a strong
intensity. The relative error of the flood volume of these events in Scheme I is relatively
small and shows that the Initial and Constant-rate model is effective for modeling the
infiltration-excess runoff. However, the increase in the error on 27 July 1982, 14 July 1987,
28 July 1995 and 8 July 1998 indicates that there was additional saturation-excess runoff, a
fact that can been seen in Figure 3, Figure 4, Figure 5 and Figure 6, which shows that the
57
 four events had a long, continuous duration. Although the SCS-CN model in Scheme II
demonstrated a better performance for runoff volume prediction, it needs to be pointed out
that the SCS-CN is an empirical model, which does not consider extreme situations.

Figure 15. Lognormal, Gambel,log Pearson III

The primary objective of frequency analysis is to relate the magnitude of extreme events to
their frequency of occurrence through the use of probability distributions (Chow et al.,
1988).

The values of the highest rainfall intensity in 24 h ever recorded in the study meteostations
(Ip) is highly variable and range between 115, 166, and 170 mm/24 h at Dire Dawa
Dengego and Kersa, respectively .

The return time interval for a rainfall intensity I24 = 108 mm/24 h was calculated using
the Gumbel method for the meteo-stations with at least 32 years of data record. The results
of this analysis show that of the meteo-stations, the probability of a rainfall intensity of 160
mm/24 h is less than one in 15 years. In figure 14.

The system uses SCS-CN method, which is a method of estimating rainfall excess from
rainfall. The dominant and influencing form land use type is Acacia and shrubs land. This
shows that climatic factors, sensitivity of parameters are influenced by land use in the
watershed area thus giving results for analysis

Annual maximum daily rainfall as 6 hours duration storm peakdischarges for the 25 years
return period result values are close to each other. Ignoring thetwo extreme values, Run_3

58
 that is using the annual maximum daily rainfall as 6 hoursduration storm peak discharges
for the 50 years return period result value is chosen for thenext step analysis which is flood
delineation.

Table 6.3.5 Runoff depth and peak flood value computed at 6th hour rainfall event

Figure 16 Graph of Dechatu peak flood of 6 hour

59
Figure 17 peak inflow & peak discharge

Figure 18. Time serious Result

60
61
0.2 35

30
R² = 0.9425 25

Return Period
0.1
20

15
0
10

R² = 0.9119 5

-0.1 0
0 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 160
RF (mm)

62
5.2. Flood Flows

The floods larger than 500 m3 s-1, calculated by the methods described in Sect. 4.2.9, are
reported in Fig. 12. The flood of August 6, 2006 was the most devastating and killed 256
people. Its peak discharge was calculated to be around 1,508 m3s-1 (2.28 m3 s-1 km-2).
Alemu (2009), using the Soil Conservation Service Curve Number method (USDA 1986),
obtained a peak discharge of 1,400 m3 s-1, i.e., a value very close to that calculated by the
simple uniform flow equation. Though the SCS-CN method is empirical and developed for
small catchments in the mid-western USA, it was found to be particularly suited for
streams with negligible base flow, i.e., rivers for which the ratio of direct runoff to total
runoff is close to one, as it is commonly observed in ephemeral streams of arid and
semiarid regions (Ponce and Hawkins 1996). This is also the case of the Dechatu that is dry
for most of the time and has some water flowing only in response to individual, intense
rainstorms. Therefore, it is not surprising that the hydraulic and hydrologic approaches
used, though conceptually different, produced a very similar result. In the last decade, the
largest flood had a peak discharge of about 2,157.6 m3 s-1 (Total volume 12.6 m3/s) and
occurred on May 20, 2005. Other seven large floods, with peak discharge (Qp) higher than
1,000 m3 s-1 occurred on April 14, 2006 (Qp = 1,682 m3 s-1), March 20, 2007 (Qp =
1,044 m3 s-1), March 25, 2010(Qp = 1,160 m3 s-1), April 6,2006 (Qp = 1,080 m3 s-1),
April 12, 2007 (Qp = 1,508 m3 s-1), and March 8, 2010 (Qp = 1,118 m3 s-1). The August
6, 2006 flood was by far the worst in terms of human lives lost because it was a typical,
very flashy flood and occurred in the night, before dawn, when most of the people were
sleeping. According to eyewitness reports, in fact, the flood wave had a high velocity of
propagation and peak discharge followed shortly after the flood onset. The largest flood of
May 20, 2005 instead occurred during the day and people had time to move to safe places.

