Rainfall Runoff Flood Modelling in Nairobi Urban
Rainfall Runoff Flood Modelling in Nairobi Urban
Rainfall Runoff Flood Modelling in Nairobi Urban
By
Department of Geography
Kenyatta University
(2008)
Muli,Nathan Mweu
Rainfall-runoff flood
modelling in Nairobi
1III1I1IIIII11
06/330627
DECLARATION
This thesis is my original work and has not been presented for degree awarding in any
other University.
This thesis has been submitted with our approval as University supervisors.
Supervisors:
1. Dr. Christopher.M.Ondieki
Department of Geography
Kenyatta University
2. Dr.Wambua Kaluli
Department of Biomechanical and Environmental Engineering
Jomo Kenyatta University of Agriculture and Technology
Signature~.~'nate···:::l..·~·I·.?.I·0'8
11
ABSTRACT
The objective of this study was flood modelling in Nairobi watershed based on a
rainfall-runoff process, considering urban development and its effects on the Nairobi
watershed. The study area map was generated from a Digital Elevation Model (DEM)
developed from Survey of Kenya (SoK) topomap sheets 14811-4 with a scale of
1:50,000, georeferenced and delineated on the basis of the natural flow boundaries.
The Nairobi river Channel data base was developed through a field survey using a
digital Theodolite (Topcon 500 series). The HEC-HMS model was calibrated and
used to optimize, extract and process the input parameters of the Nairobi watershed.
The rainfall seasons in the watershed were found to be Bimodal with a precipitation
Index of 850mm. Daily historical precipitation and stream flow data obtained from
Meteorological department and MOW &1 were used to calibrate the model. For model
calibration, 3BA29 RGS (April 1SI_May3151) 1981 stream flow data was used. Data
for (April 1SI_May 3151) 1982 was used for model verification. The Manning's
roughness coefficient 'n' and bed slope of the channel were determined by
estimation. In the HEC-HMS model; initial loss, basin losses and basin transforms
were established. The model can be used to predict floods in Nairobi under the
existing and future conditions. Among the parameters, SCS lag time was the most
sensitive parameters. Its variation depicted land use changes. The Watershed has a
composite Curve Number (CN) value of 89 and 55 % imperviousness. Soil survey
data showed that the watershed was overlain by soils of Hydrological Soil Grouping
(HSG) class C of flow infiltration rate. Scenarios were simulated to answer the "what
if' question used to depict urban changes. Increased 12avemeIlts, roads and buildings
reduced the SCS lag time. As a result, the watershed developed a quick response to
precipitation. A correlation between the observed and simulated data gave a
coefficient of determination of 0.82. Flood Frequency Analysis (FF A) was carried out
using the Gumbel Extreme Value type1 (EV1) statistical method. It showed that
Nairobi had a 43 % chance of flooding after every 2.33 years. Development in the
watershed increased the imperviousness from 55 % to 60 %. The impact of this was
an increased mean annual flow from 50m3/s to 345m3/s. This was a 600 % increment
and the present drainage cannot cope, thus making the study area to be in danger of
being flooded. The Muskingum-Cunge-8-Point Routing method was used to route the
flood along the Nairobi river channel. Flood waves generated recede by 3 % and take
2 hrs and 45minutes to travel from inlet to the outlet. The Questionnaire data was
analyzed using the (SPSS) 11.5. The stakeholders were of the opinion that; the
drainage system was poorly maintained, inadequate and it required rehabilitation.
This was one of the main causes of flooding in Nairobi. The HEC-HMS model
simulations generated threshold peak flows which can be used for planning, design of
cross drainage works and other storm water drains in Nairobi city and other Kenyan
cities. It can be customized for flood prediction in other urban watersheds. The city
requires surface and subsurface drainage to remove water pools after long and short
duration storms. There is need for structural and non structural measures to mitigate
for floods and to install the Best Management Practices (BMPs) to reduce flooding
potential in the city. Drainage rehabilitation, fresh planning and design are required
for the city to enable it cope with the increased development.
III
Dedication
"Unto Him who cares for me, I dedicate all this work. He gave me a hope like no other
could give .... "
IV
Acknowledgements
My special gratitude goes to the chairman of Geography Department; Dr.Ondieki and all
the members of lecturing staff. I am especially grateful to my supervisors; Dr. Ondieki
and Dr. Kaluli for their guidance. I would like to thank all the people who assisted me
directly or indirectly with my studies. I am indeed grateful to the Principal Kenya
Polytechnic University College for allowing me time and resources during my studies. I
am thankful to Mr. Mutune of Kenya Polytechnic University College, Mr.Wambua from
Katumani (KARI), Ministry of Water and Irrigation (MOW&I) research division,
Meteorological Department, Nairobi City Council; City Engineer's Department,
Consulting firms, NGOs and other data Centers for their valuable information. To the
PolyGIS staff in Kenya Polytechnic University College; I say thank you for the time you
spared me. Special thanks to my family for being there for me. Without you, it could not
have been easy for me. In one way or other you made my life comfortable throughout my
studies. My cherished friends Francis and family, Joash, Titus, Kaliti, Kasimu, Japheth,
Nderitu, Paul, Ruth and many other close persons. You all deserve my sincere gratitude
for the concern, moral support and the encouragement you offered me. I would like to
express my humble appreciation to my friends and classmates EUy, Mwangi, Sungu,
Maunda and the late Wanyonyi for their encouragement during our studies.
v
TABLE OF CONTENTS
Declaration .i
Abstract .ii
Dedication .iii
Acknowledgement .iv
Table of contents v
List of Tables vi
CHAPTER 1 1
1.0 INTRODUCTION 1
1.1General Introduction 1
1.1.1 Background study of the Sub-catchment area 3
1.1.2 Location 3
1.1.3 Topography 5
1.1.4 Surface Geology and Soils of the study area 6
1.1.4.1 Geology 6
1.1.4.2 Soils 7
1.1.5 Climate 9
1.2 Statement and Research Problem 11
1.3 Justification 11
1.4 Research Questions 12
1.5 Hypothesis 12
1.6 Research Objectives 12
1.7 Significance of the study 13
1.8 Limitations and Scope 13
CHAPTER 2 14
2.1.2 Urbanization 15
2.2 Hydroc1imatological Patterns in Nairobi Sub-catchment 19
2.3 Impact of Gardening .22
2.4 Drainage systems in Nairobi Sub-catchment area .23
2.4.1 The existing drainage network in Nairobi Watershed 24
2.5 Role of Mass Wasting in Urban Flooding 25
2.6 Flood Estimation Methods 27
2.7 Flood Frequency Studies 28
2.7.1 Flood Frequency 29
2.7.2 The Gumbel Extreme Value distribution type 1 (EVI) 29
2.7.3 Peak Flows 32
2.7.4 Standard Error (SE) 33
2.7.5 Analysis of Annual Maximum Series 33
2.7.6 Fitting the Extreme Value Distribution 34
2.8 Flood Hydrograph Generation 35
2.9 Modelling 36
2.9.1 IHACRES Model. 37
2.9.2 The HEC-HMS ModeL :.38
2.9.3 Modeling Runoff Losses 43
2.9.4 Precipitation Methods 45
2.9.5 Muskingum flood Routing Model. .45
2.9.6 Muskingum-Cunge-8-point Method .48
2.9.7 Flood Mitigation Measures .49
2.9.8 Best Management Practices 51
CHAPTER 3 53
3.0 METHODOLOGy 53
3.1 Data Acquisition 53
3.1.1 Secondary data 54
3.1.2 Primary data 56
3.2 Model selection and application 57
3.2.1 Model Selection 58
3.2.2 Model Application 59
3.2.3 Watershed RunoffProcesses 61
3.2.4 Rainfall-Runoffprocess 62
3.2.4.1 Model component creation 62
3.2.4.2 Basin Model. 62
3.2.4.3 Meteorological Model. 63
3.2.4.4 Control Specifications 64
3.2.4.5 Hard Constraints 65
3.2.4.6 SCS Curve Number (CN) Loss 66
3.2.4.7 The SCS UH 67
3.2.5 Model Calibration and Verification 68
3.2.5.1 Model Calibration 69
3.2.5.2 Model Verification 72
vu
CHAPTER 4 82
CHAPTERS 113
REFERENCES 117
Vlll
APPENDICES 125
List of Figures
Figure 1.1: Location map of the sub-catchment and the delineated study area 3BA. .4
Figure 1.2: Spatial distribution of gauging stations 5
Figure 1.3: Annual rainfall for the sub-catchment area 10
Figure 2.1: Distribution ofland use in Nairobi Watershed 15
Figure 3.1: HEC-HMS Run-offprocesses 62
Figure 3.2: Muskingum-Cunge-8-point configuration 78
Figure 4.1: Nairobi Sub-catchment schematic 83
Figure 4.2: Calibration graph based on (Aril 1st - May 31 st) 1981 data 84
Figure 4.3: Scatter Graph of Simulate and observed hydro graph in 3BA29 RGS 86
Figure 4.4: Residual graph of simulated and observed hydro graph 3BA29 RGS 88
Figure 4.5: Objective function graph 89
Figure 4.6: Model verification using (April 1st -may 31 st ) 1982 data 90
Figure 4.7: Correlation between observed and simulated flow 91
Figure 4.8: Hydrograph for 1981 annual stream flow data at 3BA29 RGS 92
Figure 4.9: Rainfall hyetograph for the long rain season 93
Figure 4.10: Scenario when imperviousness is 60 % 94
Figure 4.11: Scenario when imperviousness is 65 % 95
Figure 4.12: Rating curve for 3BA29 RGS 96
Figure 4.13: Frequency curve based on 3BA29 RGS stream flow data 97
Figure 4.14: Probability of exceedance on 3BA29 RGS 98
Figure 4.15: An inflow Hydrograph generated at 3BA29 RGS 99
Figure 4.16: An outflow graph at Kariobangi south bridge 100
Figure 4.17: Phase-speed of routed flood from inlet to outlet point.. 101
Figure 4.18: Nairobi floodwave profile from 3BA29 RGS to Kariobangi south bridge. 103
Figure 4.19: Level of investment to improve drainage 110
Figure 4:20: Distribution of Highway drainage system in Nairobi 111
x
List of Tables
Table 2.1: Infiltration rate for various soils (after Skagg and Khaleel, 1982) .44
Table 3.1:.Limits of hard constraints used in calibration 65
Table 4.1: Calibration parameters 85
Table 4.2: Modelled results after calibration 87
Table 4.3: Observed results after calibration 87
Table 4.4: Analyzed stream flow magnitudes and return periods 97
Table 4.5: Flood Routing results from 3BA 29 RGS 102
Xl
List of Plates
Plate 2.5: Blocked Culvert on Isiolo Road in Nairobi Industrial area .21
Plate 4.2: Garbage heap next to kiosk on Commercial Street Industrial area l05
Plate 4.6: Level crossing at Tetra pak on Enterprise road Industrial area l 09
Plate 4.7: Floods turns Dunga road in Industrial area, into an open channel.. 111
Plate 4.8: An office property off Enterprise road inaccessible due to flooding .l12
Xll
CHAPTER 1
INTRODUCTION
Urban development worldwide affects the surface runoff due to the spread of
settlements; and other auxiliary features such as roads, pavements and air fields.
During periods of high intensity storms, there is little doubt on what happens on the
city streets with respect to drainage (Bauer, 1969, Savini and Kemmerer, 1961, Hall,
1985). Due to urban development, large areas of high infiltration rate have been
reduced in most urban areas. Falling precipitation is caught on roof tops, paved roads
and then passed through designed drainage systems (Ward and Robinson, 1990). The
drainage system may not be well maintained for efficient disposal of storm water into
Nairobi River, Mathare, Gitathuru, Ngong Rivers and other streams in the sub-
and lack of maintenance are two factors which when combined increase the flood
peaks by between two and eight times in some of the cities in the U.S.A. Espey et aI,
(1966) also found that, urban development resulted in discharges which were between
100 - 300 percent greater than those in underdeveloped rural areas. According to
Pereira (1973), floods are symptoms of land misuse, but not all floods are due to
human mismanagement.
In the Nairobi watershed, it is of interest to note that apart from peak flows, urban
development has affected water quality and a wide range of other hydrological
variables. Human factors are responsible for the apparent increase in severity of
floods witnessed in Nairobi in recent times. In Nairobi area, heavy rain occurs in the
2
evenings spreading through to early morning during the short rains in November. Part
of this rain results from effects of the highlands close by after 5:00 pm (Thompson,
1957). Nairobi major storms occur between November and May, but some storms are
recorded in the 'dry' months of January and February as well. Many factors combine
to increase peak flows and magnitudes of floods, partly this could be due to rapid
movement of runoff or quick flow (Ward, 1975). Knowledge of the flow magnitudes,
volume and peak stages in the existing drainage network, contribute greatly to
Floods are defined as high rates of discharge that lead to inundation of land adjacent
to rivers and streams, damaging property and interfering with human activities and
settlements as a result of quick flow rather than base flow (Reddi, 1992). Reduced
infiltration rates, changed drainage network, efficiency and increased rainfall intensity
due to climatIc chcmg have contn·6oted to increased flood seventy in many Afiican
cities (Imbamba, 1990). Villages in Nairobi especially slum areas and to some extent
some of the up-market residential estates have always been affected by floods
(Jenkins, 2001). A study carried out showed that five African countries including
Kenya are vulnerable to climatic change, which is characterized by sea rise and
increased flooding. This is known as La Nifia which according to Ibe and Awasike
(1989), is related to periodic African drought, while El Nifio is associated with excess
rain. According to Wisler and Brater (1959), hydrologic and hydraulic analyses
The watershed area is located in the upper part of the Athi River drainage basin three.
The greater Athi basin occupies an area of 66,837 knr'. The basin has a mean annual
rainfall of 740mm (MOWD, 1992). In essence, the Nairobi watershed is part of the
upper Athi drainage area, which represents only 3 % of drainage area three.
1.1.2Location
Nairobi sub-catchment area is part of the upper Athi drainage basin three. It is
situated between Embakasi-Athi plains to south and east, and the broken landscape of
the Kenya highlands in the north and west. The larger upper Athi sub-catchment from
the farthest point to the lowest outlet point covers a total area of 1942 square
kilometers. It lies between 360 24' and 370 4' east Longitudes and 10 9' and 10 28'
Latitudes to the south of the Equator (MOWD, 1992). The upper Athi drainage area
was created from a Digital Elevation Model (DEM). On the map of Kenya at a
smaller scale, the geographical position of the sub-catchment was delineated. The
actual study area; watershed 3BA, was further delineated from the larger upper Athi
sub-catchment (figure 1.1). It shows the entire natural drainage network in the sub-
catchment. On the watershed area map; the rain gauge stations, regular gauging
stations, shopping centres, roads and natural drainage networks were georeferenced.
District boundaries, watershed boundary and the study area were marked as indicated
(figure 1.2).
4
Fig. 1.1: Location map of the upper Athi sub-catchment and study area 3BA
5
•••
1:250000
1.1.3 Topography
Nairobi watershed area is divided into two physiographic land forms; the western and
northern rising between 1905m to 1975m above mean sea level (amsl) forming parts
of the Kikuyu Plateau. The Eastern and lower parts to the South East are generally
low and flat, about 1600m above mean sea level (amsl) forming the Athi Plains. It
exhibits a low gradient of approximately 2%. When it rains, much of the land
experiences ponding which later culminates into flooding. To the West of the
escarpment, very few areas exhibit plateau characteristics. This landscape changes
further to the west of Dagoretti, Karen, Kikuyu, Kabete and the northern suburbs of
the city centre with its undulating valleys, streams and rivers, most of which are
perennial. These areas are characterized by sharp gradients and soils of low
6
infiltration rates. The watershed has numerous streams and their tributaries which
join the Athi River to the South and East. The rivers and their tributaries form a
Dendritic pattern (tree branch-like pattern) (figure 1.1). Nairobi River is one of the
main rivers in the 3BA watershed area. It originates from Ndindiri swamp in Kikuyu
The geological formations in Nairobi watershed have highly influenced the soil
patterns, land forms and the natural drainage network (Sikes, 1939). Deep soils and
gravels of Quaternary age cover all the formations. The drainage area which is
characterized by the rift faulting is endoreic. Most of the runoff seems to collect in
1.1.4.1 Geology
Nairobi sub-catchment area has a history of successive volcanic activities, with layers
of tuff and lava which have been subjected to varying degrees of weathering. The
area has layers of black tuff, normally between 2m and 10m thickness. This is
overlain by expansive clays with some phases of yellow softer tuff interposed and
underlain by agglomeratic tuffs and lava. Under the black clays are layers of tuff
underlain by phonolite lava to depths between 10m to 25m. The watershed is an area
where volcanic activity had dominated the geological history, and controlled the
Dagoretti-Karen area to the East of Nairobi and extent North of Kiambu to Githuguri.
7
The rock system in Nairobi area mainly compnses of succession of Lavas and
Pyroclastics of the Cainozoic age overlying the folded Schists and Gneiss of the
Precambrian Basement system. The western part of the sub-catchment area has been
affected by faulting at least in three different periods. The Eastern half of the sub-
catchment area, which is an erosion plain, gently dips to the South from an average
elevation of 1650m to I 540m at the confluence of the Athi and Nairobi Rivers. It is
also underlain by almost horizontally lying tertiary rocks. The north merges into
young tuffs and lavas (Gaevaerts, 1964). During the recent volcanic activity, Nairobi
seems to have had two main drainage systems; one line running in the general
direction of Moi Avenue, which was probably the old line of Nairobi River that has
changed its course to the present one. The other one runs from Parliament area
towards the Railway Station. The old watercourse is indicated by the present out
crops of the upper and middle tuff of the city centre. The existing ground surface
1.1.4.2 Soils
Soil patterns in Kenya are very intricate because of their striking differences in
altitude, landforms, shape, stability, age, geology and climate (KSS, 1980). The soil
exploratory survey incorporated all the information available on Kenyan soils upto the
end of 1979. The soil survey reveals the complex relation between landforms and
geology in many parts of Kenya (Sombroek et aI., 1982). The central parts of the
8
Nairobi watershed consists of LI1 (LBvp) type of soil. These are limestones, calcitic
mudstones, in plateaus and high level structural plains that are flat, gently undulating
and sloped at a rate less than 8%. This type of soil occupies almost two-thirds of the
watershed area. It is imperfectly drained, very deep, dark greyish, brown to black,
bouldery and stony with cracking clay. Along the Nairobi River Valley, alluvial
deposits from various sources are found. This soil is categorized as A18 (AAje). It is
a complex of well drained to imperfectly drained, very deep, dark greyish brown to
Fluvisols). The extreme western parts of the sub-catchment area are overlain by soils
ofR3 (RBne) type extending from areas ofNgong Hills all the way to Limuru. These
are well drained soils, extremely deep, dusky red to dark reddish brown, friable clay
with inclusions of well drained, moderately deep, dark red to dark greyish brown,
friable clay over rock, pisoferric or petroferric materials(Eutric Nitisols; with Nito-
There are isolated deposits of M5 (Mvbc) in Ngong area which were well drained,
shallow to moderately deep, dark reddish brown, friable, humic, rock and stony, clay
loam(Humic Cambisols, Rocky and partly Lithic Phase). Limuru area has scanty R3
(RBne) which are Eutric Nitisols; with Nito-chromic Cambisols, and Chromic
Acrisols, partly of pisoferric or petroferric phase. The Nairobi city center has patches
of well drained, shallow brown, firm gravelly clay, with a stony to bouldery surface
(Chromic Cambisols, Lithic and Boulder) mantle phase. These are tiny patches of L9
(LBhv2 soil imperfectly drained, very deep, dark greyish brown, firm clay (Verto-
luvic Phaeozems; with Eutric Planosols in the city center (KSS, 1980). Ngong area to
the extreme west has some isolated spots overlain by the complex well drained soils
of type F9. These are deep, reddish brown to very dark greyish brown, firm, sandy
9
loam to clay soils. In some cases, the soils are moderately calcareous (undifferentiated
Luvisols, Luvic Phaeozems and Chromic Vertisols). In the areas around Athi river
Town, there were patches ofV2 (Vc2) existing which are a complex of well drained
to imperfectly drained shallow to moderately very dark, greyish brown, firm, slightly
to moderately calcareous, rock stony or gravelly clay. They overlie the southern and
the south-eastern parts of the sub-catchment area. These soils combined have a
1.1.5 Climate
Nairobi watershed area is situated in the ecological zone five, which is classified as
sub-humid. The Western and the Northern parts border the high potential highland
areas. The annual mean rainfall of Nairobi watershed is 900mm, which is distributed
over long rains from mid March to May with the short rains from mid October to mid
large diurnal range; this is due to high temperatures during the day and low
temperatures in the night. The mean annual temperatures have been found to be 23° C,
characterized by small annual and large diurnal range due to high temperatures during
the day, and low temperatures during the night. Nairobi watershed experiences two
rainy seasons; short and long rains, which usually occur in November-December and
900mm. This amount of rainfall decreases from the Kikuyu escarpments to 250mm in
the Athi plains. Throughout the year, the watershed experiences both seasonal and
unseasonal storms. This was clearly exemplified in the rainfall hyetographs based on
the rainfall data for Dagoretti, Muguga, National Agricultural Laboratories (K.A.R.I),
10
Moi Airbase, Wilson Airport and JKIA Rainfall Stations. It is apparent that; at certain
times in the year, storms which are out of season and of low magnitudes are
1200~------~~~~~~~~~~--~~~
1000 +---------~----------------~------~
~ 800 +-------1..----..:...-.--..:...-....;........,--..:...-.----1
-~ 600 +----------tft--rl------.:.......-.:.......-o.--~~----___;
I:
~ 400 +---------~~~+-----~----~~~~~----~
200 +-------11-
o ~itiIn;blnml1lllt
The March to April rain season is characterized by the highest rainfall peaks recorded
of close to 1200mm (Lamba, 1994). The highest precipitation peaks are noted at the
beginning of the season, and sometimes towards the end of the season. The months of
November to December marks the short rain season, with low rainfall peaks of close
to 250mm on the average. During this period, the highest peaks are recorded around
the middle of the season. From the records available, this watershed area has very
few months in the year without precipitation. The Dagoretti meteorological Station is
found in the central region of the watershed. It recorded higher values of precipitation
compared to Muguga Station, Jomo Kenyatta International Airport and Moi Air Base.
January and September appears to be the only months with scanty amounts of rainfall
according to the records. Regardless of the direction of the storm, topography and the
11
shape of the watershed, the spatial rainfall distribution seems to be uniform over the
entire watershed.
Nairobi watershed experiences floods quite often; therefore, this has increased the
need for flood prevention over the years. The El-Nifio floods and other subsequent
and other facilities in Nairobi (Ali, 1997). At present there is inadequate knowledge
on factors, which contribute to flooding in Nairobi, and the impact thereof is not well
1.3 Justification
Flooding in the Nairobi watershed is a hazard. When floods occur, they cut off
certain areas of the city for many hours. Roads and bridges are usually submerged
and people get trapped in the city center (Jenkins, 2001). There is need therefore to
enable proper planning and provision of mitigation measures (Moriba, 2006). Results
of this study will contribute to the advent of well organized mitigation measures in the
Nairobi watershed area, and Kenya as a whole in other towns with a similar problem.
(a) What are the peak runoff rates from the Nairobi watershed areas during rains?
(b) Will the future development of the open land areas of the watershed increase or
(c) What will be the magnitudes of flow/volumes and peak stages with the existing
drainage structures?
(d) What interventions can be put in place to protect city residents against floods?
(e) What is the state of the existing natural and artificial drainage system in Nairobi?
1.5 Hypothesis
Nairobi watershed has some engineered and natural drainage systems. Because of
poor maintenance and inadequate drainage systems, the city experiences floods at
times when there are high intensity storms in season or off season.
1.6Research Objectives
The main objective of this research is to study flooding in Nairobi watershed area
for the purpose of flood mitigation and urban storm water management. The specific
objectives were:
(b) To determine the magnitude and frequency of floods in Nairobi for the
existing records.
(c) To route the floods and establish the dynamics of the flood wave for selected
13
.(c) Use mode led results to improve the current system and management practices
This study was limited to watershed 3BA in the Nairobi sub-catchment area, which
forms part of the headwaters of the Athi catchment. There was a problem in obtaining
continuous and accurate data, since most gauging stations were no longer functional.
Most of the data centres complained of lack of funds and field personnel who can
collect the data. The scope of the study was to establish Nairobi flood magnitudes
and frequencies with the continued urban development and simulate a model that can
be used to predict and route floods along the Nairobi River channel, with the aim of
CHAPTER 2
LITERATURE REVIEW
The watershed consists of communities, people, soil, water, fanning activities and
industries. These are components which exist and interact with their bio-physical and
dynamic and adaptive to systems, which change due to natural and human induced
function around a dynamic equilibrium state or multiple stable states. People respond
or rules to force the system to move towards another state (Holling, 1986). This
There are so many dynamic changes taking place in the Nairobi watershed due to
settlers and demands for additional area for habitation, manufacturing and agriculture.
Channels that are only occasionally occupied by the streams and low bottom lands are
filled and built upon. Bridges are built without provisions for extra floods. Normal
channels are restricted. This causes floods to rise in height creating increased velocity
needed to carry water through the reduced channel. Sometimes the channel banks
15
cave in, carrying stumps and trees into the stream. The channel banks are used as
Different land uses compete for space in the airobi watershed. This had resulted to
environmental issues in the watershed, a factor that led to land degradation. The main
watershed had shown that; 69 % of the land was under urban development, 30 %
under agricultural use and 1 % left under other urban uses (figure 2.1).
• Residential
• Recreational
o Institutional
o Industrial
0.40%
• Corrmercial
3.50% • Agricultural
11.00%
2.1.2 Urbanization
showed that hydrology of urban areas is quite complex (pereira, 1973). In developed
towns and springing up of new ones was inevitable due to the ever-increasing
construction of roads, roofs and other amenities, such as parking lots that transforms
the existing regime. Increased water use has had implications on the water supply
suburban areas in recent decades caused a complex merging of social, economic and
physical problems (Adede, 1988). The interrelationship of man, his use of land,
development. There has been relatively little study to date on the effects of the urban
man upon natural hydrologic conditions (Savini and Kemmerer, 1961). Cities
expenence low ground water levels, accelerated land erosion, increased stream
sediment and aggradation, overload of sewers, and increased local stream flow, flood
damage, decreased infiltration, waste and water use (May and Rinks, 1988). On
January 13th 2001, floods in Nairobi killed four people in different parts of the city.
On January is" 2001, serious flooding brought Nairobi to a stand still, where Roads
became impassable; the streams which flow through the city became brown, swirling
Plate2.1 Flooded ale Shapara Avenue in Nairobi South C (January is" 2001)
(Jenkins, 2001).
Among the worst affected areas were the sprawling slum areas, which surround the
Nairobi, most residents woke up to flooded estate roads; making it impossible for
Plate 2.2: A flood in Kitsuru estate in Nairobi May 9th 2002) (Ark, 2002))
18
On May 5th 2003, floods left Nairobi without piped water. Up to 1 million residents
of the Kenyan capital Nairobi faced water shortages after floods damaged parts of the
Sasumua dam. In some parts of the city, children stopped attending school to help in
looking for water as the city taps went dry (Karobia, 2003). Others withjerrycans and
wheelbarrows went about looking for places to buy water in some parts of airobi,
Plate 2.3: Nairobi experiences water shortage after floods (Karobia, May 5th 2003)
Scenarios of city residents getting rain trapped in the estates in the morning arriving at
work late or getting trapped in the offices after work due to flooded streets became
common in Nairobi (Karobia, 2003). The Standard News Paper of May 9th 2006
reported that: "after a heavy storm in Nairobi, floods threaten to destabilize 150,000
city residents". Unusually long traffic jams on the flooded streets became common
phenomena in the city (plate 2.4). Roads became impassable as some sections of the
Plate2.4: Flooded section of Lusaka Road Industrial area causes a traffic jam
The airobi sub-catchment area has numerous tributaries of the Athi River. The
rivers originate from Kikuyu highlands and gong hills and flow to the south-east
two spates, November to December and March to May. They originate from north,
moving east and south across the city. The major storms of the 'long rains' (mid-
March to May) commence in the early evening. There are two basic types of storms
during the rain seasons; one type of storm occurs during the night and early morning,
and the other which starts in the evening. Morning storms which usually start from
OO:50hrsto 11:59hrs are more prolonged and less intense than the evening ones,
which are usually storms starting 12:00hrs to 23:59 hrs) (Charania and Charania
1975). On average, the main storms last for 522 minutes. They have an average fall
of 73rnrn, and a mean hourly intensity of rainfall of 27.7mm/hr (Forsgate and Grigg,
1975). The evening storms last for 187 minutes on average. During this period,
20
55mm of the rain falls with an average hourly intensity of 41. 1mm/hr. The morning
According to Thompson (1957), the largest and the most intense storm recorded in
this watershed occurred in the evening and approached from the North. Air reversals
associated with the drainage of cold air from Aberdare Mountains made the evening
storms approach the city from North and East (Findlater, 1969). A major low level air
current near the Indian Ocean during the northern summer influences the rainfall
Saha, 1983). In the watershed, nearly 50% of the total rainfall occurred in 10-15% of
the total duration, though sometimes in the past, unique storms had pounded the
watershed non stop for more than 12 hrs. It is understandable that, floods hit Nairobi
Nairobi watershed has a precipitation index of 850mm. The central and eastern
sectors are characterized by low, but effective rainfall depth. Dagoretti Station for
150mm soil retention capacities. The Potential Evapo- Transpiration (PET) for the
watershed is 1236mm. In the watershed, the river gauging stations are operated by
the Ministry of Water and Irrigation (MOW&1) in watershed 3BA, though at present,
most of Stations are vandalized and are no longer functional. Report by MLRR&WD
1998, revealed that Nairobi has a soil permeability coefficient ranging between 3 to
According to Marshall (2000), El Nifio and the subsequent heavy rams of 1999
affected the poor populations living in the slums, as some of them were squatters
along the flood plains and landslide prone areas. Businesses were seriously
Kenyans were not adequately prepared for this and had no facilities in place to
cushion them against the adverse effects. Floods in airobi usually follow the major
drainage patterns of the watershed, over loading the existing drainage ways. As the
storms recede, the natural and artificial drainage ways are eroded and wasted away,
causing land slides and slumps (Wanaanen and Limerinos, 1997). Due to the
consequences of erosion and sedimentation, drainage takes the form of muddy streets,
clogged storm sewers and dirty office corridors (plate 2.5). This leads to an increase
Zigg, (1940) found out that the rate of erosion increases with the length of the slope.
Further accelerated erosion occurs at the edges of tarmac roads without kerbs. The
22
big storms experienced in Nairobi were associated with large raindrops, which can
cause dislodgement of sediments on bare surface (Ongwenyi, 1985). These have upto
8 times the erosive capacity of the runoff. Erosion of tarmac roads is literally a "pot
hole" menace that results from poor drainage, allowing water to stand on the roads for
a long time (Plate 2.6). Engineers refer to this problem as "bath-tub" menace as it
Urban Agriculture exists throughout Nairobi city on both private and public land.
Growing of crops in urban areas is an important survival strategy for the urban poor
(especially those without rural land holdings) as this reduces the amount of income
expended on food. The Legal status of urban agriculture remains unclear to date.
Urban cultivators have a way of composting waste in the city for their agricultural
23
use, the composting efforts have a negative impact on flooding in the city; because
when it rains, the footpaths become muddy (Peters,1998). The two types of gardening
(ii) Riverine gardens planted with maize, potatoes, beans and nurseries.
In both cases, the soils are left bare and then later transported to the pavement and
rivers during rainstorms. Urban watersheds contribute 30 times more sediments than
natural watersheds (Chen, 1974). McPherson (1977) found out that, runoff from
unpaved areas to the road pavements has made sediment contribution more acute,
According to Krhoda (1986), there are two types of drainage systems in Nairobi
watershed, namely; the natural and artificial channels. The natural channels include;
Nairobi, Gitathuru and Mathare Rivers and their tributaries. The headwaters of these
rivers are in the Uplands area of Limuru. The major artificial channels include, the
canalized branch of Mathare River extending from Loresho to City Park. Another one
runs from Kilimani via Uhuru Park, Railway Club and Golf course joining the Ngong
River near the Dunga road bridge, before Matermisericodiae Hospital. These add to
the numerous lined and unlined storm drains within the city. Diniz and Moore (1974)
discussed the sequence of changes in sediment yield and channel stability as most
Nairobi watershed has both natural and artificial drainage network. Krhoda (1986)
observed that, most of the catch pits on the road and the Kerb gratings had not been
serviced for a long time (plates 2.7). Garbage swept from the street is usually
deposited into these catchpits. This has led to serious drain blockages, which turns
the streets into shallow open channels whenever there is a storm. The maintenance of
In some of the cases, drainage inlet gratings had been covered by grass and other
growths of vegetation. This is a common scene in Nairobi and some of the major
towns in Kenya, whose drainage systems seemed, neglected and lacked regular
The surficial geology of Nairobi includes volcanic ejecta, alluvium and fills that are
areas aggravates the magnitude of flooding. Increased flood risks arise from soiled
streams and drainage ways, clogged storm inlets and blocked drains that are as a
result of mass movement (Senga, 1983). Cases of solid waste dumping are also
common in the streets of Nairobi. This has narrowed of some streets due to piling up
of solid waste, culminating in blockage of the existing drainage facilities (plate 2.9).
26
From field investigations (plate 2.10), it was noted that; some of the slum
constructions had encroached the reserved space on the roads and major streets which
were meant for drainage. These slums have no proper drainage and sanitation
facilities; any time there is a storm, garbage is washed to the streets creating filthy and
unaesthetic conditions which are unbearable to both motorists and pedestrians. This
An abnormal quantity of water arising from heavy rainfall causing channels or rivers
to overflow their banks, inundating low lying areas is considered a flood (Reddi,
1992). Several methods are used for flood estimation. Some of the methods used for
sophisticated among these methods, and it requires a lot of data, such as; standard rain
gauge data, continuous rainfall data, intensity-frequency duration data and temporal
patterns of rainfall (MOWD, 1975). For flood studies, precipitation data is required
since rainfall variability in space and time should be considered. The Transport Road
Research Laboratory method of flood estimation can be used where the volumetric
According to Kenrail consultants (1998), flood prediction models can be used for
flood estimation in Nairobi. Kenya Railway Corporation has used some of these
Both the TRRL and Kenya Railway methods use the Area Reduction Factor (ARP).
For the entire country, ARP was based on Nairobi rain gauge stations network
28
covering 1200 km2 (Fiddes, 1977). According to WMO (1969), the (ARF) equation
suitable for Nairobi watershed which has both humid and semi arid-climatic
I I
Where
T = Duration in hours
A = Area in km2
All these methods can be applied in the Nairobi watershed for flood studies, but
limitations exist due to their demand for certain catchment descriptors which are not
easy to establish. The four methods used by Kenya Railways can be employed
(b) Sizing and designing water control measures, if target exceedance level or
reliability is specified.
To meet the objectives of a flood frequency study, peak flows, stages and volumes for
specified annual exceedance probabilities are required. The flow and stage frequency
29
curves are often used for flood-damage calculations. Flow volumes are used for
2.7.1Flood Frequency
Frequency analysis made use of observed data in the past in order to predict future
flood events and their probabilities or return periods. Adequate and accurate data
must be used if frequency analysis is to provide useful answers. Flood data may
consist of either annual series or partial duration series. The choice depends on the
storms, floods and droughts. The magnitude of an extreme event is inversely related
to its frequency of occurrence, but very severe events occur less frequently than the
relate the magnitudes of extreme events to their frequency of occurrence through the
(Parodi, 2005).
Three models commonly used for extreme value analysis are Gumbel, Frechet and
formula is biased and plots the largest values of a sample at too small a return period.
Gumbel is easier to work with since it requires only location and scale parameters,
while Weibull and Frechet functions require location, scale and shape parameters
From observed or simulated rainfall data, when valid watershed models are applied,
flood frequencies can be estimated, according to Baker, (1988). Using the Gumbel
extreme value type I, the general formula for probability density function of the
x-p x-}J
P
f(x) = ~e-P e-e-- (2)
distribution. From this, the standard Gumbel distribution (maximum) reduces to:
This equation can be used to make a plot of p.d.f for the maximum cases.
The probability that the event can be equaled or exceeded can be obtained from:
P = 1- «:' (4)
A smooth curve of the probability density function (p.d.f) can be obtained if the data
series is imagined to be infinitely large in number and the class intervals are infinitely
small. The area under the probability curve is a unit and is obtained from;
r p(Q)dQ =1 (5)
31
The probability that an annual maximum flood Q lies between a and b is given by
p = !p(Q)dQ (6)
For any given magnitude, X, it can be shown that an annual maximum event equals or
P(X) is the probability that annual maximum event equaling or exceeding X in any
given year, since it is the relative proportion of the total number of maxima that have
r
ThenP(X) =- (10)
N
1
Thus P(X) = -- (12)
T(X)
It follows that;
1 1
T(X) = -- = (13)
P(X) 1- F(X)
T(X) -1
and F(X) = (14)
T(X)
Peak flows and their frequencies are used for design and planning in most cases, to
provide solutions to problems related to engineering for peak flow analysis. The
maximum value for each water year is selected and arranged in a decreasing order
a and b are the two parameters related to the moments of population of Q values.
Defining the first moment (the mean) by JLa and the second moment of variance (}~l '
1l
b =r=r: (17)
(J"Qv6
and:
!1Q = N:t
1 N
Q; (Sample mean) (18)
1 N 2
o-~= S~ = ~ L (Q, - Q) (Sample variance) (19)
N -1 ;=1
But,
1
x =a-blnln [T(X)
T(X)-l 1 (21)
Substituting the parameters a and b with the sample mean Q and standard
The Standard Error (SE) is a measure which would indicate how precise the
prediction is. It gives an idea about reliability and precision of the sample. The
smaller the S.E, the greater the uniformity of sampling distribution hence, the greater
the reliability of the sample. Conversely, the greater the S.E, the greater the difference
between the observed and expected frequencies. In such a situation the unreliability of
SE(X)
~
= SQr
-L1 + 1.14K(T) + 1.1O(K(T))
2]' (24)
N
The peak flows which are usually in m3is are arranged in a decreasing order of
34
magnitude, with the second column showing the rank position. The probability of
exceedance P(X) is then calculated for each value, X. According to plot position,
there is a formula devised to overcome bias by the fact that; when N is not large, ..!..-
N
is not a good estimator. Among the several formulae in use, the best is owed to
the unbiased mean value of probability associated with a given point by assuming an
r -0.44
P(X) = (25)
N+0.12
Where r is the rank, N and X is the total number of data values (table3)
Peak values are selected for flood records of over 20yrs. The data is usually arranged
in a descending order starting with the highest. Out of this, the return periods and
frequency of occurrence are established. The statistical concept of flood return period
is useful in achieving the least overall cost solution, when the design engineer tries to
balance the cost of a structure designed to pass a calculated flood against the cost of
repairs to the structure, in case the flood exceeds limits of the control structures.
Incase larger structures are required; other factors are taken into account such as the
cost of delay of transport on the major highways and the cost of damage to other
assets. Such an economic analysis results in the adoption of increasing flood return
M.O.W.D (1975), the most usual figures for economic consideration are:
35
(a) For waterways of areas up to 0.72m2, 5-10 yr flood is used for small sized
structures.
(b) For waterways of areas of 0.75-lOm2, lOyr floods are used for medium sized
structures.
Most of all the artificial and natural water ways in the Nairobi watershed fall within
the stated categories, and so this justifies the use of Gumbel distribution function in
the watershed.
They are necessary in the design of flood control and other watershed management
structures. Flood hydrographs are obtained from streams gauged with automatic
water level recorders. Such instruments are not readily available for most streams in
Kenya, because of the high initial cost and lack of proper maintenance; leading to lack
data, the flood hydrographs cannot be generated (Onyando and Chemelil, 2004).
From the hydro graph, the significant changes taking place in the watershed can be
identified by visual inspection. In general, the volumes of a stream flow at any given
hydrograph equation;
............................................................................................................ (26).
'if = J'Q.dt
J
36
Where
defined function oft, the integration cannot be performed analytically. Normally, and
to be more realistic, the volume of stream flow may can be obtained by integrating
hydrograph between two successive mid-nights covering that particular day. If the
volume of stream flow is divided by 24 hrs, the average discharge for the day is the
2.9 Modelling
Several approaches have been used in the study of hydrological problems that have
evolved over the years, as evident from proliferating hydrological books and
periodicals (Cunane, 1978). These approaches have so far been diverse and hybrid
such that, it is not possible to distinguish one approach from the other (Singh, 1995).
Numerical modeling is not widespread here in Kenya because of the scarcity and
developed a model for predicting floods for design of small hydraulic structures.
since the area is covered with some rainfall and stream gauging stations. Modelling
Nairobi floods will provide a tool for better planning and improvement of storm water
hydrologists examine emerging problems and exploit new data sources (Diskin and
37
Simon, 1977). Models enable us to study very complex problems and synthesize
different kinds of information. However, model results are only as reliable as the
model assumptions, inputs, and parameter estimates (Sorooshian and Gupta, 1983).
The emphasis here was directed towards simulation of runoff, based on daily rainfall
and Burt, 1985). Models have contributed a lot to research in hydrology and their use
makes planning easy and expedient (Luijten, 1999). Maidment (1993) noted that,
models deal with four distinct areas; namely, pollution control, flood mitigation, flood
control and water utilization. He observed that models can be grouped into surface
water quality models, ground water constituent transport models and surface water
hydrology models. Among other fields of hydrology, models have been used for
pollution control studies and urban drainage studies. James (2003) found out that,
models are selected on basis of their performance, as a good base for better planning
runoff behavior rather than the small-scale hydrological processes by which rainfall
hydrograph separation and deriving of a Slow Flow Index (SFI), Dynamic Response
continuous stream flow, rainfall data, evapotranspiration, air temperature which for
some unknown reason, data for this study area was not available from the MOW &1
records.
Hydrologic Modelling System, a computer code used which models rural and urban
flood hydrology, and calculates hydrologic routing based on input hyetograph and
of dendritic watershed systems. The model assumes that, the basin is a very wide
known input and runoff as the unknown output. The basin characteristics and runoff
(ii) Loss models which can estimate the volume of runoff, given the precipitation
(iii) Hydrologic routing models that can account for overland flow, storage and
(iv) Direct runoff model that can account for overland flow, storage and energy
processes include:
- Runoff computation,
- Direct runoff,
watershed and how much of this infiltrates a pervious surface. They also answer the
question of how much volume of runoff remains on the pervious and impervious
surfaces when the runoff is generated. The choice of the type of model to use depends
on whether the watershed is rural or urban and much more on the soil moisture
(a) Initial and constant rate model: This model is an event, lumped,
storage within an initial loss. Parameters required for this model are initial
(b) SCS curve number (CN) model: It is an event, lumped, empirical, and
surface run off for rain fall events. It computes CN by considering the land
type, hydrologic soil group and land management practices of any gi ven
area. The required parameters for this model are initial loss and curve
number.
(c) Gridded SCS (CN): This model is an event, distributed, empirical and
fitted parameter model. The Curve Numbers are specified in a grid cell
file. The parameters required for this model are abstraction ratio, potential
(d) Green and Ampt: This is an event, distributed, empirical and fitted
(e) Deficit and constant rate: It is a continuous, lumped, empirical and fitted
and transportation processes. The required parameters for this model are
initial deficit, maximum deficit, loss rate and recovery rates (HEC, 2001).
a soil moisture accounting unit for each grid cell. The ModClark being a
and processes are considered explicitly. Parameter required for this model
between one unit of excess rain fall and the resulting direct run off. The
only requirement for this model is the unit hydrograph (HEC, 2000)
sub-basin of the outlet (HEC, 2001); this model is based on the linear
reservoir model. The traditional wave theory of the kinematic wave friction
(i) Snyder's UH: It is an event, lumped, empirical, fitted parameter model. Its
G) SCS UH: This model is event, lumped, empirical, fitted parameter model
concentration for the entire subbasin, storage coefficient and grid-cell file.
The grid-cell file contains; coordinate information, area and a travel time
1979). The HEC-HMS models for base flow which simulate slow
subsurface drainage of water from the watershed system into open channels
include:
event, lumped, empirical, fitted parameter model that uses a constant rate
base flow at all simulation steps that fall within a particular month. Base
flows are required for all months that fall during a simulation time window.
Model that computes base flow from ground water storage. It can only be
(iv) Lag Method: It is an event, lumped, empirical, fitted parameter that routes
Channel flow with translation and no attenuation (HEC, 2001). Lag time is
(v) Modified puIs: This is an event, lumped, empirical, fitted model, which is
outflow curve, number of sub-reaches and initial condition for all Sub-
Models used in this research work are described in the following sections:
Runoff losses are modelled using the initial and constant loss model. The underlying
concept of the initial and constant rate-loss model is that, the maximum potential rate
of precipitation le' is constant through out an event. Thus if P, is the Mean Annual
Precipitation (MAP) depth during a time interval t to t + M ,the rainfall excess Pe,'
infiltrates or evaporates. This loss occurs prior to the onset of runoff. Until the
accumulated precipitation on the pervious area exceeds the initial volume, no run off
P,f£-F)fIP,"21 ifP
,SI&P,"2F,} (28)
P e, { Otherwise
Q
•••••••••••••••••••••••••••••••••••••••••••••••••••
Where
The model includes one parameter (the constant rate) and one initial condition, the
initial loss). Respectively, these represent physical properties of the watershed soils,
land use and antecedent condition. If the watershed is in a saturated condition, this
value will approach zero. But when the watershed is dry, then la will increase to
represent the maximum precipitation depth which can fall on watershed with no
runoff; but this will depend on the watershed terrain, land use, soil types and soil
treatment. The constant rate loss can be viewed as the ultimate infiltration capacity
(HEC, 2001). Skaggs and Khaleel (1982) published estimates of infiltration rates for
soils as shown in (table 2.1). These values may be used in the absence of any other
information. Since the initial loss rate and the initial condition model parameters
were not measured, their values are best determined by calibration and optimization.
Table 2.1:Infiltration rates for soils after (Skagg and Khaleel, 1982)
cm/hr
Several methods are available for use in the HEC-HMS model when handling
precipitation data in the meteorological model, but only one method can be used on
each new project using this model. These include; user hyetograph, user gauge
weighting, inverse distance gauge weights and standard project storm (HEC, 2000).
At the heart of the HEC-HMS routing models are the fundamental equations of open
channel flow; the momentum and continuity equations. The two equations when put
together are known as the St Venant equation or the dynamic wave equations. The
momentum equation accounts for forces that act on the body of water in an open
channel. It equates the sum of gravitational forces, pressure forces, and friction forces
to the product of fluid mass and acceleration. In one-dimensional case, the equation is
expressed as;
dy Vdv Idv
Sf = So ------ (29)
ax gax gat
Where
v = Velocity (m/s)
t = Time (s)
46
vav
- = C onvective
. acce I·eration (diImenSlOn
. 1ess )
gax
The continuity equation accounts for the volume of water on a Reach of an open
channel, including flowing into the Reach, then flowing out of the Reach, and then
stored in the Channel Reach. In the one- dimensional case, considering a unit width,
av ay B --q
A -+ VB -+ ay _
ax ax at (30)
Each term in the equation inflow and outflow from, or storage in the reach of the
A ay = Prism storage
ax
VB ay = Wedge storage
x
B ay = Rate of rise
at
The momentum and continuity equations are derived from basic principles, and the
Velocity is constant
All flow is gradually varied with hydrostatic pressure prevailing at all points,
Each model in the HEC-HMS for flood routing solves the momentum and
Storage in the reaches is modelled as a sum of the prism storage and wedge
profile, while the wedge storage is the additional volume under the profile of the
flood wave. During the rising stages of the flood, the wedge storage is positive
and is added to the prism storage. At the falling stages of the flood, the wedge
storage is subtracted from the prism storage. The volume of the prism storage is
the outflow rate 0, multiplied by travel time through the reach, K. Volume
travel time K. On this note, the Muskingum model defines storage as:
Where;
storage and outflow are highly correlated, then X = 0.0 in that case, the equation
L
K =- resolves to S = KO (33)
Vw
When equation (31) is substituted in equation (32) the final equation becomes;
hydrograph, given the inflow hydrograph ordinates (I, for the time t) an initial
This model accounts for varying conveyance between main channel and over bank
conveyance (Ponce and Yevjevich, 1978). The physical characteristics and the
geometry of a river contribute to its use for flood routing in rivers where down stream
information is not available (HEC, 2001). The model is event, lumped, quasi-
combating excess water in streams. This is more commonly called flood control in
the United States, while the term flood-damage mitigation had been adopted from the
Australian practice to emphasize that, absolute control over floods is rarely feasible
either physically or economically. For conciseness, the term flood mitigation is used
as a short hand for the longer term f1ood-damage mitigation. This broad definition
major flood, but may be able to minimize damage to crop and property within the
flood plain of the river (Linsley and Franzini, 1984). Flood control is a relative term
as it is not economical to provide protection for the largest floods that will occur.
Flood control structures can only do that up to their design capacities. Unable to cope
with enormous volumes of water involved. Flood catastrophe brings home the lesson
that, protection from floods is only a relative matter, and that nature demands its toll
from those who occupy flood. According to Linsley and Franzini (1984), the
or closed conduits
improvement
warnings
The physical factors of most flooding situations are against the planner who attempts
to provide absolute flood control and talk of adequate protection, only to imply no
more than calculated risk (Hoyt and Langbein, 1955). In hydrologic study, every
estimates. Though rare; the Probable Maximum Flood (PMF) is sometimes used for
flood control projects. In most cases, this is a rare event and therefore, the Standard
Project Flood (SPF) which is normally 50% of the (PMF) can be used (Linsley and
Franzini, 1984). The damage caused by floods can bring about losses such as;
(a) Direct losses to property, crops, land and can be determined in monetary
terms.
(b) Indirect losses; such as depreciation of property, traffic delays and loss of
income
Since we are dealing with head water flood studies, it can be abit difficult trying to
apply PMF or SPF. The maximum size of a watershed for a head water area is
small-area flood. In most cases, head water floods are typically flash floods of short
duration that occur rather frequently, two times or thrice a year, having duration of
The best programme for flood management for a river can be obtained if a master
amount of flood plain for a flood channel (Linsley and Franzini, 1984). Flood plain
encroachments initiate a cycle of higher stages which are contributing factors to flood
spread (Hoyt and Langbein, 1955). This ends up with areas which are perceived to be
safe getting inundated when it floods (plate 2.11). Building structures constructed on
the flood plains are normally left with flood marks as water level rises. Development
areas subject to flooding may be slums, and urban renewal might be used to clear the
area and convert it to uses not threatened by flooding. Careful study of each flood
plain will specify recommendation for flood plan management. (ASCE, 1951).
Evacuation and flood proofing are used by property owners. Emergency evacuation
for flood-damage reduction can be used for large catchments, but this proofs difficult
These methods have been used with success in some countries. The methods used in
most cases depended so much on the physical characteristics of the watershed area
affected by floods.
53
CHAPTER 3
METHODOLOGY
3.1 Data Acquisition
From January 2005, the research work on the Nairobi watershed to generate runoff
from rainfall data commenced in earnest. The research problem having been defined;
this led to the decision on the type of data and the method to be used for collection.
Both Secondary and Primary data were found necessary, and therefore collected for
this research. The Secondary and primary data were obtained from the various data
centres and sites in the study area. The Secondary data included; rainfall and stream
flow data and topographical map sheets. Primary data set was required for the
Nairobi River channel geometry, and information on the present conditions of the
artificial and natural drainage in Nairobi watershed. Field survey data was collected
type 500 series). Information on the present condition of the drainage system in the
number of stakeholders.
Data and information acquired made it possible to expedite on compilation of the data
and its analysis. Other information and data required for this research were obtained
from the Kenya Soil Survey (KSS) Centre, National Agricultural Laboratories (NAL),
Kenya Agricultural Research Institute (KARI), Nairobi City Council (NCC), Survey
of Kenya (SoK) and the Ministry of Roads and Public Works (MOR&PW) materials
testing branch.
54
This was data earlier collected and stored in the various data centres within the
watershed area; most of it was raw data that had not been processed. The main
in the Nairobi watershed was the daily rainfall. Other supplementary data for this
study included Stream flow, soil type, Soil parameters and topographical map sheets.
All the precipitation data from other stations within the watershed are usually
after verification. For this study, data was obtained for Dagoretti Corner Station,
Airport Station, Jomo Kenyatta International Airport Station (JKIA) and Moi Air
Base Station (MAB). Precipitation data was obtained for a 25 year period; from 1980
to 2004. It had very few missing gaps to be filled. For security reasons and
maintenance, the rain gauge stations in the watershed are located within government
Daily Stream flow data from 1970 to 31 st October 1992 was obtained from the
Ministry of Water and Irrigation (MOW &1) data bank, for 3BA29 Regular Gauging
Station. Nairobi watershed had several Regular Gauging Stations well spread in the
rivers and streams, the Ministry of Water and Irrigation (MOW &1) has the sole
responsibility of maintaining the stations and storage of data collected in a data bank.
55
Stream flow data obtained from the ministry had no continuity, as record for most of
the years was missing. A visit to some of the Regular Gauging Stations (RGS) sites
revealed that; most stations had been vandalized and were no longer functional. For
purposes of this study, there was need to have continuous stream flow and rainfall
data for periods not less than 20 years, but this was not possible; especially with the
stream flow data due to the erratic manner in which it was collected and stored. The
stream flow data from MOW &1 was not quite a reflection of the precipitation data
collected in the watershed from 1985 to 2004. Lack of connectivity in the two data
sets called for a lot of caution when using the data. Recorded figures which appeared
to be outright outliers had to be ignored, since they could not tally with the
Topographical map sheets for Ngong 14811, Limuru 148/2, Nairobi 148/3 and
Kiambu 148/4 at a scale of 1:50,000 were obtained from Survey of Kenya and used to
create a Digital Elevation Model (DEM). From a raster form, a digitizer cursor was
used to trace spatial features. Contour lines for each map were digitized into Arcinfo.
Lines were built after digitization, cleaned, projected to the Universal Transverse
Mercator (UTM) and then transferred to the ArcView. In the ArcView; using geo-
processing wizard, this formed a single map. An attribute map was created with
contour heights. The map was then interpolated using a value domain and pixel of
30m in size. This created the DEM which was exported to ArcView in Arcinfo
delineate the watershed study area map and to specify the location of the regular
gauging station 3BA29 RGS; at the casino roundabout next to Museum Hill, stream
network, rainfall stations and the road network It covered areas to the east, west,
56
south and north of 3BA29 Regular Gauging Station. Similar results could have been
(2000), hydrologic parameters such as initial loss constant, initial flow, Curve
Number, Soil Conservation Service lag time were all estimated on the basis of the
This was data collected afresh and for the first time, and thus happens to be original in
character (Kothari, 2003). This research at some point being both descriptive and
explanatory, the primary data had to be obtained through field survey, observations
carried out for the Nairobi River channel using a 'Total Station', which is a Digital
Theodolite (Topcon 500 series). This was to obtain data to facilitate in routing the
flood through the river channel, using the Muskingum-Cunge-8 point method and
A data base was developed to establish the channel geometry, by taking spot levels
along the Nairobi river channel from BA29 RGS to Kariobangi south bridge. The
main purpose of this was to estimate the channel bed slope and geometry of the
Nairobi River had undergone many geomorphological changes in its course due to
basis of the existing channel conditions. Initially, seven reaches were selected at
57
equal intervals of 500m each, starting from RGS 3BA29 to Kariobangi South Bridge,
River. This led to selection of the seven reaches at varying lengths, by considering
salient points which can be easy to identify on the ground. The reach selection was
done in the order of; Museum Hill-Globe Cinema Bridge, 700m; Globe Cinema-
Kariobangi South Bridge, 1250m. Among all the reaches selected, the least reach
length measured was 450m, while 2150m was the highest reach length measured. To
obtain the channel cross-section data, measurements were made from the left
overbank to the right overbank of the stream channel using the total station (digital
computed the reduced levels in the field, cutting down on any office computations.
Model selection was one of the major tasks in this research; to identify a model that
can simulate urban floods and establish the peak runoff and volumes in a watershed
whose natural drainage system forms a dendritic pattern. The model had to take into
consideration the complex nature of the urban watershed when accounting for
precipitation losses during flooding, and the dynamics of the simulated floods.
58
To carry out the study, two models were considered for possible use; the HEC-HMS
and the IHACRES model. The HMS was developed by HEC of USACE as a
(HEC, 2001). The HEC-HMS is an event based model, and it can effectively use only
one year data at a given time. It is a model designed for short duration storms, with
maximum duration being 24 hours (HEC, 2002). It had the advantage of its multi-
These include; large river basin water supplies and small flooding studies or natural
conjunction with other softwares for urban flood damage studies, flow forecasting,
damage reduction, flood plain regulation and systems operation. The HEC-HMS
interface was contained within the Water Management System (WMS), this made it
simple to input data and display the analyzed results. Manually; precipitation and
stream flow data can be entered into the program, or it can be loaded from Data
included other options for modelling rainfall losses, unit hydrograph and channel
routing. Results computed can be viewed from the basin model schematics. This is
unlike the llIACRES model in which data entry is made for several years. It does not
have a Graphical User Interface (GUI) like the HEC-HMS model, which makes data
entry more direct. The IHACRES model considers catchment descriptors which are
only applicable to rural watersheds; therefore, this made the model unsuitable for
characteristics.
The HEC-HMS model program was used to simulate precipitation-runoff and the
routing processes for the natural channel. Since a model is a set of equations, relating
something known (model input) and something unknown (model output). The model
there was need for more information, which was to be used for comparing the
feasibility of any proposed flood mitigation measures. The benefits of the measures
In the Nairobi watershed, there was need to predict flows, stages, velocity and their
timing in order to provide information required for decision making. In the HEC-
HMS model simulation; usually, the known is precipitation and stream flow, while the
unknown outputs are the stage, flow magnitude and the velocities which keep on
60
changing at different points of the channel reaches. While Simulating using the
other sub-models in the HEC-HMS were also found useful in simulating runoff from
the watershed, routing the flow in the channel and the behaviour of the water-control
structures'; thus predicting flow, stage and timing (Loague and Freeze, 1985).
(b) Soil Conservation Service (SCS) Curve Number (eN) for run off volume
(c) User specified (SCS) unit hydrograph (UH) for direct runoff
(e) Muskingum-Cunge-Spoint method for routing the flow through Nairobi River
Channel
The runoff volume generated dealt with the question of precipitation volume that
collected on the watershed. The HEC-HMS modelled the precipitation that was
retained on the watershed, and the runoff generated on both pervious and impervious
surfaces of the watershed. Direct runoff; including overland flow and interflow in the
runoff process, described what happened when water infiltrated or had been stored on
the watershed moved over, or just remained beneath the watershed surface. The base
flow simulated the slow subsurface drainage of water from the hydrologic system into
the watershed channels. The channel flow simulated a one-dimensional open channel
flow, thus predicting time series of the down stream flows or discharges and time.
61
detailed hydrological model, it was always necessary to account for the movement
and storage of water through all components of the system. The HEC-HMS was such
a model. In the rainfall-runoff modelling process, the main boundary condition was
runoff peaks and volumes as detailed in the framework. Implicitly, the HEC-HMS
omitted any detailed accounting of water movement within the soil, given that; this
was an urban watershed. The near surface flows and overland flows were modelled
as direct runoff into the stream channel. A detailed model of interflow to the
ground water aquifer was represented only as a combined outflow to base flow (HEC,
2000)
Precipitation
Evapo
transpiration
Water body
I
il1filtJralio n
basenow
watershed
discharge
The HEC-HMS model operated in form of projects. Each project accepted daily
precipitation data collected over a period of one year from January to December, for
365 days. Every annual data input made into a meteorological model, basin model
constituted a starting time and an ending time; intervals within which data collection
was made. The model exhibits flexibility such that; daily precipitation and stream
flow data for short time durations (in hours and minutes) can be used to study other
short events and run simulations. Also, data collected for short intervals of time in
minutes can be used to conveniently carry out any other study using this model
There were three key components in running the HEC-HMS model; basin model,
treated as models in every sense of the word 'model'. Through the components, it
was possible to input data in order to create a basin schematic, input precipitation
data, stream flow data and the control specifications. These components immensely
3.2.4.2Basin Model
From the component's menu, a selection was made for the basin model as a new
63
menu item in the project definition screen (HEC, 2001). This was applied to create
the Nairobi sub-basin schematic and other attributes linked to the basin model
configuration. The watershed basin model schematic was limited to one sub-basin,
because the study was limited to 3BA watershed. The sub-basin was interconnected to
the rest of the stream Reaches by a junction at 3BA29. From this junction, 7 stream
Reaches were selected from the 3BA29 RGS and terminated at Kariobangi South
Bridge on Outer ring road as the outlet. The outlet point was chosen arbitrarily for
case study, considering the position of the sub-basin model and the reach connections.
