WWW - Itc.nl Library Papers 2010 MSC Wrem Munyao

USE OF SATELLITE PRODUCTS TO ASSESS WATER
HARVESTING POTENTIAL IN REMOTE AREAS OF AFRICA
A CASE STUDY OF UNGUJA ISLAND
ZANZIBAR
John Ngila Munyao

February, 2010
USE OF SATELLITE PRODUCTS TO ASSESS WATER HARVESTING
POTENTIAL IN REMOTE AREAS OF AFRICA
by
John Ngila Munyao
Thesis submitted to the International Institute of Geo-information Science and Earth

Observation in partial fulfilment of the requirements for the degree of Master of Science in
Geo-information Science and Earth Observation, Specialisation: Water Resources and
Environmental Management
Thesis Assessment Board
Dr. C.M.M Mannaerts Chairman

Dr. M.S Krol External Examiner, University of Twente
Drs. R. Becht (1st Supervisor / Member), WRS Department, ITC
Ir. A.M van Lieshout (2nd Supervisor / Member), WRS Department, ITC
INTERNATIONAL INSTITUTE FOR GEO-INFORMATION SCIENCE AND EARTH OBSERVATION

ENSCHEDE, THE NETHERLANDS
ii
Disclaimer
This document describes work undertaken as part of a programme of study at the

International Institute for Geo-information Science and Earth Observation. All views
and opinions expressed therein remain the sole responsibility of the author, and do
not necessarily represent those of the institute.
iii
Abstract
This study presents a methodology that can be easily applied to identify Rain Water
Harvesting (RWH) sites using freely available RS products and GIS for data scarce areas of
Africa. The potential of data integration (use of historical and near real time RS data, GIS and
hydrological modelling) to assess the potential of RWH in combination with analytic hierarchy
process (AHP) using spatial multi-criteria evaluation (SMCE) model as the GIS platform is
exploited.
The Integrated Land and Water Information System (ILWIS), a GIS software package is used
to derive all the key spatial layers that are used for various analysis. Input layers derived for
use in this model include rainfall, slope, soil groups, land use/cover, CN and runoff index with
a spatial resolution of 30 metres. RWH maps indicating spatial extents of suitable areas for
roof catchment (RC), Micro and Macro Catchment are the key outputs.
Soil Conservation Service (SCS) is used for runoff modelling at pixel scale. About 84% of the
total runoff is generated within flat and undulating slope classes. Masika rains (March to
June) contribute 64% while the Vuli rains (October to December) accounts for 20% of the
total annual runoff.
Based on the developed model, the RWH sites identified relative to runoff generating areas
produced 10.18 km2 suitable areas for roof catchment, generating 4.6 Million Cubic Metres
(MCM) which can meet 33% of the total annual water demand. 30% of the island is suitable
for micro-catchment RWH and 23% suitable for macro-catchment RWH representing a total
area of 44,000 and 35,000 hectares respectively.
Validation for micro-catchment RWH (based on existing and expert knowledge) shows that
10 % of the sites identified as suitable are unsuitable, 10 % in marginally suitable areas and
80 % within suitable and highly suitable areas. For macro-catchment RWH, 12% of the sites
are in unsuitable areas, 20 % in marginally suitable and 68 % within suitable and highly
suitable areas.
The capabilities of using RS, GIS and field data for identifying potential sites for RWH
technologies for decision making on development and management of RWH programmes is
well demonstrated.
The main constraint to the adoption of RWH could be associated with lack of knowledge
among the decision makers and the community on existing potential for RWH for the island.
RWH suitability maps developed in this study that give a clear indication of the spatial extents
and the existing potential can be a starting point for creating awareness among stakeholder
at the local and national scale.
Key words: Remote Sensing; GIS; AHP; SMCE; Roof catchment; Micro and Macro
catchment RWH
iv
Acknowledgements
I like to thank the almighty for his blessing throughout my period of study for all is possible if
we truly trust and believe.
I am grateful to the Dutch government through the Netherlands Fellowship Program (NFP)
for granting me the financial support to enable me pursue the studies. It would have been a
dream impossible without their support. I also do wish to extend my sincere gratitude to my
employer, Water Resources Management Authority (WRMA) for granting me the opportunity
to undertake my studies.
My supervisors Drs. Robert Becht and MSc. Ir. Arno .M. van Lieshout have given me
immerse support to ensure that my work conforms to the set standards. Their advice and
positive criticism are highly appreciated. Msc. Jeniffer Kinoti’s advice and constant
encouragement during the formulation of the proposal and thesis phase has been of great
help.
Useful comments and contributions were received from Dr. Ben Maathuis and Dr. C.
Mannaerts during the processing and analysis of the MPE rainfall product. My colleagues in
WREM were always supportive and I highly appreciate.
I would also like to thank all the lecturers who were involved in one way or another in
disseminating the knowledge, which I am now proud of.
I would like to extend my thanks to the following people and organisations in Zanzibar who
were very supportive during my field work.
 ZAWA and all the staff for their support in terms of Vehicles and other logistics
during the fieldwork
 Salum Rehami and Haji Khamis Fundi of Kizimbani Agricultural Research Institute for
providing valuable assistance especially on the soils of the island.
 Tanzania Meteorological and Irrigation department for providing some of the key data
used in this study.
 Special thanks go to Uledi M. Uledi a retired soil expert who sacrificed his time to
ensure that we got the necessary soil information that helped to greatly improve the
soil map of the study area.
 My fellow students who accompanied me for the fieldwork, for the brotherhood
exhibited during our stay in Zanzibar ( Mwale, Chou, