JOMO KENYATTA UNIVERSITY OF AGRICULTURE AND TECHNOLOGY

DEPARTMENT OF CIVIL, CONSTRUCTION AND ENVIRONMENTAL ENGINEERING

FINAL YEAR PROJECT PROPOSAL

TITLE: IMPACTS OF LAND USE CHANGES ON HYDROLOGICAL PROCESSES AND
IMPLICATION FOR HYDROPOWER POTENTIAL: CASE STUDY OF MAU CATCHMENT
AND SONDU-MIRIU PROJECT

KEVIN OCHOLA OBAMBO

E25-0111/04

SUPERVISOR
DR. M. O. NYADAWA

A Project submitted in partial fulfillment of the award of BSc. Civil Engineering of the Jomo
Kenyatta University of Agriculture and Technology
i

DECLARATION
I, Kevin Ochola Obambo do declare that this report is my original work and to the best of my knowledge,
it has not been submitted for any degree award in any University or Institution.

Signed______________________________________________ Date ____________

Kevin Ochola Obambo

CERTIFICATION
I have read this report and approve it for examination.

Signed_______________________________________________Date_____________
Dr. M. O. Nyadawa

ii

DEDICATION
This Design work is dedicated to my dear parents and entire family. Its through you that God has
provided for me in all ways up to this point in my life. Thank you for being there

iii

ACKNOWLEDGEMENTS
I wish to thank my supervisor Dr. M.O.Nyadawa for his guidance and assistance at every stage of this
project, Eng. Wangai for his help on the watershed model and Mr.J.K. Mwangi on his invaluable support
and advice during data collection and analysis.
I also wish to thank the Mr. Muriuki GEGIS dept. for providing some Landsat images, Mr. Lawrence
Okello of the RCMRD (regional centre of mapping for resource development) for advice on the GIS and
remote sensing methods and providing recent Landsat images.
I am appreciative of the assistance offered to me at the Meteorological department, on meteorological
data of the study area, the Chief Electrical Engineer Ministry of Energy, Eng. Muiru and the Director of
Operations KENGEN, Eng Richard Nderitu for assistance in obtaining generation data on the Sondu
miriu power project.
I am indebted to Ken Kemucie for his assistance in GIS related procedures during preparation of data.
I am grateful to Mr. Thooko and Mr. Njiru of WRMA (Water Resources Management Authority), for
providing streamflow data and the operation of the rating curve.
Finally, I would like to thank my friends and colleagues for assistance and encouragement.

iv

Impacts of land use changes on hydrological processes and implication for hydropower potential:
case study of Mau catchment and Sondu-miriu project

ABSTRACT
To demonstrate the relation in land use change to streamflow variability, a hydrological modelling at a
catchment scale has been used to investigate the impact of land use change on streamflow in Mau
catchment using a distributed deterministic modelling system. The model divides a catchment into
equally sized grid cells, providing the ability to impose changes in land use spatially. Model parameters
were established by calibration against measured discharge data available over the year 1986. The model
simulation-period was divided into calibration 1986 and validation 1995 periods. The model provided
relatively good fits between measured and simulated discharge both at a daily and monthly scales.
To illustrate the effect of land use change on streamflow, the challenge is to determine whether the
hydrological variability is as a result of change in land use/cover. This was done by keeping all other
datasets in the model constant and varying only land use / land cover data. This study provides an
approach to assess the actual changes in hydrologic response of a watershed to land use transformations
made in the past 30 years. The approach combines land sat image data from satellites from 1973- 2008 of
the SW. Mau catchment, climatic data and hydrometric data. In 1973, 262 km2 (42%) of the watershed
area was covered by natural forest. Due to continued encroachment into forest land, the cover decreased
to percent 212 km2 (34%) of the area in 2008. Most of the cover change occurred between (2002-2008)
i.e. from 237km2 (38%) in 2002 to 212km2 (34%) in 2008 a decrease of about 10%. Three main classes
of land use have replaced the closed canopy. These are: poor natural cover, agricultural areas and
residential areas. The hydrology of the Sondu River also has changed, with sharp increases in peak flows
and an earlier occurrence of these peaks between 1973 and 2008. It can be concluded that land use change
is a factor affecting the (changes in the) hydrologic regime of the Sondu river drainage.
An active management strategy aimed at the conservation and regeneration of the natural vegetation is
recommended, in order to improve the consistency of streamflow, during both dry and wet periods.
In order to use the results of this kind of model for decision making and water resources management, the
model should be tested under different environment and different scenario conditions. Rainfall
measurement and stage-discharge rating curves should be given priority to improve model performance.

Table of contents
DECLARATION .......................................................................................................................................... ii
CERTIFICATION ........................................................................................................................................ ii
DEDICATION ............................................................................................................................................. iii
ACKNOWLEDGEMENTS ......................................................................................................................... iv
ABSTRACT .................................................................................................................................................. v
LIST OF FIGURES ................................................................................................................................... viii
LIST OF TABLES ....................................................................................................................................... ix
LIST OF IMAGES ........................................................................................................................................ x
LIST OF ABBREVIATIONS ...................................................................................................................... xi
CHAPTER ONE ........................................................................................................................................... 1
1.0 INTRODUCTION .................................................................................................................................. 1
1.1 Background ......................................................................................................................................... 1
1.2 Study justification ............................................................................................................................... 3
1.3 Problem statement ............................................................................................................................... 3
1.4 Objectives ........................................................................................................................................... 4
1.5 Research hypothesis ............................................................................................................................ 4
1.6 Scope of study ..................................................................................................................................... 4
CHAPTER TWO .......................................................................................................................................... 5
2.0 LITERATURE REVIEW ....................................................................................................................... 5
2.1 Introduction ......................................................................................................................................... 5
2.2 Model overview ................................................................................................................................ 15
CHAPTER THREE .................................................................................................................................... 17
3.0 RESEARCH METHODOLOGY. ......................................................................................................... 17
3.1 Study Area ........................................................................................................................................ 17
3.1.1 Location ..................................................................................................................................... 17
3.2 Input data .......................................................................................................................................... 19
3.2.1 Remote sensing and GIS geo-data. ............................................................................................ 19
3.2.2 Meteorological data ................................................................................................................... 22
3.2.3 Hydrometric data ....................................................................................................................... 23
3.3 Data Analyses ................................................................................................................................... 24
3.3.1 Remote Sensing Data Processing ............................................................................................... 24
3.3.2 GIS Processing and Analysis ..................................................................................................... 24
vi

3.3.3 Simulation of streamflow with CEQUEAU ............................................................................... 25

CHAPTER FOUR ....................................................................................................................................... 29
4.0

RESULTS AND DISCUSSIONS ................................................................................................... 29

4.1 Comparisons of Streamflow discharge and generation of the power project ................................... 35
CHAPTER FIVE ........................................................................................................................................ 37
5.0 CONCLUSIONS AND RECOMMENDATIONS ............................................................................... 37
5.1 Conclusions ....................................................................................................................................... 37
5.2 Recommendations ............................................................................................................................. 37
REFERENCES ........................................................................................................................................... 38
APPENDICES ............................................................................................................................................ 40
Appendices 1: STEAMFLOW DATA 1986 (sample) ............................................................................ 40
Appendices 2: sondu miriu power generation ......................................................................................... 42
Appendices 3:METEOROLOGICAL DATA ......................................................................................... 43
Appendices 4:The parameters file lists of vectors of the Sondu river .................................................... 44
Appendices 5: Changes of river flow response with change land use/ land cover data .......................... 45

vii

LIST OF FIGURES
Figure 1: Sondu Miriu power project layout ................................................................................................. 2
Figure 2: Schematization of the principal components of an integrated flow model ................................... 6
Figure 3: Schematization of the relations between the model and the system's environment ...................... 7
Figure 4: General layout of model .............................................................................................................. 16
Figure 5: Geographical location of the Sondu river basin .......................................................................... 17
Figure 8: Upper catchment of Sondu river within the Southwest Mau area ............................................... 18
Figure 6: General layout of the entire Sondu river basin ............................................................................ 18
Figure 7: General layout of the entire Sondu river basin ............................................................................ 18
Figure 9: General structure of the CEQUEAU parameters optimization module ....................................... 26
Figure 10: Model user interface .................................................................................................................. 28

viii

LIST OF TABLES
Table 1: Characteristics of the images used in the study ............................................................................ 19
Table 2: Meteorological data ...................................................................................................................... 22
Table 3: Values of rating curve ................................................................................................................... 24
Table 4: List of parameters that can be optimized ...................................................................................... 27
Table 5: Lists of vectors of the parameters file of Sondu river. .................................................................. 31
Table 6: Change of river flow response with change land use/ land cover data ......................................... 33

ix

LIST OF IMAGES
Image 1: Large scale destruction of forests and bamboo forests in South West Mau forest reserve .......... 14
Image 2: Land sat image P182R60 year 1973, showing watershed box ..................................................... 20
Image 3: Land sat image P169R60 year 1986, showing watershed box ..................................................... 20
Image 4: Land sat image P169R60 year 1995, showing watershed box ..................................................... 21
Image 5: Land sat image P169R60 year 2002, showing watershed box ..................................................... 21
Image 6: Land sat image P169R60 year 2008, showing watershed box ..................................................... 22

LIST OF ABBREVIATIONS
ArcGIS---------------------------Geographic information system software
DHSVM--------------------------Distributed Hydrology Soil Vegetation Model
ERDAS Imagine-----------------Remote sensing software
GIS--------------------------------Geographic Information System
GoK-------------------------------Government of Kenya
HRU-------------------------------Hydrologic response units
HSPF-----------------------------Hydrologic Simulation Program - Fortran
KenGen---------------------------Kenya Electricity Generating Company Ltd
KPLC------------------------------Kenya Power and Lighting Company
MIKE-SHE-----------------------Model Systeme Hydrologique European

n----------------------------------Mannings roughness
PRMS (MMS) -------------------Precipitation-Runoff Modeling System
SWAT----------------------------Soil-Water Assessment Tool
UBCWM--------------------------UBC Watershed Model
UNEP------------------------------United Nations Environmental Programme

WMS-----------------------------Watershed Modeling System

xi

CHAPTER ONE
1.0 INTRODUCTION
1.1 Background
Forests cover only about three per cent of Kenyas land area they provide the following ecosystem
services: they trap and store rain water; regulate river flows and prevent flooding; help recharge groundwater tables; improve soil fertility; reduce soil erosion and sediment loads in river water; help regulate
local climate conditions; and act as carbon reservoirs and sinks.
There has been progressive reduction in forest cover over the years as economic pressures lead to
encroachment into protected forest areas; this has been experienced in the Mau catchment which is the
largest closed canopy in Kenya about 400,000ha, it is also the catchment to most drainage systems
including, Sondu, Mara, Nzoia,Yala, Ewaso Nyiro rivers.
The Sondu river is one of the major drainage systems whose catchment lies within the Southwest Mau
region. It is one of the large river basins across the western Kenya region. The basin is considered one of
the important sub catchments of the Lake Victoria basin as the Sondu River is one of the six major rivers
that drain into the lake; it has a catchment area of 3470sq.km and a length of 200km. With the recent
commissioning of the Sondu Miriu power project the hydrology of the river has attracted attention due to
the low flows that have impeded the achievement of the power plant capacity. Hence focus has been on
the Sondu river drainage basin and there has been growing concerns regarding degradation in the rivers
watershed that lies in the Southwest Mau forest reserve, suspected to be affecting the flow regime and
discharge. The river originates from the western slopes of the Mau escarpment and flows through a
narrow gorge penetrating the Nyakach escarpment, it then meanders into Odino falls before entering the
flood plains of Nyakwere where it drains into the Winam gulf of Lake Victoria. The Sondu Miriu power
project is based on a run-of-river diversion from a weir structure on the Sondu River above Nyakach
escarpment. Accurate scientific data is required to forecast and avert any more degradation including
planning, and implementing appropriate mitigation measures on the river.
This study utilizes the CEQUEAU hydrological model, a distributed deterministic model for stream flow
simulation and geographical information system (GIS) to determine the extent of land use/cover changes
in the Sondu river watershed and analyze the effects of these changes on the river flows over the last three
decades.

In 2001, 61,023 ha of forest in the Mau Complex were excised including one quarter of Southwest Mau
forest reserve was excised, the Southwest Mau forest is the primary source of the Sondu River. Between
1973 and 2005, Maasai Mau Forest lost over 8,214 ha of forest within its official boundaries, which were
established to protect the forest. Almost 43 per cent of that loss occurred in just two years from 2003 to
2005.Forest loss in critical catchment areas for the Sondu, Mara, Molo, Naishi, Makalia Nderit, and Njoro
Rivers will result in ecological and hydrological changes, which threaten the sustainable future of areas
downstream. In addition, people have encroached into some 43,700 ha in the Mau Complexs remaining
protected forests. The hydrology of the Sondu River has also increased with sharp increases in the flood
peak flows. There is evidence of increased soil erosion in the upper catchments with silt built-up in the
downstream especially in the flood plains of Nyakwere where it drains into the Winam gulf of Lake
Victoria. (UNEP, 2009), Kenya: Atlas of Our Changing Environment.Division of Early Warning and
Assessment (DEWA), Kenya
The Sondu River is the lifeline of the power project at the Sondu town and the basin is considered one of
the more serene sub catchments of the Lake Victoria Basin and ultimately the Nile Basin. The basin also
contains forests, large-scale farms, smallholder farms, pastoral grazing lands and fishers. Hence there is
need for urgent action to stem land degradation of the Sondu river basin including planning and
implementing appropriate measures.

Figure 1: Sondu Miriu power project layout

1.2 Study justification

The catchments potential hydropower generation capacity is approximately 535 megawatts, which
represents 57 per cent of Kenyas current total electricity generation. To date, projects that have already
been developed, are under construction, or are proposed within the Mau catchment will generate only
about 190 MW. (UNEP, 2009), Kenya: Atlas of Our Changing Environment.Division of Early Warning
and Assessment (DEWA), Kenya
When environmental pressures are added to likely increases in fossil-fuel prices, increased electricity
generation from renewable sources, including hydropower appears to be attractive. Accordingly, global
hydropower production, which at present meets around one-fifth of global demand, is anticipated to
increase by three times by 2100. Despite this, new hydroelectric development will occur in a troubled
environment. Firstly, adverse environmental degradation of drainage basins has caused fluctuations river
discharges hence making it difficult to establish their trend. Secondly, while precipitation is anticipated to
increase on a global level, many parts of the world are anticipated to see significant drying. A wide range
of studies have found that river flows and consequently hydropower production are sensitive to changes
in precipitation and temperature. Declining hydroelectric production potential will be detrimental to the
economic viability of schemes, reducing financial returns, raising unit prices and, ultimately, making
investment in hydropower less likely.
By alleviation of the land use activities within the watershed we gain benefits from the lush vegetation i.e.
its montane forests are an important part of water-flow regulation, flood mitigation, water storage,
groundwater recharge, water purification, micro-climate regulation, and reduced soil erosion and siltation.
The forests also provide other major environmental services, including nutrient cycling and soil
formation. In addition, their role in storing carbon makes the Mau Forest globally important for mitigating
climate change.

1.3 Problem statement

Extensive land use/cover change has had an effect on the streamflow trends and variation in peak flows
and low flows. This affects the generation of power at the power project.
Extensive degradation and human encroachment have serious consequences both within the forest and
downstream in the form of water shortages, health risks, desertification, habitat destruction,
sedimentation, erosion and even alteration of the micro-climate.

1.4 Objectives
The basic objective of the study is to investigate the spatio-temporal information on the status of land use/
cover changes on the watershed and subsequent variability on the hydrological regime hence the changes
on the streamflow discharge.
The Specific objectives of the study include:

To process and display remotely sensed land use / land cover datasets at chosen time intervals
over 30 years in the Sondu river upper catchment in the Southwest Mau forest reserve.

To calibrate and validate the CEQUEAU model hence simulate streamflow and flow
characteristics of the river.

To ascertain whether or not changes in the flow regimes were attributable to land cover changes
in the basin.

1.5 Research hypothesis

The land use/cover change in the watershed has no effect in the Sondu river flows/discharge

1.6 Scope of study

The scope of the study entails simulating streamflow for different scenarios of land use and assessing
whether the cover change causes variation in streamflow trends and discharge.