From the basin model attributes, other parameters inputs were made into the model.
These included; watershed area, loss rate, initial constant, base flow, SCS lag time,
recession constant, and other parameters related to the physical characteristics of the
stream channel and basin transforms. Each of the attributes was essential in executing
The meteorological model was created as one of the main model components. This
facilitated the entry of gauging station identification codes, gauge weights and
precipitation data for all the selected rain gauge stations in the watershed. The rain
Agricultural Laboratories, Muguga Forest Station, Wilson Airport, Moi Airbase and
Jomo Kenyatta International Airport Stations (JKIA). They were all none recording
stations and their gauge weights added up to 1.0. As a prerequisite for successful
HEC-HMS model run computation, an entry had to be made for an arbitrary recording
rain gauge station from among the non-recording rain gauge stations, even if it never
64
existed, so that the model can run (HEC, 2001). The recording station was not
stations. All the data entries made were saved in the model project, for use in
simulation runs.
The e control specifications were the start and ending times. Their choice depended
on when the meteorological events were recorded. They had to be clearly specified
before the model can be configured for simulation. The HEC-HMS model accepts
annual daily or hourly data sets for new projects, starting from 1SI January to 31 SI
December of any year selected. Nairobi watershed had precipitation data for non-
recording stations whose readings were made daily at 08:00hrs, this is after every 24
hours. The Starting time and ending time control specification was from 08:00 hrs-
08:00hrs of the following day. This time interval which is a computation step had to
specification representing the 24-hour clock. This determines the resolution of model
results computed during a run (HEC, 2001). The model provided an option of
choosing between any time durations, so long as they were within the months of a
particular year for model simulation. For ease of calibration and model run
simulation, the April 1SI to May 31 si dates were chosen. Data record for these months
was continuous, besides falling within the long rain season of the watershed area.
65
The Hard constraints were specific calibration constraints used in the HEC-HMS
model. They limited the range each parameter value can be estimated to during
optimization (HEC, 2001). These constraints were chosen after a sensitivity analysis
during model creation and used as a guide during the research, to account for the
many watershed variations within reasonable limits. They preclude variables that
could cause numeric instabilities or errors in the model computations. The parameters
used were all within the specified limits in table 3.1. Through the calibration process,
The SCS Curve Number CN is simple, predictable, and a stable method relying on
only one parameter. It varies as a function of soil group, land use and treatment,
surface conditions and the antecedent moisture conditions. This is a method widely
accepted for use all over the world (Ponce and Hawkins, 1996). The SCS Curve
Number was used in this study to account for precipitation excess as a function of
cumulative precipitation, soil cover, land use and the antecedent moisture based on
(p-Ia)2
the equation: P = (35)
e P-Ia+S
Where;
Until the accumulated rainfall exceeds the initial abstraction, the precipitation excess,
Since I a = 0.2S
(P-0.2S)2
Pe = .- (36)
P+0.8S
25400 - 254CN
S= (37)
CN
67
CN has a range from 100 (for water bodies) to approximately 30 for permeable soils.
composite CN was found suitable and was a function of; land use, soil type, and the
IACN
CN composite = I'A , ••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••• (38)
j
Where;
i-An index of watershed subdivision of uniform land use and soil type
CN j = CN For sub-division i
calibration, on t~e basis of the amount of development realized over the years and the
present land use changes in the watershed. It was to account for the directly-
connected impervious areas as required in the HEC-HMS model for all the paved
urban areas, residential estate areas as well as the newly graded areas.
This is a model based on averages of UH derived from gauged rainfall and runoff
throughout the USA, at the heart of which is a dimensionless single peaked Unit
peak discharge Qp for any timet, a fraction oft pk > the time to UH peak.
68
SCS research suggests that the UH peak and time of UH peak are related by
A
Qp=C- (39)
tp
Time to peak is also known as time of rise and is related to the duration of unit of
!1t
'» =-+t lllg .••.••••••••••••••••..••••••..•••••..••.••••••..••••..•••••••.•.••.•.•••••.•••••••••.•••••••..••..••••.••.•.. (40)
2
In which;
tlag =Time difference between center of mass of rainfall excess and the peak of the
UH. For adequate computational interval in application of the model for this
study.Ar of the ordinates on the rising limb was to be found and tlag estimated from
the calibration.
The process of calibration is like turning the knobs again and again until the correct
parameter values are obtained (Singh, 1995). It matches the observed and the
simulated (HEC, 2000). The HEC-HMS model used in this study was calibrated
using part of precipitation and stream flow data for 1981, which was found to be
search for parameters that can yield the best of the computed results to match the
2000). Calibration was done automatically using 1st April to 31 st May 1981 data. The
model parameters were selected in such a way that closely simulated the behavior of
the watershed. Two parts of this process were parameter specifications and parameter
estimation. Normally, data for calibration should be in longer sets in order to improve
satisfactory (Gupta and Sorooshian, 1985). Though this was not the case in the
present set up, the only data available was used for this study. The main goal of
calibrating the HEC-HMS model was to identify reasonable parameters that could
yield the best of fit results for the computed and observed hydrographs. Rainfall-
runoff observations were made from the same storm. The runoff time series data
represented all the runoff due to selected rainfall time series data. The rainfall data
aimed at providing adequate spatial coverage of the watershed; this kind of data can
also be used in the methods of computing Mean Annual Probability (MAP). Since the
data available was secondary, the only option was to use it as it was.
The HEC-HMS model computed an index for the goodness-of-fit during calibration.
These indices are called Algorithms, and were included in this model to search for
model parameters that can yield the best value for the objective function. The
algorithms were incorporated in the model and displayed after model execution as:
70
gives greater weight to large errors and lesser weight to small errors. It is the sum of
the squared difference between observed and computed flow which is defined as;
n 2
Z = I(Qo(t)-Q,.(t») (41)
1=1
Where
This objective function is evaluated for all times t in the objective time window.
timing. It is a percentage difference between the observed and the computed peak
Where
This method could be a logical choice if the information needed for designing or
planning is limited to peak flow or peak stages. The method can be used to measure
useful in the case of flood plain management study that will seek to limit development
71
in the areas subject to inundation, when flow and stage are uniquely related (HEC,
2000).
Program HEC-1 (USACE, 1998). It compares all the ordinates, squared differences,
~(Q _ QS(I) + QA )
L. 0(1) 2Q
I~ A
Z= - (43)
n
1
QA =- IQo n
(44)
n 1=1
The weight assigned to each ordinate was proportional to the magnitude of the
ordinates. Ordinates greater than the mean of the observed hydrograph were assigned
a value greater than 1.00, when the flow was greater than average, and 0.5, when the
flow was smaller than average. The peak observed ordinate were assigned the
maximum weight using this method. According to HEC-HMS (2000), this function is
an implicit measure of comparison of the magnitudes of the flow peaks, volumes and
comparison that could permit visualization of the model fit to the observations of
hydrologic system.
72
This was an objective function which gives an equal weight to both small and large
differences between the observed and computed hydrographs. It was the sum of the
absolute differences between observed and computed flow. The function was defined
as:
n
Z = ~]Qo (t) - Qs (t)I ······ (45)
t=l
Where
The results of any calibration process are conditional on several factors such as
practice to conduct a verification test. Verification tries and detects any biases that
might have crept into the parameter estimates due to imperfections in the calibration
procedure (Duan, 1991). In the Nairobi watershed, verification was carried out using
1st April to 31SI May 1982 data. This data was chosen since it had not been used to
Though search methods were used to estimate the optimal parameter values, they
73
did not indicate the parameters with the greatest impact on the solution. For this
reason, sensitivity of each parameter with respect to the objective function was
estimated from:
0.995X -1.005X
S= (46)
X
Where
S = sensitivity measure
response surface in the region of the best parameter value (Sorooshian and Arfi,
considered to be well determined (sensitive) and which ones are poorly determined
Crippen, 1977).
The univariate gradient search algorithm was used as a search method in HEC-HMS.
moved towards the parameter that yielded minimum value of the objective function.
Where
!li
k
= the correction of the parameter throughout all the iterations in calibration, this
best fit of Computed to observed Hydrograph. In case of any error, the HEC-HMS
was supposed to change the trial parameters and reiterate. All this depended on the
univariate gradient search algorithm. When this method is used, it evaluates and
adjusts one parameter at a time, while holding the other parameters constant (HEC,
2000).
Simulation runs were executed any time to update the results whenever there was data
adjustment to signify a new scenario. The results were viewed through the basin
model screen. Whenever the schematic basin model was activated in the HEC-HMS
model screen saver after each successful run, this gave options to view a variety of
graphical, tabulated results, summary tables, base flow, channel reach connections
These were estimated using the 1981 streamflow data. The 1981 data was used
75
because it had a continuous record for all the days and months of the year. The
rainfall data was also found to be continuous and the rain gauge stations well
Based on observed data at 3BA29 RGS for 1981 year, a hydrograph was generated
using the discharge component of the HEC-HMS model which uses daily stream flow
and precipitation data for the whole year. From the hydrograph, peaks flows can be
identified for each of the twelve months of the year. According to Jones (1975),
isolated storms which are unseasonal are also experienced in Nairobi watershed area
throughout the year. Further to this, runoff Peaks were determined by projecting the
observed stream flow data for 1981 water year, and by simulating runoff from
precipitation data using the HEC-HMS model. These methods facilitated in obtaining
the magnitudes of the peak discharge values and volumes. After calibration and
verification of the model, different scenarios were simulated to answer the 'what if?'
question. These were to depict the changing trends on the watershed area as a result of
stored, evaporated and transpired, and so, removing all these abstractions from
precipitation left us with the rainfall excess to generate the peak runoff and volume.
In the watershed, there was no linkage between the immediate soils overlying the
aquifer below the ground surface, since this was an urban watershed. Ground water
held in the aquifer might have originated far from the area of precipitation. Evapo-
transpiration and infiltration are always treated as losses, but in this study case, the
76
assumption was that; evaporation and evapotranspiration were quite negligible at that
Flood frequencies are estimated by use of observed or simulated rainfall data and
valid watershed models (Parodi, 2005). The statistical concept of flood return period
is useful and economic exercises can be carried out to determine the optimum return
period of floods for various structures. It is usually necessary to balance the net
present values of the proposed large structures against the maintenance cost of smaller
structures. Incase of larger structures being preferred, other factors such as the cost of
delay of transport on major highways and the cost of damage to other assets may have
increasing flood return period, to increase the structure sizes (Anderson and Burt,
1985).
The Gumbel distribution method was used for frequency analysis of the identified the
extreme values of the peak discharges (Shaw, 1988). Estimation of parameters and
selection of distribution methods becomes unreliable when the observed data is for a
period less than 15 years (Chow et aI., 1988). The primary objective of the frequency
analysis was to; relate the flow magnitudes of the extreme events to the frequency of
occurrence through the use of probability distribution functions. Data observed over
an extended period of time in a river system was analyzed. A Gumbel plot of the
extreme values was made using stream flow data obtained from 3BA29 RGS. This
was applied to determine flood magnitudes with return periods of 2.33; 5; 10; 50; 100;
77
500 and 1000 years. Though the stream flow data obtained from the Ministry of
Water and Irrigation data bank had so many missing gaps, the data was used as it was,
since only the peak value for each year was required. Through continuous calibration,
the observed stream flow data was made compatible with the observed precipitation
data. To establish the probability and risks of these floods, the following relations
(ii) Probability that certain event will not occur in n successive years
P; =(1-~)n (49)
(iii)Probability risk that an event will occur at least once in n successive years
R=l-(l- ~ )n (50)
Where
n = number of years
T = Return period
The process of computing travel time and attenuation of a flood wave is referred to as
routing (HEC, 2001). This took into consideration that, there was no local flow into
Nairobi river channel between 3BA29 RGS and the Kariobangi south bridge;
78
therefore, the unsteady flow in the channel was computed using the combined
Muskingum-Cunge 8-point method was used for flood routing along the Nairobi
River channel. Since overbanks existed as established from reconnaissance and the
actual survey, this method was found to be the most suitable for flood routing in this
case. This is primarily due to its advantage that, it is a physically based model and
there was no downstream data available for calibration (USACE, 2000). The
Muskingum-Cunge-8 Point section method uses the same solution techniques as the
with eight station coordinates (figure 3.2). Parameters required this included; reach
length, energy slope, Manning's roughness coefficient 'n' for each reach partition and
The Nairobi river channel was divided into left overbank, main channel, and the right
over bank partitions. The Flow was computed separately in each channel partition
(HEC, 2001). At each of the 8 points, measurements were made using a digital
Theodolite to establish the reduced levels. These reduced levels were used in model
simulations and routing. According to Ponce and Yevjevich 1978, the error in
From this condition, the flood wave can spread beyond the river bank limits into the
. TSU
flood plains. The channel slope should be >0.002 and -- o ~ 171 ; (52)
do
These were among the conditions which justified the use of Muskingum-Cunge 8-
Where;
T = Hydrograph duration
So = Bed slope
u = Reference
0 mean velocity
g = Gravitational acceleration
The channel properties and parameter values used for M-C-8 routing method; reaches
lengths, energy slopes, and cross section geometry have been estimated from survey
the flood magnitudes. The corresponding stage levels determined using the rating
equation. The stage heights made were for estimating the extent of inundation by the
flood wave; either on the left overbank or the right overbank of the stream channel,
depending on the topography. Using the observed stream flow records for 3BA29
RGS, the simulated flows and their gauge heights, a plot was made of the stage-
discharge curve. This enabled the establishment of a rating equation of the form;
Where;
Q = The discharge in m 3 / s
a and b are constants obtained from the y-intercept and the gradient of the curve.
collapsed river banks, erosion and the river capture caused by flood waters in every
rain season.