Table 7 skew ness

Skewness/Return 5 10 25 50 100 200
Period
1,709.62 1,851.4 2,518.26 3,126.80 3,844.15 4,691.37

63
 Discharge Vs Return period Discharge Vs…
5,000
4,500
4,000
Discharge in Cumecs
3,500
3,000
2,500
2,000
1,500
1,000
500
0
Return Period
5 10 25 50 100 200

Figure 19. Discharge Vs. Return Period

The river reach within the town of Dire Dawa and the subsequent floods, though a couple
of
them were of the same order of magnitude as the previous ones, had limited impact.
In order to quantify the magnitude of the August 6, 2006 flood and the largest flood of
May 20, 2005 (Demessie 2007; Alemu 2009), a flood frequency curve was constructed
with the discharge data calculated on the basis of the water level measured by the flow
gauge. The discharges with the return times of 5, 10, 20, 50, and 100 years were calculated
using the Gumbel EV method and by interpolation (Table 13). The results
in Table 13 indicate that the return interval of the largest flood of May 20, 2005 is 20 years,
whereas that of August 6, 2006 flood is about 4.5 years. The data used to calculate the
flood frequency curve, though from one side may reflect the most recent flood trend, from
the other cover a very short time interval (March 2003 to September 2015) that can make
the results obtained questionable. Given the limitation of the data available, in an attempt
to, at least, indirectly confirm the results obtained, the most renown empirical equations to
calculate the maximum flood (i.e., Q100) reported in the literature were applied to the
Dechatu. They are the equations of Pagliaro (1936, in Maione 1977); Rodier and Roche
(1984); and Griffiths and McKerchar (2008) which returned the following discharges for
Q100, 1,709.62 1,851.4 2,518.26 3,126.80 3,844.15 4,691.37 m3 s-1,
respectively. None of these criteria were specifically developed for ephemeral streams of
arid and semiarid regions.

64
5.3. Land use-land cover and changes in Dechatu watershed

Dechatu watershed is not an exception to the global processes of land use/ cover changes
though the driving forces and their relative impacts might be different. Understanding
changes and the driving forces for such changes are vital for timely, cost effective, socially
acceptable and environmentally sound interventions. With these assumptions, satellite
images of 1985, 2000 and 2014 were used to evaluate the land use/cover in the stated years
and changes happening overtime in the watershed. The 1985 satellite image has shown that
Acacia woodland had covered 84.14% of the land mass of the Dechatu watershed followed
by Shrub land (11.34%) and plantation forests (0.9117%). The coverage of agricultural
land and natural forest were 0.87 and 0.11%, respectively in the specified year. Similarly,
though there was extensive decline in coverage, the Acacia woodland (79.25%), shrub land
(14.13%) and plantation forests (1.92%) were the largest land covers observed in 2000
satellite image. In 2014 satellite image, the Acacia woodland, shrub land, and bare lands
had a coverage of 42.62, 15.02 and 1.29%, respectively, of the total area of the dechatu
watershed. Comparison of the vegetation cover for two time intervals (1985-2000 and
2000- 2014) shows that large areas of woodlands have been converted into other land uses
during 1985-2000 than 2000-2014. This is primarily because of the massive deforestation
took place in 1991 following government change in the country. The annual rate of change
was also high for the same time interval. Contrary to woodlands, agricultural land and
settlement has shown increment in both time intervals. The coverage of natural forest was
small in all satellite images. As mentioned above, the coverage of natural forest was 0.11%
of the total area of the dechatu watershed in 1985, which became almost nil in 2014. The
small increment in the size of agricultural land in 2000-2014 could be attributed to the
absence of suitable land for crop production and inhospitable climatic condition of the area
for agricultural activities at least with the currently used technologies. Shrub land, on the
other hand, had increased during 1985-2000 with annual decreased of 140.68 ha, which
might be credited to the destruction of plantation and natural forests. But in 2000-2014 the
size of the shrub land declined with annual change of -3.12%. Acacia woodland have been
converted to other land uses such as agricultural land, bare land and settlements during
1985-2014. Out of these 83.72% of the natural forest, 54.31% of plantation forest, and
67.63% Acacia woodland conversions occurred during 1985-2000. Figures 5.1-5.3 show
the land use land cover changes of the three time periods considered in this study

65
66
6. Discussion

In this study flood hazard forecasting and early warning system is developed. The system
usesSCS-CN method, which is a method of estimating rainfall excess from rainfall. Dire
Dawa, thesecond largest city of Ethiopia, has been suffering from disastrous floods in its
history. The 2006flooding was unheard-of disaster that caused severe impacts on human
lives and property(Yonas, 2015).It is used to determine flood prone areas using rainfall
event, if the amount of rainfall is greater than initial abstraction there will be surface runoff
on the area. Potential maximum retention is also related to soil and land use condition of
the area. It is potential maximum retention of water by the soil (Bansode and Patil, 2014),
initial abstraction is generated from potential maximum retention. CN value is a primary
input parameter for the SCS runoff equation (ponce and Hawkins, 1996). According
toShadeed and Almasri (2010), a high curve number means high runoff and low
infiltration, whereas a low curve number means low runoff and high infiltration. CN
converter, in the system, is used to convert the value of CN in different antecedent moisture
content conditions. Flood analysis application requires rainfall data from the user. Initial
abstraction is used to analyze flood prone areas, as described here above if initial
abstraction exceeds rainfall there will be flood and vice versa.