CHAPTER TWO

2.0 LITERATURE REVIEW

2.1 Hydrological models overview
One of the earliest and very simple hydrological model is the quantitative expression of flood magnitudes
in relation to rainfall and watershed area known as the rational formula (Mulvaney 1851). Since then,
further progress in basic and applied hydrology has contributed to a better understanding of the terrestrial
components of the hydrological cycle. This, in turn, supported by an increase in computation capabilities,
permitted the development of more realistic hydrological models.
Today, many types of deterministic physical based watershed models are reported in the literature. These
models can be included in one of two main categories: streamflow-overland flow models and
groundwater flow models. In the surface flow models, the groundwater flow is partially treated and often
taken into account with a lumped parameter. This category is valid where groundwater does not constitute
an important part of the overall flow domain. The Stanford Watershed Model (Crawford and Linsley
1966), the SSARR model (U.S. Army Corps of Engineers 1975), the CEQUEAU model (Morin et al.
1981) and MDOR (Villeneuve et al. 1984) are a few examples of this type of model. Fleming (1979)
presents a good review of many surface models. Fleming, G. (1990). Deterministic models in hydrology,
FA0 irrigation and drainage paper, Rome 80 p.
The second category is used where groundwater resources dominate. In groundwater models, overland
and streamflow are treated as boundary conditions and may even be completely neglected. Many
groundwater models have been built and are widely used (Bachmat et al. 1980). The Prickett and
Lonquist model (1971), the model put forward by Trescott et al. (1976) and Newsam (Ledoux and Tillie
1980; Ledoux 1975) are a few examples.
However there is a watershed category where surface flow and groundwater flow are closely linked
together. In these watersheds, both the surface water and groundwater and their interactions are of a
significant importance and their proper management requires that both be studied simultaneously. It is
therefore necessary, in these cases, to use a hydrological model that simulates the conjunctive use of
surface water and groundwater, and their interactions, in a realistic way.
In using such a model one has the advantage of taking into account all water contributions to the
hydrological basin as well as the eventual interactions between the water cycle's terrestrial components.
This type of model constitutes an adequate tool for the efficient control of the water budget components
5

and has proved very useful to answer questions raised by regional water resources problems associated
with surface water and groundwater (Hansen and Dyhr-Nielsen 1983). For example, such a tool can
enable the study of the effects of aquifer abstraction on surface water regimen, particularly during lowwater periods (Miles and Rushton 1983; Refsgaard and Hansen 1982b; Besbes et al. 1981; Emsellem
1971).
Characteristics of Integrated Flow Models
The model cannot represent all the mechanisms involved in the hydrological cycle and, to avoid undue
complexity, must simplify reality by integrating only its principal components. The aim of modeling is to
estimate the total flow at the outlet of the watershed. It is recognized that total flow is a combination of
three processes: surface flow, interflow and base flow (Fig. 2). We are also interested in the eventual
interactions between these hydrological processes and the systems' environment (Fig. 3): exchanges with
the atmosphere, surface outflows, lateral contribution by surface water, exchanges between aquifers and
exchanges between the non-modelled surface areas and aquifers.

Figure 2: Schematization of the principal components of

an integrated flow model

However, in addition to the integration of the main processes and relations, the model should, for
economic and operational considerations, satisfy particular characteristics (Freeze and Harlan 1969;
Jacquet 1971). First, the model must take into account the system's physical properties. The model's
representation of the system must be based on the maximum amount of geological, geomorphological,
topographical and physiographical information. In addition, the functions used in describing the
hydrological processes must be derived from known physical law or include conceptual functions in
which parameters represent a physical reality. Because the model is indeed an image of the system's
physical structure, simulation results can be interpreted with an increased level of confidence.
Second, the model must represent the spatial and temporal variations in the system's structure and
behavior. This condition is relevant to the variability of hydrological phenomena and to the model's
usefulness as a tool for planning, development and management purposes. In most cases, modeling work
is not concerned with the conditions that prevail at the monitoring stations themselves:

Figure 3: Schematization of the relations between the model and the system's environment
(after Ledoux 1980).

The tool must therefore be helpful in extrapolating the watershed's response to points where there may be
particular needs.
Third, the model must be suitable for a vast range of hydrological and geographical conditions. In practice
this is desirable because it imparts a great deal of versatility with respect to the model's eventual
applications to different hydrological system configurations. The model must use a minimum number of
parameters for calibration and must incorporate simple relations to represent the hydrological processes
so that its usefulness is unhindered when going from one watershed to another.
7

Furthermore, it is preferable to use a model that can be easily adapted to data obtained at different levels
of discretisation at any time of the modeling process. This permits 1) to use the model as soon as the
minimum amount of data required to run the program is collected; 2) to update the model as new
information and data become available. This gives, for example, the modeler the possibility of
progressively defining areas where particular studies have to be undertaken to respond more accurately to
the objectives of the modeling study.
Since the model is to be applied by users who have not worked on its design, it is necessary as Bachmat et
al. (1980) point out, that the model be supplied with a set of relevant documents, including a complete
description of the model, a user's manual, print out copies of the program and modeling examples.
Existing Integrated Flow Models
Until now, much attention has been focused on management models that integrate the use of surface
water and groundwater with respect to economic, technological and political objectives (Bachmat et al.
1980). In watershed hydrology, the first attempts at modeling the interaction between surface and
groundwater flows were aimed at the study of the hydrologic cycle's mechanisms and to the development
of new tools to improve hydrologic methods and research. Pinder and Sauer (1971) were the first to
present a model that treats alluvial-aquifer-stream interactions to calculate the modification of a flood
wave due to bank storage effects. Freeze (1972) introduced a deterministic mathematical model that
couples three-dimensional transient saturated-unsaturated subsurface flow and one-dimensional gradually
varied unsteady channel flow to study the role of subsurface flow in generating storm runoff.
In practice, usable models have been built for various objectives according to two approaches. The
general approach includes models conceived and designed to be applied to watershed of variable sizes
and characteristics. The Knapp et al. (1975) model is a very good example of an integrated watershed
flow model designed to be general in practice. However, it should be noted that it requires the estimation
of a relatively large number of parameters (- 40). Another example is the Cunningham and Sinclair (1979)
coupled flow model involving the simultaneous numerical solution of the surface and subsurface flow
equations using the Garlekin finite-element method. It includes specific methodologies worked out by
Pinder and Sauer (1971).
The second approach is more specific to a particular watershed's physical characteristics. With this
approach, model application is restricted to watershed where the general configuration and the physical
characteristics are similar. The main objective of the model reported by Refsgaard and Hansen (1982a)

was to predict the hydrological consequences of groundwater abstraction on the river discharges and on
the hydraulic heads of the aquifers for the SusA watershed in Denmark.
Hydrological model- is a conceptual or physically based procedure for numerically simulating a process
or processes which occur in a watershed.
Modelling is the process of implementation of the model in simulating the actual phenomena
Most of the knowledge on the effects of land use change on catchment runoff comes from experimental
catchment studies, statistical methods and hydrological modelling. Some of models used include:

DHSVM
The Distributed Hydrology Soil Vegetation Model (DHSVM) is a physically based model developed at
the University of Washington to represent the effects of topography and vegetation on water fluxes
through the landscape (Wigmosta et al.1994; Wigmosta et al. 2002).
Model Description
The DHSVM is a fully distributed hydrologic model that subdivides the model domain (typically the
watershed or group of watersheds under investigation) into small computational grid elements using the
spatial resolution of an underlying digital elevation model (DEM). It is typically applied at high spatial
resolutions on the order of 10 to 100 m (Beckers and Alila 2004) for watersheds of up to 10 000 km2 and
at sub-daily time scales (best results are obtained at hourly time steps) for multi-year simulations. The
model includes two vegetation layers (forest canopy overstorey and understorey) and multiple soil layers.
Model Applications
The DHSVM has been widely applied in the PNW in a forest hydrology and management context, for
assessing the effects of forest roads on flood flows, for hydrologic forecasting, and for assessing the
potential effects of climate change, including but not limited to the following studies:

HBV-EC
The HBV (Hydrologiska Byrns Vattenbalansavdelning) model is a conceptual hydrological model
designed for use in mountainous environments. It was originally developed in the early 1970s at the
Swedish Meteorological and Hydrological Institute (SMHI), and has been used extensively for
hydrological forecasting, engineering design, and climate change studies (Lindstrm et al.1997). The
HBV-EC model, reviewed here, was adapted by Environment Canada and UBC (Moore 1993) to better
represent glacier processes. Several additional improvements have been made since, as described below.
9

Model Description
The HBV-EC model uses the Grouped Response Unit (GRU) concept to group DEM/GIS grid cells into
bins having similar land cover, elevation, slope, and aspect (Stahl et al. 2008). The model incorporates
two soil layers, no explicit vegetation layer, and operates at a daily time step

HSPF
The Hydrologic Simulation Program-Fortran (HSPF13) was designed by the US Envionmental Protection
Agency (EPA) to simulate a broad range of surface and sub-surface hydrologic and water quality
processes in watersheds (Bicknell et al. 2001). The HSPS model is based on the Stanford Watershed
Model and can simulate one or many pervious or impervious unit areas discharging to one or many river
reaches or reservoirs.
Model Description
HSPF is a lumped parameter hydrologic model that can simulate all primary natural hydrological
processes (Bicknell et al. 2001). The unsaturated zone is approximated using a single storage reservoir,
while explicit representation of vegetation in the model is limited. The model runs at any time step, from
1 minute to 1 day, which can be divided equally into 1 day.
Model Applications
Many of the published applications of the HSPF model have been in the central and east portions of
theUSA (Pike 1995). In British Columbia, data from Carnation Creek, on the eastern coast of Vancouver
Island, have been used in a test calibration for using HSPF in forested watersheds (Hetherington et al.
1995). An important finding of this study was that forest cover in the HSPF is represented as a single
index, making the representation of differing forest cover types very limited

MIKE-SHE
The integrated hydrological modelling system, MIKE-SHE17, simulates the major processes in the
hydrologic cycle and includes process models for evapotranspiration, overland flow, unsaturated flow,
groundwater flow, channel flow, as well as modelling how all these interact. The MIKE-SHE model was
developed, starting in 1977, as consortium of three European organizations (Systme Hydrologique
10

Europen, [SHE]) (Abbott et al. 1986a, 1986b) and the full model is currently distributed by the Danish
Hydrological Institute (DHI).

Model Description
MIKE-SHE is a fully distributed, grid-based hydrologic modelling system that incorporates multiple soil
layers and a single vegetation layer (i.e., overstorey and understorey are not distinguished) and flexible
model time step.

Model Applications
MIKE-SHE has proven valuable in hundreds of research and consultancy projects covering a wide range
of climatic and hydrologic regimes, many of which are referenced in Graham and Butts (2005).

PRMS (MMS)
The Precipitation-Runoff Modeling System (PRMS) is a modular-design, deterministic, distributed
parameter modeling system developed by the USGS23 to evaluate the impacts of various combinations of
precipitation, climate, and land use on streamflow, sediment yields, and general watershed hydrology
(Leavesley et al. 2005). To better allow for model updates and adding process components, the
architecture and modular structure of PRMS were redesigned and formed the basis for the Modular
Modeling System (MMS) (Leavesley et al. 2005).
Model Description
Watersheds are divided into HRUs based on characteristics including slope, aspect, elevation, vegetation
type, soil type, land use, and precipitation distribution (Leavesley et al. 2005). The model incorporates
a single vegetation layer, two conceptual soil zones (recharge zone and a lower zone), subsurface and
groundwater reservoirs, and operates at a daily time step.

RHESSys
RHESSys24, the Regional Hydro-Ecological Simulation System, is a GIS-based, modelling framework
that integrates water and biogeochemical cycling and transport over spatially variable terrain at small
(first order streams) to medium (fourth- and fifth-order streams) scales (Tague and Band 2004). RHESSys
combines both physically based process models and a method for partitioning and parameterizing the
landscape (Tague and Band 2004).
Model Description

11

RHESSys uses a hierarchical approach to partition a landscape into hydrologically distinct units,
including patches (essentially HRUs, areas with similar soil and land use characteristics), hillslopes
(portions of a watershed that drain to a single point or stream reach), climate zones and watersheds (to
organize stream routing). Multiple vegetation strata can be included in the model, while soil water storage
is conceptually divided into an unsaturated and a saturated zone (Tague and Band 2004). The model
operates at a daily time step

SSARR
The Streamflow Synthesis and Reservoir Regulation (SSARR26) model was developed by the US Army
Corps of Engineers, North Pacific Region to provide mathematical hydrologic simulations for systems
analysis as required for the planning, design, and operation of water control works (US-Army Corps of
Engineers 1991). The SSARR model has been further developed for operational river forecasting and
river management activities (US-Army Corps of Engineers 1991).
Model Description
The model is semi-distributed and works with elevation bands for snowmelt calculations. It includes a
single soil layer and a single vegetation layer (i.e., overstorey and understorey are not distinguished). The
model can operate at time intervals ranging from 0.1 to 24 hours

SWAT
The Soil Water Assessment Tool (SWAT) is a watershed-scale model developed to predict the impact of
management on water, sediment, and agricultural chemical yields (Gassman et al. 2007).
Model Description
In SWAT, watersheds are divided into multiple sub-watersheds, which are then subdivided into
hydrologic response units (HRUs) (Gassman et al. 2007). The HRUs are areas of homogeneous land use
and soil characteristics, and represent percentages of the sub-watershed area that are not considred in a
spatially explicitly manner in the model simulation (Gassman et al. 2007). The model considered two soil
layers (root zone and unsaturated zone), together with conceptual shallow and deeper aquifer stores, a
single vegetation layer, and operates on a daily time step.
Model Applications
Gassman et al. (2007) reviewed historical SWAT development as well as a total of 115 model
applications in the US, Europe, and worldwide. These studies illustrate the versatility of SWAT in

12

simulating the hydrology of diverse watersheds. Generally, this review illustrated the ability of SWAT to
replicate hydrologic processes at a variety of spatial scales on an annual or monthly basis. However,
model performance has been inadequate in some studies, especially when comparisons of predicted
output were made with time series of measured daily flow. These shortcomings in simulating watershed
hydrology at relatively short time scales (days) is attributed to the simplified representation of runoff
processes with the SCS curve number method. One drawback of the HRU approach is that explicit spatial
representation of riparian buffer zones, wetlands, and other features is not possible.

UBCWM
The UBC Watershed Model (UBCWM) was originally developed for daily streamflow forecasting on the
Fraser River system in British Columbia (Quick et al. 1995). The UBCWM is being used operationally
for forecasting daily streamflows in sub-watersheds of the Fraser River system that are subject to
snowmelt floods from the mountain snowpacks of the Coast, Columbia, and Rocky Mountains in British
Columbia.
Model Description
The model was originally designed for forecasting runoff from mountain watersheds and, for this reason,
the model is divided into area-elevation bands (Quick et al. 1995). The model incorporates a conceptual
soil moisture storage reservoir and a single canopy layer

WaSiM-ETH
The Wasserhaushalts-Simulations-Modell (WaSiM-ETH30) is a fully distributed, physically based model
for estimating climate change impacts for subalpine and alpine regions. Compared to PREVAH,
WaSiMETH is a physically based runoff model and has more flexibility in separating surface runoff from
interflow, allowing, in general, a better reproduction of flood events (Gurtz et al. 1999). The application
of WaSiM-ETH to climate change studies in cold mountainous terrain has been reviewed by Werner and
Bennett (2009).
Model Description
The spatial and temporal discretizations of the model can accommodate grid cell sizes of virtually any
dimensions (from centimeters to several kilometers), and time steps between 1 minute to several days
(Werner and Bennett 2009). The model includes a layered soil model and layered vegetation

13

The Mau Complex forms the largest closedcanopy forest ecosystem of Kenya, as large as the forests of
Mt. Kenya and the Aberdares combined. It is the single most important water catchment in Rift Valley
and western Kenya. Through the ecological services provided by its forests, the Mau Complex is a natural
asset of national importance that supports key economic sectors in Rift Valley and western Kenya,
including energy, tourism, agriculture and water supply. The Mau Complex is particularly important for
two of the three largest foreign currency earners: tea and tourism. UNEP (2009), Kenya: Atlas of Our
Changing Environment. Division of Early Warning and Assessment (DEWA), Kenya

Image 1: Large scale destruction of forests and bamboo forests in South West Mau forest reserve

14

2.2 CEQUEAU Model overview

The background of model development is well documented in CEQUEAU Users Manual (Morin et al.
,1998). Relevant excerpts are reproduced here to assist readers understand the history of the model.
In the last decades, the CEQUEAU model has been used for many watersheds, varying from 1 to 100,000
km2 in the Province of Quebec, in Canada and elsewhere in North and South America. It is also used in
Europe and Africa. In the Province of Quebec, it was applied on some sixty watersheds and used to
determine probable maximal floods (PMF) in many watersheds of northern Quebec. The CEQUEAU
model is presently used on a regular basis by some institutions in the Province of Quebec to forecast flow
rates in real time. The accuracy of CEQUEAU model was tested in comparison with other well known
hydrological models in the world in the framework of two inter-comparisons of hydrological models
fostered by the World Meteorological Organization (WMO). The first inter-comparison, (WMO, 1986),
was dealing with the simulation of flow rates including snowmelt. The second inter-comparison, (WMO
1992) dealt with forecasting of streamflows in real time.
Many of the datasets involved in these processes are raster grids. The spatially distributed nature of the
raster grids used in these processes pointed to the adoption of a customizable geographic information
system with excellent raster functionality. The ArcView 9.x GIS series was adopted for the
implementation because it provided a visual, customizable development environment with excellent
support of raster operations. The hydrological model comprises two main parts and provides an improved
description of the eventual water flow towards the exit of the watershed.
The first part deals with the vertical flow, the principal related phenomena being: rain, evapotranspiration, infiltration and variations of near-surface and deeper water reserves. This first part is
labeled: "production function" and is calculated on each "whole square" and for each temporal step.
The second part deals with water flow transfer in the drainage network. The processes included in this
second part take into account the influence of lakes, marshes, artificial waterworks such as dams,
deviations, canals, water intakes, etc. This part is labeled "routing function" and requires the use of
"partial squares".
CEQUEAU requires five types of datasets for its operation. These are:
o

precipitation at the meteorological stations

maximum and minimum air temperatures at the meteorological stations

Observed streamflow for the calibration period.

Land use/ land cover

15

Digital Elevation Model data (DEM)

Detailed model operation is schematically illustrated in Fig. Fundamental hydrologic equations are Water
balance, Runoff Curve Number method and Muskingum channel routing method.

Figure 4: general layout of model

16

CHAPTER THREE
3.0 RESEARCH METHODOLOGY
3.1 Study Area
3.1.1 Location
The study area shown in the figure below is the Southwest Mau forest reserve a part of the expansive Mau
forest, located southeast of Kericho town. It forms the upper catchment of the Sondu River that drains to
Lake Victoria. The western flanks of the Mau forest complex are influenced by Lake Victoria
macroclimatic region and are generally wetter with rainfall greater than 2000mm and more evenly
distributed with a mean annual temperature minimum of 12 C maximum of 16 C. The soil of the Mau
forest is rich volcanic loam.
The watershed outlet is taken at the Itare gauging station 1JA04 the gauging station used for calibration
of the model is Kiptiget 1JA02 whose position is 35 15E
upper catchment of the river in the Southwest Mau forest reserve.

Figure 5: Geographical location of the Sondu river basin

17

0 33 S and delineated upwards to the

Figure 7:
6: General
general layout
layoutof
ofthe
theentire
entireSondu
Sonduriver
riverbasin
basin

Figure 8: Upper catchment of Sondu river within the Southwest Mau area

18

3.2 Input data

The input data required for the study included remote sensing and GIS data, stream flow data and
meteorological data. This data was available from various sources which included RCMRD (regional
centre of mapping for resource development), WRMA (water resources management authority) and the
Kenya Meteorological Department (KMD).

3.2.1 Remote sensing and GIS geo-data.

Table 1: Characteristics of the images used in the study

Type of sensor

period

Bands available

Scene

Spatial resolution

Landsat TM

1973

1-5 and 7

P182r60

28.5m

Landsat ETM

1986

P169r60

1995

P169r60

2002

1-5 and 7

P169r60

2008

28.5m

P169r60

The data set consisted of the following formats;

1. Raster Data:
Landsat image year 1973, 1986, 1995, 2002 and 2008

Years 1973, 1986 and 1995 Thematic Mapper scanners

Years 2002 and 2008 Enhanced Thematic Mapper scanners

Landsat Spectral Resolution 7 bands image type

Spatial Resolution 28.5 metres

Landsat 1986, 1995, 2002 and 2008 image was in path 169 and row 60

Landsat 1973 image was in path 182 row 60

These were in Datum WGS 1984 but different projected coordinate system zones.
2. GPS points of:

the areas Meteorological stations

Kiptiget Gauging station

3. Scanned Topographic-sheets number 131-2, 131-1, 117-4 and 117-3. These were in the datum
Arc 1960.
4. Vector Data Set-Sondu digitized rivers. This was in the datum Arc 1960.

19

Image 2: Land sat image P182R60 year 1973, showing watershed box

Image 3: Land sat image P169R60 year 1986, showing watershed box

20

Image 4: Land sat image P169R60 year 1995, showing watershed box

Image 5: Land sat image P169R60 year 2002, showing watershed box

21

Image 6: Land sat image P169R60 year 2008, showing watershed box

3.2.2 Meteorological data

Table 2: Meteorological data
STATION NAME

STATION NO.

GRIDS

KARINGET FOREST STATION

9035324

29S

35

38E

KAPKATET AGRI. OFFICE

9035323

37S

35

12E

SAINO FOREST STATION

9035298

21S

35

31E

NDOINET FOREST STATION

9035292

27S

35

29E

KERICHO NGOINA ESTATE

9035261

33S

35

3E

KERICHO TIMBILIL

9035244

21S

35

21E

ELBURGON,BARAGET FOREST STATION

9035241

25S

35

44E

SOTIK, TENWIK MISSION

9035079

45S

35

22E

BALLOBRACH,CHEMELIL

9035053

4S

35

9E

KERICHO KABIANGA H. SCHOOL

9035044

26S

35

8E

HAIL RESEARCH STATION KERICHO

9035279

22S

35

16E

22

Meteorological data were obtained from the Kenya Meteorological Department. The data included mean
daily temperature and rainfall data of the years 1973, 1986, 1995, 2002 and 2008.
The mean daily temperatures were obtained from the Kericho meteorological station, station number
9035244 while precipitation data were obtained from volunteer stations within the watershed their station
numbers are 9035044, 9035053, 9035079, 9035241, 9035244, 9035261, 9035292, 9035298, 9035323,
9035324 and 9035330.
Apart from the meteorological data the grid positions of the stations were obtained to map out their
positions in the watershed area. The mapping out is carried out using GIS methods hence the geographic
coordinate system of the stations is very important as alignment and accuracy problems may arise unless
there is correct transformation between geographic coordinate systems.
The digitized topographical sheets used for watershed delineation used the Arc 1960 coordinate system
while the stations used the WGS1984 coordinate system, this information assisted in the accurate
positioning of the meteorological stations.
The meteorological data provide was very consisted for the study periods hence minimized errors of
inconsistent data. The stations are;

3.2.3 Hydrometric data

Hydrometric data was collected from the water resources management authority (WRMA) at Sondu and
Kiptiget gauging station along the Sondu River.
The outlet of the Southwest Mau catchment is at the confluence of Itare and Chepkoisi tributaries at the
Itare gauging station 1JA04, but since this gauging station had inconsistent data with regard to observed
streamflow a smaller watershed was delineated further upstream at the Kiptiget gauging station 1JA02
that had reliable data to enable calibration of the calculated streamflow levels and the observed
streamflow levels.
The data obtained was in form of gauge heights hence required a rating equation for conversion to
discharge. The discharge rating equation gives the relationship between the discharge and the stage. Each
rating curve segment is defined by the equation
Q = C*(H-Ho)n ..eq.(i)
C, n, Ho, are constants

H is the gauge height

23

Table 3: Values of rating curve

Constant

Value provided

20.6718

1.2786

Ho

0.3

Hmin

0.37

Above a Hmin, Each segment applies from Hmin until Hmin of the next segment or until Hmax of the rating
curve it belongs to. In our case we had only one segment. The constants are all given. All that is required
is to input the water level (H) for each day and get the discharge (Q) from the equation.
MS excel spreadsheet was utilized in the conversion from gauge heights to discharges.
The gauging stations grid positions were also noted for positioning in the watershed area using GIS their
geographic coordinate system is WGS1984 hence required conversion to Arc 1960 for accuracy in
mapping them out.
The observed streamflow data had inconsistencies but was sufficient enough for calibration purposes.