There were so many changes took place in the Nairobi watershed. Urbanization and
very drastic changes (SCS, 1986). According to Nderitu (2006), vegetation changes
had a profound effect on water resources and hydrology in Kaiti watershed. Though
81
this watershed is different from the Nairobi watershed due physical factors and
and soil texture in the watershed would bring a lot of hydrological changes. Whenever
changes are made on SCS lag time, degree of imperviousness, initial loss, constant
loss rate, initial abstraction and the recession constant through element editor, this in
To investigate the conditions of the current drainage network in Nairobi with respect
to storm water management; the main focus was on urban storm drainage, urban land
drainage and highway drainage. Baseline information on the state of the current
drainage system was obtained through the use of questionnaires, to gauge the
According to Dixon and Leach (1978), the smallest sample size acceptable in
sampling theory is thirty one, so, thirty five questionnaires were circulated to some of
the key stakeholders in the city. These questionnaires had both structured and open
ended questions. Responses from the questionnaire were analyzed using the Statistical
Package for Social Sciences (SPSS) version 11.5. This package has the capability to
analyze data which is descriptive and explanatory, by converting the data into the
desired variables (SPSS, 2005). A Digital Camera was used to capture some of the
scenarios to show how the situation of the city drainage system can be; during a
storm. The scenes taken were fro industrial area, Central Business District and some
CHAPTER 4
Basic concepts were adopted in developing a project in the HEC-HMS model. This
began from a new project file by creating a basin model and meteorological model. In
these models, the control specifications were selected. This was followed by a run
configuration; there after, several trial simulation runs as parameter adjustments were
made. Each trial simulation run was followed by an optimization of the parameters,
and every optimization resulted in an automatic computation run after saving the
changes. Each of the optimized computation runs generated a flow comparison graph,
The Basin model created in the basin model file produced the Nairobi sub-basin in a
Reservoirs, Inlets and Outlets. The sub-basin was connected to the reaches by a
junction created at 3BA29 RGS. From the junction, seven stream reaches were
selected starting from 3BA29 RGS in sequence upto the end of the ih reach at
Kariobangi South Bridge, which is an arbitrary outlet. Reach lengths were measured
from the junction (inlet) at 3BA29 to the outlet junction at the Kariobangi South
Bridge. The reaches were not equal in distance due to river geomorphological
changes caused by erosion, collapsing of river banks, deposition and river capture
from season to season. Salient points were selected for easy identification on the
ground (figure 4.1). The meteorological model had the attributes to analyze the areal
83
precipitation, fill in missmg gaps usmg the square mverse distance method and
B.!Ih.!lti·Kim.!lthi 5
Kim.!lthi·U huru 6
U huru·S, K.!Irib.!lngi 7
Kariob.!lngi s bridge
This was the process oftuming the model knobs repeatedly to match the observed and
the simulated data sets. During calibration, three parameters were found to be the
most sensitive when using the HMS model; SCS lag time, area of the watershed, and
the degree of imperviousness. The SCS lag was the most sensitive of all the
parameters. It affected the outcome of all the computation results and influenced the
shapes of the hydro graphs generated. Though the percentage of imperviousness was a
sensitive parameter; its overall effect on the results and the hydro graph were less,
84
compared to the SCS lag time. The Curve Number (CN) was chosen on the basis of
the Hydrological Soil Groupings (HSG's), which depicts the physical changes of the
watershed. The watershed had an optimum composite Curve Number value of 89,
because of the existing impervious and pervious conditions. The curve number was
condition. However, it did not seem to have much effect on the overall calibrated
results. The generation of flow magnitude and the total volumes depended on the
size of the watershed, the degree of imperviousness and the SCS lag time. Through
calibration, an optimum SCS lag time value of 500 hours was established. Whenever
the SCS lag was increased or decreased for each run simulation; this caused a drastic
change on the resultant hydrograph. In all the HEC-HMS model operations, one
parameter at a time should always be dealt with. A computation run should always be
executed to effect changes made on any of the parameters during and after calibration.
140
-- Calibrated
-- Observed
120 --Base flow
--Timestep
100
SO
~
M
g
3 60
0
;:;:
40
20
0
01 11 21 01 11 21 01 11
~r81 MaySl JunSl
Fig. 4.2: Calibration Graph Based on (April 151 to May 315) 1981 data
85
The HEC-HMS model calibration was successfully carried out using the optimized
parameters in table 4.1. According to the hydrograph generated (figure 4.2), the
observed and the simulated flows matched at the optimization location. The vertical
green lines on the output graph denoted the start time and end time of the objective
function. From the hydrograph, it was worth noting that, the existing base flow
changes continued taking place in the watershed, parameter values will also change
over the time. These parameters when optimized can be used for flood predictions in
the Nairobi watershed. Whenever a parameter adjustment was made, this signified
Parameter Value
Imperviousness 55 %
Curve Number 89
Through calibration, a scatter graph in figure 4.3 was produced. This implied that;
there existed a correlation between the simulated and the observed flow data. The
scatter graph in HEC-HMS model terms produced a near 45° line to show that; the
model can neither over predict nor can it under predict floods in the watershed. At the
86
peak point, the model generated deviation of 0.18, which was a good fit considering
140
120 ·0
~
"....,
100 C!)
..l(! C!)
M
g 80 C!)
~
3
0 o
-e
GI 60 .C!).
1ii
"5
E
(i) 40
C!)
C!l
. . (!J .
20
0
,,$' . C!)
• IIII r e n I
HEe • e 10 I e Pe 11k BIIS In:
Pun: 0p I 1_28
sub billS
The computed results in table 4.2 and 4.3 were indicator of a successful calibration.
This implied that; the full objective of calibration had been achieved. It further
showed that, the most reasonable parameters for the watershed at the time when data
In the scatter graph, the straight line indicated the equality of the computed and the
observed flows. If all the plotted points fell above the near 45° equality line; the
87
model could have been biased; implying that, it was going be an over predicting
model. Similarly, if all the points plotted fell below the near 45° straight line; the
model should consistently be an under predicting model. If the entire points plotted
fell in equal numbers above or below the near 45° straight line as it was the case; this
was an indicator that the model was no more likely to over predict than under predict.
The spread of the points also indicated that; the modelled and the observed flow data
matched well, such that, random errors in flow predictions were not going to be large
relative to the calibrated flow. This meant that; the parameters chosen were suitable
j
Total direct runoff (m 'zs) 317858 x lO
The residual graph was an indicator of the behavior of the residual flow at the
optimization location, for each time step during model calibration (figure 4.4). From
the residual graph, the start time was in April 1st, 1981; while the peak time was in
April zs", 1981. The end time of the objective function was on May 31 st , 1981. The
residual flows computed were as a result of the simulated flow, minus the observed
flow. This graph indicated how prediction errors were distributed throughout the
duration of the simulation. In this case study, majority of the residuals values were
not grouped below zero at the start of the runoff event, but above zero, which was an
~~Trnn~Trnn~Trnn~~~~OT~rn~TTrn~TTTrrn~Trnn~Trnn~Tn
.
~
Cl'
:
I
~ Q:i :
-20 ..... ~ I .
1
1
1
-30fTTrnn~Trnn~Trnn~~Tn~OT~rn~TTrn~TTTrrn~Trnn~Trrn~oh
01 11 21 01 11 21 01 11
Ap~1 May81 Jun81
The convergence of the model solution was evaluated after several iterations. Each of
the parameter estimations sharply oscillated up and down, until a point was reached
when the parameters were brought to a convergence through the objective function
89
(figure 4.5). At this point, the parameters started yielding reasonable results. The
implication of this was that; some of the parameters obtained through calibration did
not need any adjustment, either downwards or upwards, otherwise, the results would
MO · -
· -
· -
500 · r-
~O~~~~~~~~~~~~~~~~~~~~~~~~
o 5 10 15 20
It er.rti on
graph between the simulated and the observed data sets for a similar period from
April 1SI to May 31 S\ 1982 (figure 4.6). It implied that; both the calibration and
90
verification data were collected under similar conditions. This made the model valid
to simulate flood hazards in the Nairobi watershed, taking into consideration that;
~
Q)
150
- Simulated
E -Observed
::s
CJ
100
c:
3:
o
;;:::: 50
E
-
CO
Q)
~
rn
1 5 9 13 17 21 25 29 33 37 41 45 49 53 57 61
Time in days (1st Apr-31st May)1982
Fig 4.6: Model verification using1982 (April I" to May 31s~ data
In addition to model verification, a scatter plot was made to correlate the simulated
and observed data sets for the same period (figure 4.7). This produced a linear
correlation between the two data sets, with a linear equation Y = 1.0489 X - 2.1323
data for 1981 and 1982 verification data were collected under similar condi tions. This
model validation made it useful in simulating flood hazards in the Nairobi urban
140
120
100
80
60
y = 1.0489x - 2.1323
40
20
o
o 50 100 150
Observed flow in C••.•
mecs
Though modelling was successful, there was no ruling out errors in some of the
secondary data records obtained. It was noted in some of the cases that; there existed
Flood magnitudes and their frequencies were derived from the observed and the
precipitation inputs and stream flow regimes in the watershed were bound to change
over time. This was reflected in the flood magnitudes and their corresponding stage
levels.
Observed stream flow data set for 1981 at 3BA29 RGS was used to generate a flow
hydrograph for Nairobi River using the HEC-HMS model (figure 4.8). From the
hydrograph, it was noted that peak flows occurred in the months of April to June. In
92
this period, the watershed experiences the long rain season. The highest peaks flow
value were noted in the month of May; a trend which continued upto the middle of the
month of June. The same trend continued during the short rain season between the
month of November and December, though for a short period. There were isolated
cases of low peaked flows noted in the month of February. This was due to the
had a range between 20m3/s and 115m3/s. These flows were characterized by low
120
110
100
90
80
70
'Ol'
~ 60
~
i;:
50
40
30
20
10
0
Jan Mar May Jul sep Nov Jan Mar
1980 1981
---,...------------------- ------------------
HEC
H\JS [
3BA29 Nai robi river Flow Gage
Using the HEC-HMS model, two scenanos were selected to depict the future
watershed changes. They were simulated to answer the question of; 'what if? Their
choice was based on the assumption that; land use changes and urbanization
continued taking place in the watershed. This affected the physical characteristics of
93
the watershed. The results obtained were in form of time series, which expressed
flood magnitudes in cubic metres per second and cumulative volumes in cubic metres
per day.
The precipitation data used for calibration was the same data used to generate flow
magnitudes under urbanized and developed conditions (figure 4.9). In the scenario
where the imperviousness was assumed to have increased to 60 %, the SCS lag time
reduced to 300hrs; with all the other parameters remaining constant, the total
precipitation was 2254.8mm and a precipitation loss of 35.7mm. This reduced the
total precipitation to an excess of 2219.1 mm. When a simulation run was executed; it
generated a hydro graph with a peak flow of 345 m3/s in magnitude (figure 4.10). It
was noted that, the stage level rose abruptly to 6.8m. This flood seemed to take
several days to recede because of the time scale of the precipitation and stream flow
1000
-
E
E
800
c: 600
-
-
CtI
e
CtI
0:::
400
200
0
1 5 9 13 17 21 25 29 33 37 41 45 49 53 57 61
Time in days(1st April-31st May)1981
Fig. 4.9: Rainfall hyetographfor the long rain season (April-May) 1981
94
Apr81 May81
02 07 12 17 22 27 02 07 12 17 22 27 01
i~OI!!.!i!!.!!!.!,!!,!!i'!:!!!!:.:,!!!i!:.!!!'!:!"'!I!'.'i
350
-- Baseflow
300 --Observed flow
--Simulated flow
250
~
fOO
..:~ 150
100
50
0 I
01 11 21 01 11 21 01 11
Be s e t t cw Run. Run 6
H\IS Ne t r cb r sub-basin Time: 010ct06, 17.03
The total simulated volume was 133453 x 103 (cubic metres per day). This was quite
an increment in runoff, given that the observed peak flow was 136.56 m3/s. In the
outlet at Kariobangi south bridge, the peak flow had reduced to 329 m3 Is. This
translated to a reduction of 3.24 %. This flood was able inundate areas adjacent to the
airobi river banks, considering the rise in stage, and the topography of the adjacent
In another simulation, when the imperviousness was increased to 65%, the SCS lag
time reduced to 150hours; with all the other parameters remaining constant, the total
precipitation was 2254.8 mm; since it was from the same storm, the precipitation loss
this loss, the precipitation excess reduced to 2239.6mm. This generated a peak flood
of 460 m3/s in magnitude (figure 4.11). The stage level rose to 7.68m. Though this
95
flood took a long time to peak, its magnitude was far much higher than in the previous
one.
Miy81
02 07 12 17 22 27 02 07 12 17 22 27 01
~[l:::::::::::::::::::::::::::::::::::::::::::::::::::::::r
-- Baseflow
400
--Observed flow
--Simulated flow
~ 300
I--.-.~~~~_~............J .
(Y)
g
g 200
..::
100
O~~~~~~~~~~~~~~~~;;~~~~~~~I
01 11 21 01 11 21 01 11
May81 Jun81
b.
HEC .r.Jb I I
H\oIS u. e
Routed results indicated that the flood reduced by 15m3 Is between the inlet at 3BA29
RGS and the outlet at Kariobangi south bridge. It also implied that; with the current
channel characteristics, a flood wave can sustain a high stage for many hours. The
flood of such a magnitude can spread far into the flood plain and inundate larger areas
along the river course. This may cause a lot of damage to property and infrastructure.
A stage-discharge graph was plotted on the basis on observed and simulated flows. It
drew a relationship between stage levels and the stream flow. The main purpose of
this was to establish a rating curve (figure 4.12). From the rating curve, a rating
equation was derived of the form Q = aHb; where the constants a = 4.43 and b =
96
2.25. From this information, the rating equation was expressed as Q = 4.43H 2.25. The
equation was used to compute the river stages for the observed and simulated flood
magnitudes. Based on the equation; with a simulated flow of 345 m3/s, the stage level
was 6.8m and when the simulated flow increased to 460m3/s, the river stage rose to
7.68m. From these stage levels, it was evident that; the flood wave would spread far
beyond the overbanks and inundate large areas in the flood plains.
20
•
'E'
--
.£«:.-
10
15
I r 5
o
) 1000 2000 3000 4 ODO
-5
Rood magnitudes in Cumecs
From the topographical survey data, some of the lower reaches of the river channel
after 3BA29 RGS had cross-sections whose depth was 4m or less. Unfortunately,
some of the areas are slum settlements areas built along the river channel. It implies
that most structures and properties were not likely to be spared by the floods.
The mam objective of the frequency analysis was to relate the magnitude of the
distribution. The assumption made was that; the stream flow data was independent
and identically distributed. Peak annual discharge values for 21 years' record were
used for frequency analysis. This generated 2.33year, 5year, lOyear, 25year,50
97
year,l00year,250 year and 500year (table 4.4). The significance of these floods
magnitudes and frequencies was how they can be used as design floods. The 10-
25year floods can be used for the design of culverts and open drainage systems.
2.33 50
5 120
10 220
25 330
50 455
100 510
250 580
500 650
From the Gumbel plot based on peak values identified from the observed data, the
simulated flows of 345m3/s and 460m3/s in magnitude were captured in the frequency
curve (figure 4.13). Their return periods were 25 and 53 years. On the basis of the
frequency analysis, it was established that, Nairobi watershed had a mean annual
flood of 50 m3ts with a return period of 2.33 years. This flood had a 43 % chance of
~ 800
a 600 +---------- ...•
~------l
.5 400 +-------------~~--------------~
~
;;: 200
l o +---~--~------~~------~------~
o 2 4 6 8
Return Period in years (In)
Fig 4.13: Frequency Curve based on 3BA29 RGS Stream flow data
98
From the frequency analysis, a 100 year flood of magnitude 510 m3/s was established
with a 1 % chance of being equaled or exceeded. A flood with this return period can
be used for designing large structures like bridges. From the probability curve in
figure 4.14, most of the high magnitude flood events had a very low chance of being
equaled or exceeded. However, these were high risk floods, even though their
0.6
\
.~ 0.5
0.4 .\
~ 0.3 -~-- -------- - --
e 0.2 <,
a. 0.1 ~
~
o
o 200 400 600 800
Flow in Cumecs
Within the scope of the probability analysis, flood events lower than the mean flood
of 50m3 Is did not seem to have more than 60 % chance of being equaled or exceeded.