The coverage of natural forest was small in all satellite images. As mentioned above, the
coverage of natural forest was 0.11% of the total area of the dechatu watershed in 1985,
which became almost nil in 2014. The small increment in the size of agricultural land in
2000-2014 could be attributed to the absence of suitable land for crop production and
inhospitable climatic condition of the area for agricultural activities at least with the
currently used technologies. Shrub land, on the other hand, had increased during 1985-
2000 with annual decreased of 140.68 ha, which might be credited to the destruction of
plantation and natural forests. But in 2000-2014 the size of the shrub land declined with
annual change of -3.12%. Acacia woodland have been converted to other land uses such as
agricultural land, bare land and settlements during 1985-2014

The system also has the capability of disseminating information to community or

governmental and non-governmental organizations. To make it secured, messaging
application is available to only concerned organizations. The SCS-CN method, adapted for

67
the analysis, has been widely applied in Ethiopia watersheds in particular and the Ethiopian
highlands in general (Setegn et al., 2009; 2010)

68
7.0 Conclusion

The increase in extreme rains, paired by a marked change in land use/cover and
management practices, is considered the main factor responsible for the increased
frequency of high flash floods in the town of Dire Dawa during the last decade, though the
increase in rainfall intensity is likely playing a more relevant role.

For flash floods with a value above 12.5mm/hour, can cause high peak discharge, capable
inflict hazards depending the type of topography, soil property, effect of rainfall over
upstream determine the occurrence of flood over flood sensitive areas, the flood hazards in
Dire Dawa is favored by its topography, land cover, runoff from highland and intensive
rainfall condition.

More intensive and frequent heavy falls with relatively higher rainfall intensity were
recorded in 1956, 1977, 1981, 1983, 1988 and 1998, of wet months, slightly frequent but
heavy falls were record in April & August 2004 & 2006 (after intensive, frequent hourly
rainfalls and wet days).

There exist the likelihood of getting 50.0 mm of daily maximum heavy rainfall in every
five years cycle for the month of April and every nine years for the month of August in
Dire Dawa. For the maximum hourly rainfall intensity there exist a chance of getting 15.2
and 16.6 mm/hr in every three years of cycle for the months of August and April,
respectively.

Lack of optimum areal coverage of representative index hydromet stations defects the work
of hydromet at microclimate level.

The prospect of utilization of new technological outputs: from Radar/optical higher

resolution satellites enable us fully enhance flood and flush flood early warning systems.

Major investments based on hydromet models are essential in managing rainwater flows
and develop rainwater channels along the basin, thus, it will be able minimize flood related
hazards.

Develop dikes and flood breaks and environmental rehabilitation at strategic places is the
only remedy in alleviating flood risks.Thus the responsible bodies of Dire
Dawa Administrative Council should incorporate the flood hazard and flood risk maps in

69
the currently ongoing activities related to flood disaster, as one of the short term
intervention recommendations of the town is identification of critical flood areas

Lack of awareness on the importance of hydro meteorological information made the

problem severe during those flood periods.

70
Reference

Sorman and Abdulrazzaq 1993; Al-Qurashi ,2008 areas Pilgrim ,1988; Wheater et al. 1991;
Al-Qurashi et al. 2008 Al-Qurashi et al 2008 (Derib, 2009), (Steenhuis et. al.2009).
(Mulugeta et al. 2007; Di Baldassarre et al. 2010). Terefe et al. (2006) (Haregeweyn et al.
2016; Nyssen et al. 2010; Sultan et al. 2017; Zégre et al. 2010) (Hadadin 2013; Patil et al.
2008; Sultan et al. 2017).models (Haregeweyn et al. 2016; Jakeman and Hornberger
1993).2005 (Demessie 2007;Alemu 2009; DDAEPA 2011)(Alemu, 2015)
(Alemu, 2009).Federal Highway Administration. 1990. HYDRAIN Documentation.
Gebeyehu, Admasu, Regional Flood Frequency Analysis, Hydraulics Laboratory, Royal
Institute of Technology, Stockholm, Sweden, 1989.
Newton, D. W., and Herin, Janet C. 1982. Assessment of Commonly Used Methods of
Estimating Flood Frequency. Transportation Research Board. National Academy of
Sciences, Record Number 896.
Potter, W. D. Upper and Lower Frequency Curves for Peak Rates of Runoff. Transactions,
American Geophysical Union, Vol. 39, No. 1, February 1958, pp. 100-105. Alemu, Y.T.,
2015. Billi, P., Alemu, Y.T. & Ciampalini, R., 2015. Increased frequency of flash floods in
Dire Dawa, Ethiopia: Change in rainfall intensity or human impact? Haregeweyn, N. &
Tsunekawa, A., 1975. Reservoir sedimentation and its mitigation strategies : a case study of
the Ethiopian highlands. , (2).
Kliamenakis, A., 2011. School of Graduate Studies School of Graduate Studies. , p.112.
Available at: http://spectrum.library.concordia.ca/36191/1/Kliamenakis_MSc_S2012.pdf.
Mulugeta, M., Tolossa, D. & Abebe, G., 2017Partial, I.N. & Of, F., 2008. No Title,

71