3.3 Data Analyses

3.3.1 Remote Sensing Data Processing
Landsat images came with a spectral resolution of 7 bands. The images came with proper Geo-referencing
with the datum as WGS 1984. In ERDAS Imagine remote sensing software, the bands 2 (Green), 3
(Red) and 4 (Near IR, infrared) were layer stacked to come up with a false color composite image. These
composite depicts vegetation as red and is easier to pick put the forest while digitizing.
The images of 1973 required some spectral enhancement which was done by slightly adjusting bands
histograms to get a clearer image. This was done in ERDAS Imagine remote sensing software.

3.3.2 GIS Processing and Analysis

This was all done in ArcGIS 9.3 GIS software. In ArcMap all the data was projected to datum WGS
1984 UTM (Universal Transverse Mercator) zone 36N (Kenya falls in sector and sections 36 and 37 N
and S). This is because the area of study falls in that region.

24

All the shape files used were created and attribute set up in Arc Catalogue
1. Shape files of the area of study i.e. Sondu river watershed was created and digitized from the
scanned topographic sheets (ArcMap).
2. The rivers in the watershed were clipped from the provided data set of rivers (ArcMap).
3. Watershed box was also delineated to cover the entire area of study (ArcMap).
4. Using the Landsat images as the template and the watershed box, the land use classification of the
area was done. Forest area was digitized in all the years to give forest cover of 1973, 1986, 1995,
2002 and 2008. This was also done in ArcMap.
5. The GPS points of meteorological and Kiptiget gauging stations were displayed and projected to
WGS 1984 UTM Zone 36N.
6. In ArcMap a grid dividing the entire watershed box into equal regions was created.
7. Map composites consisting of the grid, clipped rivers, watershed box, delineated watershed,
meteorological stations, Kiptiget gauging station and each of the years forest cover were
developed in ArcMap.

3.3.3 Simulation of streamflow with CEQUEAU

3.3.3.1 Calibration of the model
The application of the CEQUEAU model to a watershed requires the adjustment of the parameters so that
the model reproduces the observed streamflows to the best extent possible. The adjustment of the
parameters of the CEQUEAU model is done by trials and errors or by optimization.
3.3.3.1.1 Adjustment of the parameters by trials
In this case one carries out a first trial, and then modifies the parameters and analyzes the new results to
determine if one must continue to modify the parameters and in what direction. The trial also allows one a
familiarization with the interactions of the model parameters, i.e. to know the direction and the amplitude
of the modifications of the simulated hydrograph brought by the modification of parameter(s). The
procedure to adjust the parameters varies from a watershed to the other; one may however determine the
necessary steps and the general rules.
To obtain the first simulation, the parameters and option file of the model must be prepared. The value of
the parameters for the first trial may be determined in the following way:

The constants are determined using the hydrological and physiographical characteristics of the
watershed under study (e.g.: average latitude, concentration time, etc...)
25

The parameters relating to the physical phenomena are fixed by studies external to the model
(e.g.: snowmelt parameter, coefficient of correction of precipitations as a function of the altitude,
etc...)

As for the other parameters, no fixed rule is used to determine their value; only experiment and
knowledge of their interactions provide the better adjustment. To determine the values of these
parameters for the first trial, one can take the values used for the simulation of neighboring
watersheds.

The verification of the adjustment of the model is done by analysis of numerical criteria and the analysis
of the results presented on graphs. Then if necessary, some parameters are modified and the simulation is
redone. This process is repeated until satisfactory results are obtained.
To facilitate the adjustment of the parameters and to allow the analysis of the accuracy of the simulations,
the CEQUEAU model computes some numerical criteria and can produce several types of graphs
showing the observed and calculated streamflows. These numerical criteria and the possible types of
graphs will be presented in the section on analytical results
3.3.3.1.2 Adjustment of the parameters by optimization
In the Tools menu of Project window, the CEQUEAU model makes it possible to launch the optimization
program. This program allows finding the value of some parameters of the model, while maximizing or
minimizing an objective function in order to reproduce the observed streamflows with minimal error.
The optimization algorithm (BOTM) based on M.J.D. Powell's method (1964) is used. The main program
and an account of the method are given in Optimization Techniques with FORTRAN (1973).

Figure 9: general structure of the CEQUEAU parameters optimization module

The program makes it possible to optimize simultaneously twenty parameters chosen amongst a total of
twenty-eight

26

Table 4: List of parameters that can be optimized

Nbr.
1
2
3
4
5
6
7

Name
CIN
CVMAR
CVNB
CVNH
CVSB
XINFMA
HINF

8
9
10
11

HINT
HMAR
HNAP
HPOT

12
13

HSOL
HRIMP

14
15

COEP
EVNAP

16
17
18
19
20

TRI
XAA
XIT
COET
EXXKT

Description
Infiltration coefficient in the LOWER ZONE reservoir
Drainage coefficient of the LAKES and MARSHES reservoir
Coefficient of low draining of the LOWER ZONE reservoir
High draining coefficient of the LOWER ZONE reservoir.
Low draining coefficient of the UPPER ZONE reservoir.
Maximum infiltration (mm/time step).
Infiltration threshold of the UPPER ZONE reservoir towards the
LOWER ZONE reservoir (mm).
Intermediate draining threshold of the UPPER ZONE reservoir (mm).
Draining threshold of the LAKES and MARSHES reservoir (mm).
Threshold of high draining of the LOWER ZONE reservoir (mm).
Threshold of water taken away at a potential rate by evapo-transpiration
(mm).
Height of the UPPER ZONE reservoir (mm).
Depth of water required to begin the runoff on impervious surfaces
(mm).
Correction of the precipitations according to altitude (mm/meter/year).
Fraction of the evaporation-transpiration taken in the LOWER ZONE reservoir (from 0.0 to
1.0).
Fraction of impervious area on whole squares (from 0.0 to 1.0).
Exponent of Thornthwaite formula.
Value of Thornthwaite thermal index.
Correction of the temperatures according to altitude (oC/1 000 m).
Adjustment parameter of the transfer coefficients from a partial square
to the next, for a time step of one day.

Assessment of Model performance (R2)

The Nash coefficient (Nash and Sutcliffe 1970) is defined by:

=( )
( )

eq. (ii)

Where:
NTD: Nash Coefficient;
qoi, qoi and observed flows of day i;
qoi : averages of observed streamflows for the n days used for the calculation of the coefficient .
The Nash coefficient represents the ratio of the residual variance to the variance of the observed
streamflows. Its value is 1 when the simulated streamflows are identical to the observed streamflows. As
the difference between the calculated and observed streamflows increases, the coefficient decrease and
can even become negative.
27

Figure 10: model user interface

The calibration was carried out at the Kiptiget 1JA02 outlet as the gauging station had sufficient observed
streamflow data. The parameters were used to simulate the streamflow at the Itare gauging station that is
the outlet of the catchment.
The following simulations were conducted with CEQUEAU to evaluate the impact of the change in land
use on streamflow:
1. 1986 was considered as model calibration period and the same parameters used in 1995 as the
validation period.
2. To demonstrate the effect of change in land use/ land cover data on river flow time distribution; was
done by keeping all datasets in the model constant.
The CEQUEAU model was calibrated for 1986 using the model parameters such as curve number
(abstraction coefficient), available soil water capacity, and soil evaporation compensation factor were
adjusted interactively to get reasonable match between observed and simulated mean monthly stream
flow. Streamflow was simulated using the calibrated parameters for the validation period and results were
compared.

28

CHAPTER FOUR
4.0 RESULTS AND DISCUSSIONS

Figure 9: Watershed forest cover 1973

Figure 11: Watershed forest cover 1995

Figure 10: Watershed forest cover 1986

Figure 12: Watershed forest cover 2002

29

Figure 13: Watershed forest cover 2008

After classification of the land sat images the forest area was digitized to give the forest cover in
the watershed area vis--vis the non-forested area for every time period selected. The area is
subdivided into grids to enable spatial analysis.

Figure 14: Calibration using 1986 streamflow data

30

Figure 15: Validation using 1995 streamflow data

Table 5: Lists of vectors of the parameters file of Sondu river.

SIMULATION
NEIGE
OPTION
SOL1
SOL2
SOL3
SOLINITIAL
TRANSFERT
POSTMETEO KAPTATET
POSTMETEO NDOINET
POSTMETEO KERICHO
POSTMETEO KIMBILIL
STATIONFIC
CARTEFONTE
CARTENAPPE
CARTENEIGE
CARTEPLUIE
CARTESOL
DATERELEVE
RELEVE
RELEVE
SURFACE
SURFACE
EXECUTION

19860101
-1.
80
80
0.15
65.
.75
70.
0.001

19861231 365 0101

3.5
4.0
0
0
0
0.025
0.020
65.
250.
0.
.05
30.
250.
1.00
1
9035336
9035292
9035261
9035244
0
0
0
0

37
47
19860330
19860724
19860220
19860323
19860630
19860718
19860629
19861225
0
0
100. 125. 100. 125.
100.
100. 125. 100. 125.
100.
622.89
20.71
191.58
151.97

31

0803
-1.0
0
0.
50.
1.
5.663
11
15
14
15
0

10

28
1
1
-3.0
0.7
3
3
0
0.
0.35
60.
75.
30.
4530.
+14.
0.0

16
11
15
13
0

0
125.
100.
125.
88.02

1138.68
1188.72
1149.86
1273.56
0

125
85.43

1
1.0
24
10.
0.
0.

237.15

Mean monthly simulated streamflow and associated standard deviations at the basin output (1JA04) and
the Kiptiget (1JA02) sub basin was compared well with the observed data. However, simulated values
were lower to observed values at the sub basin stream gauge (1JA02). Monthly flow patterns compared
reasonably well with the pattern of rainfall data used in the simulation, since volunteer weather stations
were used for the entire basin, the spatial variability of rainfall was adequately represented
The purpose of calibration is to adjust the model parameters so that the model closely matches the real
system. The real system is represented by observed discharge data. Adjustment of parameters is essential
to overcome errors due to unknowns and, unmeasured quantities in the basin.
From available observed discharge data at 1JA02 complete datasets were only found in the years 1986
and 1995, hence the year 1986 was used for calibration. Fig.14 shows a comparison of observed
streamflow and simulated. There were larger disparities at high flows but the simulated hydrograph
matched the trend of the observed during low flows. Parameter adjustments resulted in a coefficient of
determination (R2) of 0.63. Coefficient of determination (R2) equals one for a perfect case where
simulated values equal. These results were used to simulate streamflow at the 1JA04 gauging station. The
same parameters were used to demonstrate the effect of change in land use on basin runoff response. The
reason for poor predictability by the model at high flows could be attributed to inaccuracies in
interpolated rainfall due to poorly distributed observation network. Because the correlation is good at low
flows it is reasonable to say that rain gauge positions are not dense enough to capture extents of spatial
rainfall accurately.

Figure 16: Hydrograph of 1973 rainfall on land use of 1973 and 2008

32

Table 6: Change of river flow response with change land use/ land cover data

33

This is expected as the rainfall pattern and magnitude are similar in the two cases leaving any cause of
shift to change in area of different classes land use / land cover, the model parameters are maintained
from calibration stage.
There was a general declining trend in the streamflow with reducing land use cover from 1973 to 2008
and this was coupled with a larger variability in streamflow values between minimum and maximum
discharges.

Figure 17: Comparison of streamflow trends of the Itare and Sondu gauging stations

The discharge at the Itare gauging station 1JA02 was compared against discharge at Sondu gauging
station 1JG03.

34

There is a general trend of reduction in power production against with reducing runoff at the Sondu
gauging station, though this wasnt very consistent as power production is influenced by several other
factors apart from discharge that were not considered in the scope of this research.

36

CHAPTER FIVE
5.0 CONCLUSIONS AND RECOMMENDATIONS
5.1 Conclusions
The objectives which were set for this study were generally achieved to a satisfactory extent taking into
account the constraints of data and available models. The results are summarized as follows:

1. Land sat images were processed and classified the cover change assessed; most of the decline
occurring between (2002-2008) about 11% then (1973-1986) 1.5 % , (1986 and 1995) 2.9%,
(1995-2002) 3%.
2. There has been a declining trend in streamflow discharge over the duration characterized by erratic
flows with high peaks and very low flow discharges, hence a high hydrological variability,
attributable to land use change.
3. Calibration of CEQUEAU model in Sondu basin is accomplished and results has shown a coefficient
of determination of 0.63.

5.2 Recommendations
1. Replanting of trees and agro forestry should be undertaken in the catchment. To improve the
reliability of the streamflow hence of power project.
2. A large watershed model should be used to assess the entire Sondu river catchment as in this study
we are only able to assess the sub catchment within the Southwest Mau area.
3. Gauging stations data should be regularly recorded as this was a major problem in the study since
most gauging stations had data inconsistencies.
4. Analysis of effect of land use / land cover on river regime should be attempted by high spatial
resolution images and close time intervals.

37

REFERENCES
Akotsi, E., Gachanja, M., Ndirangu, J. (2006). Changes in the Forest Cover in Kenyas Five Water
Towers 20032005. Department of Resource Surveys and Remote Sensing, Kenya Forests Working
Group. Nairobi. Kenya Forests Working Group.
Bachmat, Y., Bredehoeft, J., Andrews, B., Holtz, D., and Sebastian, S. (1980). Groundwater management:
the use of numerical models, Water resources monograph, series 5, American Geophysical Union,
Washington, D. C., 127 p
Besbes, M., Duermael, B., Ledoux, E., De Marsily, G., and Talbot, A. (1981). Soutien des Ctiages d'une
riviere par pompage dans la nappe qu'elle draine: mythe ou rCalitt? Bulletin du B.R.G.M. 2 (111).
Crawford, N. G., and Linsley, R. K. (1966). Digital simulation in hydrology: Stanford watershed model
IV. Dept. of Civil Engineering, Stanford University, tech. rep. no. 39, 220 p.
Fleming, G. (1979). Deterministic models in hydrology, FAO irrigation and drainage paper, Rome 80 p.
Freeze, R. A., and Harlan, R. L. (1969). Blueprint for a physically - based digitally - simulated hydrologic
response model, Journal of Hydrology, Vol. 9, 237-25
Landsat satellite images for the years 1973, 1986, 1995, 2002 and 2008.
Knapp, R. M., Green, D. W., Pogge, C., and Stanford, C. (1975). Development and field testing of a basin
hydrology simulator, Water Resources Research Vol. 11(6), 879-888.
Miles, J. C., and Rushton, K. R. (1983). A coupled surface water and groundwater catchment model,
Journal of Hydrology, Vol. 62, 159-177
MORIN, G. (2002). Chapter 13 CEQUEAU hydrological model. p. 507- 576. Singh, V. P. and D. K.
Frevert (Ed.), 2002. Mathematical Models of Large Watershed Hydrology, Water Resources Publications,
Highlands Ranch, CO, 80163-0026 USA, (2002) 914 p.
Mulvaney, T. J. (1851). On the use of self-registering rain and flood gauges in making observations of the
relations of rainfall and of flood discharges in a given catchment, Trans. Inst. Civil Engrs. Ir. (Dublin)
Vol. 4(2): 18.
Nash, J.E. and Sutcliffe, J.V. (1970). Riverflow forecasting through conceptual model. Journal of
Hydrology, 10: 282-290
Ododo, E., Olago, D., Ochola, W. Eds. (2006). Environment for Development: An Ecosystems
Assessment of Lake Victoria Basin. Nairobi, United Nations Environment Programme, Pan African
START Secretariat.