Model simulations generated flood high magnitude floods. In the Nairobi River
channel, a flood wave usually turns out to be a cascade of mud and water from one
reach to another. Starting from 3BA29 RGS, 7 reaches of varying lengths were
selected upto Kariobangi south Bridge for routing the flood. Since there were
significant variations in channel properties along the selected reaches; each reach had
99
reach defined separately. In order to route the flood wave successfully, the
The method clearly defined separately each reach length as L). x , Manning's roughness
Manning's roughness coefficient for each reach length was estimated. From 3BA29
RGS, an inflow flood hydro graph of 460 m3/s magnitude was simulated using the
HEC-HMS model (figure 4.15). The flow translated to a stage of 7.68m, which was
established from the rating equation. This was quite a high stage, considering that;
the average depth of the stream channel was less than 4m at most cross-sections,
3BA29RGS
400
.--..
300
~
M
g
3
.£ 200
100
o~~~~~~~~~~~~~~~~~~~~~~~~~~~
01 11 21 01 11 21 01 11
'....
b.u' .•. J
Recession was expected in the lower reaches of the river channel, to signify that, the
channel had the capacity to store some of the flood waters. However, the out flow
hydrograph of last reach at Kariobangi south bridge did not show any significant
recession (figure 4.16). From the computed results, the peak flow at the outlet had
receded to 445 rrr' Is. This was a reduction of 15 rrr' Is, which was only a translation of
3.3 %. This made it appear like as if there was no remarkable difference between the
inflow and outflow hydrographs, after comparing the hydro graph shapes.
Kariobangi s bridge GJ
400
'--"300
~
M
g
3
~ 200
100
O~~~~~~~~~~~~~~~~~~~~~~~~~
01 11 21 01 11 21 01
May81 Jun81
The flood wave seemed to be uniform in shape from the first to the last reach. This
implied that, the river channel in its present state was not capable of attenuating the
floods.
101
The speed of a flood wave was measured by determining the travel distance from
maps and recording arrival time of wave at specific sites (USGS, 1994). A graph of
travel time against the distance was plotted; which gave the phase speed or the speed
of the wave (figure 4.17). A reciprocal of this slope would translate into velocities at
different reaches. When there are more of permeable alluvial aquifer sediments along
an unlined channel perimeter, the great depth of water table in the alluvial aquifer
produces a slow rate of flood wave travel (Donald, 2003). This was not the case with
Nairobi river channel, since the channel is perennial in nature and the underlying
aquifers must be fully saturated given the nature of the underlying soils.
~ 700
"C
~~--=-~~------------------------,
§ 600 +-----------------------------~~--~
~ 500 +--------------------------~~~----~
~
.5 400 +---------------------~~------------I
E
;:
300
~ 200 +---------~----------------------__4
~ 100 +-~~~=-----------------------------~
~ 0 +---~--.-~----.-----~.---~--~------~
o 2000 4000 6000 8000 10000
Distance from RGS 3BA29 in (metres)
Fig. 4.17: Phase speed of the routed flood from inlet to the outlet point
From the current shape of the channel bed, the flow wave profiles was expected to
follow an instantaneous changing pattern of the river from the inlet point at 3BA29
RGS to the outlet point at Kariobangi south bridge (figure 4.18). As the flood wave
Kariobangi south bridge, the time taken by the flood wave from inlet point to the
outlet point was computed; since the routing results gave the reach length and the
102
travel time it took the flood wave at each reach (table 4.5). It implied that the flood
wave took 2 hours 45 minutes from 3BA29 RGS to Kariobangi south bridge, which is
very close to 3 hours. This is enough time for people living in some of the most
vulr.erable areas move out, incase a flood wave is spotted upstream and they are
warned early enough. The basis for installing an early warning system should be the
The wave transformation was an indicator of the extent the amplitude of the flood
wave decreased over the selected reach lengths as a sign of recession. From 3BA29
even though, the flood wave being monitored in this case had increased by more than
five times the normal river stage. Since the river channel reaches were not engineered;
For the Nairobi River channel, a longitudinal profile of stream bed elevation against
distance was plotted based on the stage levels (figure 4.17). The purpose of the profile
was to exemplify the dynamism of the flood wave as it cascaded from the 3BA29
103
RGS inlet to Kariobangi south bridge outlet. There was no remarkable variation on
the flood wave as noticed from points along the channel during the routing period,
regardless of the flood magnitude. The channel bed slope seemed to vary at certain
points, although the physical shape of the water profile at almost all the points and
1670
~...
1660+--~~::::---=""---.;:- -=--=--=--=-=W:.::.a""te"-;Jr
t:-'-ro:;-:fi.:.:..:lle'--;o-;-_
--- River bed profile
~ 1650+-----~~----------
E
.S: 1640+-------'<----",-----------1
c 1630+------~=_-~~-----~
.S!
1ii 1620
~ 1610
W
1600+---.---.-----.-----.-----,
o 2000 4000 6000 8000 10000
Distance in metres
Fig. 4.18: Nairobi river flood wave profile from 3BA29 RGS to
Kariobangi South
The shape of the wave profile was not enough to define the type of flow in the river
channel under the existing conditions. Measures that can be taken for flood proofing
and flood protection should be also be based on the river stage levels.
A survey was carried out to establish the conditions of the existing drainage in the
including; the City Council, Gas, Corporate Organizations, Consulting Firms, civil
professional background gave them an insight on urban storm drainage and what
104
returned with responses. The questionnaire had both structured and open ended
questions (Appendix 7). Focus was directed towards the existing drainage conditions,
urban land drainage, urban storm drainage, highway drainage and factors that affect
the drainage system in the city and the resultant flooding whenever there were
seasonal or an out of season storms. Analyzed results were generated using the SPSS
Results from the survey revealed that, Nairobi watershed had both artificial and
natural drainage systems. These handled storm water whose origin was the paved,
unpaved and other urban undeveloped lands. Some of the systems which are
combined, separate and partially separate received sewage from predetermined areas,
stakeholders, blockage of the city drains was due to lack of maintenance and
conversion of the existing drains into solid waste dumping chutes (figure 4.1).
The space left on the road side for construction of open drains was occupied by heaps
of solid waste which seemed to originate from the roadside food kiosks constructed
Plate 4.2: Garbage heap next to a kiosk on Commercial Street Industrial area
The general perception was that, this was one of the major factors contributing to
flood hazards. Opinion by 100% of the stakeholders was that the existing drainage
system was in a pathetic state. The stakeholders were all of the opinion that, the
current system of drainage was not good; and therefore, measures should be taken to
respondents, will increase the efficiency of the current drainage in handling storm
water, but 5 % of those interviewed felt that; the efficiency of the present drainage
system cannot improve unless there is a total overhaul of the entire system.
According to the response from Nairobi City Council, who happened to be charged
with the whole responsibility of operation and maintenance of the drainage facilities
106
in the city, the current drainage systems were combined, separate or partially separate,
depending on the area being drained. The opinion of 100 % of the respondents was
that, the present system was inadequate to meet the increased drainage needs caused
of the respondents, enough maintenance was not carried out to desilt the catchpits,
replace broken kerb gratings and remove debris trapped in culverts before and after
the end of each rain season. They felt this contributed immensely to the problem of
flooding often experienced in the city and its surrounding estates, whenever there was
a high intensity storm in the city. Continued sedimentation had made sections of the
city's open drains to become completely filled up with sediments leaving no single
drain to accommodate stormwater. Notably, fast or sheet flows characterize the road
sides where they have created their own flow paths during storms (plate 4.3).
stagnation, which resulted to overflows on the main section of the road causing a lot
Garbage swept into the drains reduced their capacity to adequately accommodate
storm water. The end result was road sections; where, road shoulders were worn out
of the respondents. In some cases, open drains leading to the cross drainage works
were overgrown with vegetation which has never been cleared before the rains started
(plate 4.5).
108
The implication of this was a reduction in the economical capacity of the drain such
that; when the rainfall intensity increases, the open channel cannot accommodate the
flow generated. It was unfortunate to note that; most of the urban storm drains existed
respondents were of the opinion that, the present drainage system had inadequate
capacity and was poorly constructed. They were of the view that; the whole drainage
and other land use changes. The new system planned should cater for some of the
densely populated estates without proper drainage like; Kawangware, Riruta, Satellite
and slums areas which do not have any organized drainage system.
Whenever there is need for removal of excess water from developed and undeveloped
areas in a city, the modem tendencies are to use large closed conduits in lieu of
ditches where possible. The Nairobi City Council and other major stakeholders
109
through the interview attested to the fact that; open ditches were mostly used as a
quick means of disposing off surface runoff. Response by 96% of the stakeholders
was that; the existing open drainage ditches in the city were not lined with any
impervious materials. The few channels that used to be lined have had the concrete
lining collapsed. This led to reduced depth of the drainage channel and consequent
inundation of adjacent areas during floods. The stakeholder felt there was need to
rehabilitation all the existing open drains in the city to ensure they conformed to the
city's rapid development and expansion. There was lack of proper urban drainage
of large volumes of debris and sediments into some of the streets by overland flow
(plate, 4.6).
Plate 4.6: Level crossing at Tetra pak on Enterprise road Industrial area
The view of 86.4 % of the respondents was that; to rehabilitate the existing drainage
system, the amount of financial investment required was high, but only 13.6 % of
110
them were of the opimon that; the amount of financial investment required to
100.00%
80.00%
60.00%
40.00%
20.00%
0.00%
Heavy investment Low investment
Regardless of the cost of rehabilitation; 100 % of the stakeholders were of the opinion
that, rehabilitation of the drainage system should be given a top priority, if the
Highways occupy long narrow strips of land and therefore pose two drainage
problems; water collecting on the roadway and on the adjacent land slopes if the road
was a cutting. The survey revealed that; the current drainage distribution in the city
and a tile drain system constituting 5 %, for draining selected marshy areas. In
addition; the city had a separate system contributing to 21 % for storm water and a
partially separate system which accounts for 14 % of the drainage system (figure
4.20). An observation by 86 % of the stakeholders was that; most of the drains were
installed in the 1970s and 1980s before the city reached the current level of growth in
terms of population and infrastructure. This implied they have already been overtaken
5% 14%
Tile drain
I!!Ipartially separate
o Combined
o Seperate
• Longitudinal drain
The city's main drainage systems handled a combination of storm and foul water,
according to 60 % of the stakeholders. Even though this may be the case, 59% of the
stakeholders were of the opinion that; the highway drainage facilities that existed
debris and garbage. Some of them felt this was the cause flooding on the city
highways, which resulted to heavy traffic jams, especially where some of roads turned
into open channels during storms in the city (plate 4.7). Some of the vehicles end up
stalling in the middle of the road, making it difficult for other motorists to move.
Plate 4.7: Floods turns Dunga road in Industrial area, into an open channel
112
Field investigations also revealed that; poor highway drainage provision had so badly
affected some of city roads to the extent that, when it rained, some of the offices and
business premises became inaccessible for several hours (plate 4.8). Solid waste
dumped in the open drains caused blockages and impeded the flow of water for
several hours.
Plate 4.8: An office property off Enterprise road inaccessible due to flooding
pools on the roads, which is a major source of inconvenience to both motorists and
programmes undertaken in the city seemed not to include any drainage works
improvement. An opinion by 100 % of the respondents was that; it was not enough to
have smooth roads without proper drainage, because soon or later areas covered by
CHAPTERS
5.1 Conclusions
The HEC-HMS model can be used to simulate floods in an urban watershed when the
urban watershed under the existing conditions as well as future conditions. However,
the HEC-HMS model was not able to give the extent the floodwaters can spread into
the flood plains. When urbanization and development in the watershed increased, and
the degree of imperviousness changed from 55% to 60%, this increased the mean
annual flow from 50m3/s to 345m3/s (an over 600% increment in runoff). The 600%
increase in runoff may be too much for the existing drainage infrastructure.
to avoid floods.
Frequency analyses established that; Nairobi has a mean annual flood magnitude of
50 m3/s with a return period of 2.33 years. The watershed has a 43 % chance of
experiencing a flood in every 2.33 years. From the same analysis, it was established
that; Nairobi has a 100year flood magnitude of 510 m3/s. This can be used to design
large cross drainage structures like; bridges and culverts. Routing the flood through
the Nairobi river channel showed that, the flood waves cascaded from the inlet to the
outlet; they can take a minimum time of 2 hours and 45 minutes. The floods recede
hypothesized; it has inadequate capacity to accommodate all runoff from roads, roofs
and other paved areas. Maintenance of the system is poor and irregular, a factor that
contributes immensely to the problem of flooding. Besides this, there is total lack of
any form of drainage system in some of the estates. The slums settlements have
encroached the road reserves in some areas. The inhabitants dump solid waste into
the open drains blocking them. This has limited the space reserved for drainage. The
river valleys have become slum settlements; the inhabitants have converted the rivers
into sites for dumping solid waste, reducing space for water. This increases stage
5.2 Recommendations
The HEC-HMS model should be customized for use in drainage planning and design
in Nairobi City and for other Kenyan Cities having similar flooding problems. The
model should also be used for flood prediction and installation of early warning
concrete protection walls should be constructed in the most vulnerable areas along the
River valleys and stream channels. This will prevent the high stage floods from
spreading; protect property and the inhabitants of the slums which are situated in the
flood plains. Besides this, there will be need to create awareness on the importance
of reserving space for water and better solid waste disposal methods, rather than in the
river valleys as a non-structural mitigation measure against floods. The Nairobi river
channel can be improved to increase its hydraulic efficiency to enable it attenuate the
floods. Further research is needed to map out the spread of floodwaters along the
115
Nairobi River. The HEC-RAS model can be customized at research level for such
purposes.
There will be need to use the 100 year flood magnitude for safe design of culverts and
bridges in the watershed. Also, a combined surface and subsurface drainage to get rid
of the pools that characterize some parts of the City whenever there is a storm in the
watershed. The City Council should come up with a scheduled and regular
maintenance programmes to ensure that, all the catchpits and Storm water gutters are
inspected, desilted and repaired regularly. A concerted effort by all the key
stakeholders should be made to ensure that; open channels are cleared of vegetation,
Burs, Brushes and Snags. Channel lining, straightening and deepening the open
channels may be necessary in some of the areas. All the major stakeholders should
embrace some of the new technologies by installing some of the Best Management
Practices (BMPs) like; Bioswales in river valleys and Bioretention Islands in the
residential estate areas. There will be need to Rehabilitate all the existing drainage
systems and design new ones using the state of the art storm water design models, to
increase drainage capacity and efficiency in order to cope with the present and future
developments in the watershed. The City Council should enact By-laws that
systems. There is need to install automatic rainfall recorders and automatic stream
gauging stations to avoid the problems arising from lack of real-time data. The
government should treat this as a top priority; for without proper data, no meaningful
Further research should be carried out to establish other factors which are likely to
contribute to flooding in the watershed using Soil Water Assessment Tools (SWATs)
and other geospatial methods to identify areas in the watershed that are most
vulnerable to flooding for protection. There will be need for further research to
establish the runoff contribution from developed and undeveloped areas of the
REFERENCES
Anderson, M.G. and Burt, T.P. (1985) .Lumped Models. Hydrological forecasting.
Wiley & Sons, Inc. pp 604.
Bauer, W.J. (1969). Urban Hydrology. The Progress of Hydrology. Vol.I. University
of Illinois.pp (605-637)
Diskin, M.H. and Simon. (1977).A procedure for selection of objective functions for
hydrologic simulation model.Journal of Hydrology, 34,129-149.
Duan, Q. (1991).A global optimization strategy for efficient and effective calibration
of hydrologic models, Ph.D Dissertation, Department of Hydrology and Water
Resources, the University of Arizona, Tucson, Arizona.
Donald, O.W., (2003) Cascade of mud and water. Seattle annual meeting paper
No. 95-9
Fiddes, D. (1977). Area Reduction Factors for East Africa. The TRRL supplementary
report 259
Findlater, J. (1969).A major low-level air current near the Indian Ocean.
Forsgate, J.A. and Grigg A.O. (1975).The prediction of daily storm rainfall in East
Africa. Flood Hydrograph symposium., Nairobi.
Gupta, V.K. and Sorooshian, S. (l985).The relation between data and the precision
of estimated parameters, Journal of Hydrology, 81, 57-77
Harrold, L.L. (l949)."Has small area Flood been neglected?" Civil Engineering
journal. Pp (38-39).
HEC (2001). Performing simulations and viewing results. User's Manual; version
2.1. USACE, Davis, CA. pp 112-119.
HEC (2001). Basin models. User's Manual; version 2.1. USACE, Davis, CA. pp 39-
77
HEC (2002).Urban flooding studies. Application guide; USACE, Davis CA. ppll-29
HEC (2002). Flood Frequency Studies. Application guide; USACE, Davis CA. pp35
Hoyt, W.G. and Langbein, W.B. (l955).Water facts for the nation's future. Ronald
press, New York.
Ibe, A.C. and Awasike, L.F. (1989). Sea rise on African Coastal zone. pp (105-122).
Nigeria Institute of oceanograph and Marine Research. Lagos.
Jackson, J.D., Beik, S.E. and Kenner, S.J. (2006). Urban watershed management
(B.M.P's).Orlando, Fl
James, W., (2003).On reason why traditional single valued, single event hydrology
(typical design storm methodology) has become simple-minded, dishonest and
unethical, University of Guelph, Ontario, Canada.
Jones, M., Court, P. and Woodward, P.G. (1975).The effects of storm movement
and type of Design storm. Proceedings of a symposium on flood hydrographs,
Transport and Road Research Laboratory, Supplementary report No 259.
Linsley, R.K., and Franzini, J.B. (1984). Flood Damage Mitigation. Water
Resources engineering. McGraw-book Co. (UK) Ltd. pp 631-658
Linsley, R.K., and Franzini, J.B. (1984). Quantitative Hydrology. Water Resources
Engineering. McGraw-Hill Book Co. (UK) Ltd.pp 40-67.
Linsley, R.K, Kohler M.A. Jr. (1988). Hydrology for Engineers: Flood Probability.
McGraw-Hill Book Co. (UK) Ltd.pp 508.
Lujiten, J .c., (1999). A tool for community based water resources management in
hillside watersheds. PhD Dissertation. University of Florida, Gannesville,FL
May, L.W. and Hinks R.W. (1988). Hydrology for urban excess
management.McGraw-Hill Book Co. (UK) Ltd.
121
Ongwenyi, G.S. (1985). Problems of land use and Water Resources Management in
upper Tana Catchment in Kenya. Strategies for River Basin Management. pp (123-
130). Reidel publishing Co., Amsterdam.
Onyando, J. and Chemelil, M.C. (2004). Flood hydrograph generation from small
Catchments in Kenya. East African journal of physical sciences.5 (1): 31-42.
Pereira, H.C. (1973). Land use and water resources organization for mutual
protection of a watershed area. Cambridge University Press. William Clowes and Sons
Ltd. London.pp369.
Pilgrim, D.H, and Cordery, I. (1975)."Rainfall temporal patterns for design floods".
Journal of Hydraulics Division ASCE, 101(Hyl)81-85
Ponce, V.M., and Hawkins, R.H. (1996)."Run off curve number: Has it reached
maturity?" Journal of Hydrologic Engineering, ASCE, 1(1), 11-19
Price R.K. (1978). A River Catchment Research Model .Hydraulic Research Station
Report No./T.168, 2nd Imp. Wallingford England pp (360-363).
Saggerson, E.P (1991).Geology of Nairobi area, MENR, report No. 98, Nairobi.
Saha, V.L. 1983). Impact of climatic change on small Islands. Ministry of Housing
and Environment, Port Louis, Mauritius. pp 254.
Savini, J and Kemmerer J.C. (1961). Urban growth and Water Regimen. u.s.G.S.
Water supply paper 1591-A pp 43.
Schwab, G.O., Fravert R.K., Edminster, T.W., and Barnes K.K. (1981). Head
Water Flood Control. Soil and Water Conservation Engineering. John Wiley & Sons.
pp 241-258.
SCS (1986) urban hydrology for small watersheds, Technical release 55, USDA
Springfield, VA.
Senga, W. (Ed), (1983). Soil and Water Conservation in Kenya. Proceedings of the
second National Workshop./.D.S occasional Paper No.42, U.O.N.
Shaw, E.M. (1988) 2nd eds. Extreme value precipitation. In hydrology in practice. Van
Nostrand Reinhold (international) Co. Ltd.
Sikes, H.L., (1939). Notes on the geology of the country surrounding Nairobi.
Singh, V.P. (1995) (Ed.). Computer models of watershed hydrology. Water resources
publication. Colorado. United Nations Environment Programme.
Sombroek, W.G, Braun H.M.H and Van der Pouw B.J.A. (1982).Exploratory Soil
map and Agro-climatic zone map of Kenya.
SPSS (2005). (IzttplllV"'VW.Spss.com)(spss 11.5) brief guide.spss inc. 233 South wacker
drive Chicago IL.2002.( accessed 12006)
USDA (1986). Urban hydrology for small watersheds, Technical release 55,
Springfield, VA.
124
USGS (1994).Water resources data for Colorado water year 1994. Data report 94-1
Waananen, J.T., Limerinos W.J. (1997). Flood - prone areas and land use planning.
Geological Survey PaperNo.942. W.J kockelman, US.A. San Francisco Bay region.
Ward, R.e. and Robinson M. (1990). Urban Hydrology. Principles of Hydrology 3rd
Ed. McGraw-Hill Book Co. (UK) Ltd
Wisler C.O. and Brater E.F. (1959). 2nd (Bd). Floods. Water resources engineering.
McGrawl-Hill Book Co. (UK) Ltd.
WilIiams, L.A. (1967). Geology for Nairobi city and region. Hydraulic Engineering,
University of Michigan. John Wiley &Sons. Inc.
Zigg, A.W. (1940). Degree and length of landscape as it affects soil loss in runoff.
Agricultural Engineering journal 2:59
125
APPENDICES
Simulated Observed
1175 920
778 720
670 650
580 545.25
250 245.19
245 240.88
147 146.35
136.18 122.15
112 111.66
111.45 110.23
104.67 104.23
100 97.5
92.89 87.92
78.54 76.75
67 ..26 65.89
55 51.35
52 49.32
48 46.21
25.43 20.55
21.98 19.02
19.75 17.25
17.43 16.5
12.56 12.38
10..22 9.7
127
Q H Log Q Log H
920 10.46 2.9638 1.0195
720 9.39 2.8573 0.9723
650 8.98 2.8129 0.9533
545.25 8.84 2.7366 0.9284
245.19 5.84 2.3895 0.7664
240.88 5.8 2.3818 0.7634
146.35 4.66 2.1654 0.6684
122.15 4.3 2.0869 0.6335
111.66 4.13 2.0479 0.616
110.23 4.11 2.0423 0.6138
104.23 4.01 2.018 0.6031
97.5 3.9 1.989 0.5911
87.92 3.72 1.9441 ·0.5705
76.75 3.5 1.8851 0.5441
65.89 3.28 1.8188 0.5159
51.35 2.94 1.7105 0.4684
49.32 2.9 1.693 0.4624
46.21 2.8 1.6647 0.4472
20.55 1.96 1.3128 0.2923
19.02 1.9 1.2792 0.2788
17.25 1.82 1.2368 0.2601
16.5 1.78 1.2175 0.2504
12.38 1.57 1.0927 0.1959
9.7 1.42 0.9943 0.1523
128
0
0.3
1.5
2.3 0.65 0.65 0.5 0 0.01 0
8.3 (3BA29 RGS)
13.3
14.3
18.3
0
1.5
2.3
6.3
11.3 0.65 0.65 0.5 700 0.002 1
13.1·
14.6
14.9
0
1.5
3.0
4.5 0.65 0.65 0.5 2150 0.01 2
6.0
7.0
10.0
12.5
0
1.5
4.0
7.5
9.5 0.65 0.65 0.5 450 0.01 3
13.0
14.5
16.0
0
0.5
0.9
3.4
5.4 0.65 0.65 0.5 2000 0.01 5
7.4
7.9
8.6
0
0.8
2.3
5.3
129
0
0.85
2.6
5.1 0.65 0.65 0.5 1250 0.002 7
8.0
8.7
12
12.9
130
Appendix 5
.u = 104.591
131
specifications.
-Enter control specifications for start date and time and stop date and time.
Incase of optimization:
-Move to tools
-save
-View results; flow comparison graph, scatter graph, residual graph and objective
Function graph
131
specifications.
-Enter control specifications for start date and time and stop date and time.
Incase of optimization:
-Move to tools
-save
-View results; flow comparison graph, scatter graph, residual graph and objective
Function graph
132
NETWORK IN NAIROBI
Questionnaire No .
Date .
ORGANIZATION .
Name (optional) .
Sex: male/female
BACKGROUND INFORMATION
(i) Artificial
(ii) Natural
(ii) Sewage
(iii) Combined
(e) Does the storm water contribute to natural hazards like floods, blockage of pipes
(i) Yes
(ii)) No
(f) Is the drainage system efficient in handling the above hazard? Explain .
(i) Good
(ii) Moderate
(iii) Poor
(i) Yes
(ii) No.
(i) If the answer is 'Yes' please suggest the kind of improvement you would like to
see .
(j) Will this improvement increase the efficiency of the current 'System? Please
comment. .
2 Drainage activities
(iv) None
(b) Does the system have any problems of the conveyance during and after storms?
(i)Yes
(ii) No
(c) If the answer is 'Yes' what would you attribute this to? Please explain .
(i) Yes
(ii) No
(f) If your answer is 'No' what would you suggest can be done to cater for the future
(i) Yes
(ii) No
Will this address all the inadequacies of the current drainage system ?
(a) What drainage methods are existing for the urban land areas in Nairobi?
(b) What are the conditions for the method you have chosen?
(i) Poor
(ii) Good
(c) Would you please comment on the answer you have chosen? .
(g) What would you comment on the amount of investment required both material
(i) What reasons would you give for the answer you have given? .
U) Could lack of proper urban land drainage be a factor contributing to debris and
4 Highway drainage
(a) What drainage facilities are there for the city road network?
(b) What would you comment on the effectiveness of the available system?
(ii) Effective
(d) What would you attribute to water which pools on the roadway or on the lands
(e) Does this inconvenience the city residents and in any ways? Please comment
(f) What measures would you suggest to be put in place to improve this kind of a
(a) Who is responsible for the maintenance of drainage system in the city?
(b) Does the city have any maintenance program in the calendar of the years?
(i) Yes
(ii) No
(i) Major
(ii) Periodic
(iii) Routine
(f) If the answer is no, what measures would you suggest should be installed to
(g)Will this be adequate to cater for the problem of drainage before and after
every storm? .
138
Frequencies
Notes
IOutput Created 24-0CT-200613:15:34
IComments
I Filter I <none>
If-w-e-i-g-ht------I <none>
'Input
ISPlit File 1 <none>
N of Rows in
22
Working Data File
. Missing Value Definition of Missing User-defined missing values are treated as missing .
IHandling Cases Used Statistics are based on all cases with valid data.
FREQUENCIES
VARIABLES=drainsym; orwater; draietf; tlotype;
floharzd; exdra ;impdra; impro;
inceff ;nrbdrtyp problms; procause ;ameliora
;sysadeq; tutuexp; systpl ;reexpan;
curdradq ;urbladra; condra; concom; impmetho;
Syntax
suggimp ;effinc; invest; priodran;
reason ;Iackdra; longdra ;avasyeff ;reas watpools;
inconvin; imprmes; respman;
mainpro; mansched; maincho; main mode; instsmot
;enoistal;
IORDER= ANAL YSIS .
Elapsed Time 0:00:00.08
IResources Total Values Allowed 149796
I
Frequency Table
DRAINSYM
ORWATER
1.00
, Valid 3 13.6 13.6 13.6
3.00 1 4.5 4.5 18.2
\
4.00 2 9.1 9.1 27.3
139
,
12.00 1 4.5 4.5 31.8
,
14.00 1 4.5 4.5 36.4
,
123.00 7 31.8 31.8 68.2
I
DRAIEFF
FLOHARZD
1
Frequency Percent Valid Percent Cumulative Percent
EXDRA
IMPDRA
IMPRO
I
Frequency Percent Valid Percent Cumulative Percent
INCEFF
NRBDRTYP
PROCAUSE
AMELlORA
SYSADEQ
FUTUEXP
SVSTPL
I
I 1.00
Frequency
20
Percent
90.9
Valid Percent
90.9 90.9
Cumulative Percent
,
2.00 1 4.5 4.5 95.5
,Valid
11.00 1 4.5 4.5 100.0
REEXPAN
CURDRADQ
I
1.00 18 81.8 81.8 81.8
I
12.00 13 113.6 113.6 195.5
I Valid 11-1-23-.-00-1-1----+14-.-5--+1-4-.5----+-11-0-0-.0------
URBLADRA
CONDRA
142
CONCOM
IMPMETHO
.Valid
, 1.00 21 95.5 100.0 100.0
Missing System 1 4.5
,
_T_o_t_a_I ~1_2_2 1_1_0_0._0
__ ~ ~ _
SUGGIMP
Total 1100.0
EFFINC
PRIODRAN
143
,
Frequency Percent Valid Percent Cumulative Percent
LACKDRA
LONGDRA
AVASYEFF
144
REAS
r- I_T_o_ta_I
__ TI2_1 9_5_.5
__ -+1_10_0_._0 1---------------
"M
__is_s_in_g_I'--s_y_s_te_m-+I_1 4_.5 +1 1 _
INCONVIN
86.4
, 1.00 19 90.5 90.5
I
,Valid
12.00 11 14.5 14.8 195.2
3.00 1 4.5 4.8 100.0
,
Total 21 95.5 100.0
Missing System 1 4.5
!
Total 1100.0
RESPMAN
145
MAINPRO
I
2.00 7 31.8 31.8 54.5
I
13.00 16 27.3 127.3 181.8
IValid I23.00 1---+---I-----lI'------
1 4.5 4.5 86.4
I 1123.0013 13.6 1-13-.6------l-1-10-0-.0-------'
MAINCHO
MAINMODE
,
1.00 20 90.9 95.2 95.2
I
Valid 12.00 11 14.5 4.8 100.0
, I Total 121 195.5 100.0
I Missing 1System 1-1----+14-.-5--+-----+--------'
INSTSMOT
J
I
r -
I Percent
I Frequency Valid Percent Cumulative Percent