38

Refsgaard, C.J. ,1996. Parameterization, calibration and validation of distributed hydrological models.
Journal of hydrology.198:69-67
Siriwardena, L., Finlayson, B.L., and McMahon, T.A, 2006. The impact of land use change on catchment
hydrology in large catchments: The Comet River, Central Queensland, Australia. Journal of Hydrology
326, 199-214.
Trescott, P. C., Pinder, G. F., and Larson, S. P. (1976). Finite difference model for aquifer simulation in
two dimensions with results of numerical experiments. Techniques of water resources investigations of
the U. S. Geol. SUN. book, 7, chap. cl., 116 p.
UNEP (2009), Kenya: Atlas of Our Changing Environment. Division of Early Warning and Assessment
(DEWA), Kenya

39

APPENDICES
Appendices 1: STEAMFLOW DATA 1986 (sample)
1JG03 MIRIU SONDU- 3447'40.157"E 020'46.869"S
1JA02 KIPTIGET- 3515'24.655"E 033'2.832"S

C -20.6718 20.6718
n -1.2786 1.2786
HO - 0.3

Hmin-0.37

1-Jan-86

0.3

0.3

0.21451

1-Jan-86

0.3

0.3

0.21451

2-Jan-86

0.29

0.29

0.20541

2-Jan-86

0.29

0.29

0.20541

3-Jan-86

0.28

0.28

0.196398

3-Jan-86

0.28

0.28

0.196398

4-Jan-86

0.27

0.27

0.187474

4-Jan-86

0.27

0.27

0.187474

5-Jan-86

0.27

0.27

0.187474

5-Jan-86

0.26

0.26

0.178643

6-Jan-86

0.26

0.26

0.178643

6-Jan-86

0.26

0.26

0.178643

7-Jan-86

0.26

0.26

0.178643

7-Jan-86

0.26

0.26

0.178643

8-Jan-86

0.27

0.27

0.187474

8-Jan-86

0.27

0.27

0.187474

9-Jan-86
9-Jan-86

0.26
0.26

0.26
0.26

0.178643
0.178643
40

4.4343
05
4.4343
05
4.2461
99
4.2461
99
4.0598
93
4.0598
93
3.8754
31
3.8754
31
3.8754
31
3.6928
63
3.6928
63
3.6928
63
3.6928
63
3.6928
63
3.8754
31
3.8754
31
3.6928
63
3.6928

10-Jan-86

0.26

0.26

0.178643

10-Jan-86

0.26

0.26

0.178643

11-Jan-86

0.26

0.26

0.178643

11-Jan-86

0.26

0.26

0.178643

12-Jan-86

0.26

0.26

0.178643

12-Jan-86

0.26

0.26

0.178643

13-Jan-86

0.26

0.26

0.178643

13-Jan-86

0.26

0.26

0.178643

14-Jan-86

0.25

0.25

0.169905

14-Jan-86

0.25

0.25

0.169905

15-Jan-86

0.25

0.25

0.169905

41

63
3.6928
63
3.6928
63
3.6928
63
3.6928
63
3.6928
63
3.6928
63
3.6928
63
3.6928
63
3.5122
42
3.5122
42
3.5122
42

Appendices 2:
sondu miriu
power
generation
Date
(Month
)
Oct-07
Nov-07
Dec-07
Jan-08
Feb-08
Mar-08
Apr-08
May-08
Jun-08
Jul-08
Aug-08
Sep-08
Oct-08
Nov-08
Dec-08
Jan-09
Feb-09
Mar-09
Apr-09
May-09

TGU
(kWh)
5,563,987
7,201,000
11,395,000
5,344,000
3,403,000
13,404,000
29,572,000
35,549,000
38,741,000
30,016,000
40,777,000
40,432,000
41,948,000
40,251,000
30,507,000
14,295,000
8,600,000
5,135,000
17,333,000
36,816,000

(MWh)
5563.987
7201
11395
5344
3403
13404
29572
35549
38741
30016
40777
40432
41948
40251
30507
14295
8600
5135
17333
36816

Jun-09

27,039,000

27039

11464,000
16,404,000
21,113,000
22,755,000
16,582,000
18,442,000
39,092,000
21,862,000

14164
16404
21113
22755
16582
18442
39092
21862

Jul-09
Aug-09
Sep-09
Oct-09
Nov-09
Dec-09
Jan-10
Feb-10

42

Appendices 3: METEOROLOGICAL DATA

90350
58
90350
58
90350
59
90352
60
90352
60
90352
60
90352
70
90352
70
90352
79
90352
79
90352
79
90352
79
90352
79
90352
91
90352
91

KAMINJEIWET
SECONDARY SCHOOL
KAMINJEIWET
SECONDARY SCHOOL
LITEIN MISSION KERICHO
KOIWA ESTATE - KERICHO
KOIWA ESTATE - KERICHO
KOIWA ESTATE - KERICHO
KAPLONG GIRLS
SECONDARY SCHOOL
KAPLONG GIRLS
SECONDARY SCHOOL
KERICHO
METEOROLOGICAL
STATION
KERICHO
METEOROLOGICAL
STATION
KERICHO
METEOROLOGICAL
STATION
KERICHO
METEOROLOGICAL
STATION
KERICHO
METEOROLOGICAL
STATION
AROKET TEA ESTATE
SOTIK
AROKET TEA ESTATE
SOTIK

Precipitation;
daily total
Precipitation;
daily total
Precipitation;
daily total
Precipitation;
daily total
Precipitation;
daily total
Precipitation;
daily total
Precipitation;
daily total
Precipitation;
daily total

20
03
20
08
19
73
19
73
19
83
19
93
19
73
19
83

Precipitation;
daily total

19
73

Precipitation;
daily total

19
83

60.7

45

64.1

24
6.7

21
2.9

38
1.7

22
7.3

Precipitation;
daily total

19
93

145

126.
6

38.8

18
3.6

28
2.2

21
3

15
0.9

Precipitation;
daily total

20
03

16.2

25

191

33
5.6

34
2.4

15
1.6

12
9.1

Precipitation;
daily total
Precipitation;
daily total
Precipitation;
daily total

20
08
19
83
19
93

42.3

67.5

294

19
5

20
8.9

13
7.8

83
160.
7

91.3

79.5

20
3.3

57

82.
4

16
6.1
13
3

68.3

64.9

47.
6

18
1.5

26
3.2

44

43

90.6

49.6

151.2

30
6

27
6.4

13
3.7

70.2
147.
2
140.
4
109.
3
195.
2

48.4

416.8

45
2.7

17
2.3

19
0.7

70.4
225.
9

7.8

15
6

16
9.2

59.
3

8.1

19
4.5

35
0.1

14
0.1

66.6
154.
6

89.4

26
1.8

24
6.5

15
4.4

74.5

21
4.1

24
5.1

16
7.2

22.2

23.1

40.3

44
6

32
0

27
5

36.1

81.9

55.7

24
2.9

88.
3

67.
7

12
7.7
24
6.6
31.
8
88.
7
10
9.4
96.
7
19.
1
12
6.5
95.
2

Appendices 4: The parameters file lists of vectors of the Sondu river

SIMULATION
NEIGE
OPTION
SOL1
SOL2
SOL3
SOLINITIAL
TRANSFERT
POSTMETEO KAPTATET
POSTMETEO NDOINET
POSTMETEO KERICHO
POSTMETEO KIMBILIL
STATIONFIC
CARTEFONTE
CARTENAPPE
CARTENEIGE
CARTEPLUIE
CARTESOL
DATERELEVE
RELEVE
RELEVE
SURFACE
SURFACE
EXECUTION

19860101
-1.
80
80
0.15
65.
.75
70.
0.001

19861231 365 0101

3.5
4.0
0
0
0
0.025
0.020
65.
250.
0.
.05
30.
250.
1.00
1
9035336
9035292
9035261
9035244
0
0
0
0

37
47
19860330
19860724
19860220
19860323
19860630
19860718
19860629
19861225
0
0
100. 125. 100. 125.
100.
100. 125. 100. 125.
100.
622.89
20.71
191.58
151.97

44

0803
-1.0
0
0.
50.
1.
5.663
11
15
14
15
0

10

28
1
1
-3.0
0.7
3
3
0
0.
0.35
60.
75.
30.
4530.
+14.
0.0

16
11
15
13
0

0
125.
100.
125.
88.02

1138.68
1188.72
1149.86
1273.56
0

125
85.43

1
1.0
24
10.
0.
0.

237.15

Appendices 5: Changes of river flow response with change land use/ land
cover data
DEC. 1973 (day
334 to day 365)
1

2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
Totals

Simulated streamflow based on 1973

land use data at grid (4,3)
0.9391
20.8478
19.8620
4.7616
14.1871
18.2962
17.3210
15.1240
13.5029
11.1661
7.7112
3.2057
1.5560
1.0961
0.7904
2.0270
13.4863
13.2112
15.1060
9.4379
3.4039
1.4623
17.2312
3.4976
1.4596
1.5284
0.7463
3.3764
7.0998
5.5961
3.60774
252.6444 m3/s

45

Simulated streamflow based on 2008 land use

data at grid (4,3)
3.0865
14.1267
13.2012
10.3914
8.5122
14.1128
14.3223
13.4199
9.1542
7.0921
8.1728
7.2861
5.4562
5.9634
6.5178
4.1172
6.3753
12.2173
9.0636
5.6628
6.2193
7.3287
13.1191
4.2713
3.1720
3.5022
3.9862
6.1230
4.2599
3.3577
4.0110
237.5909m3/s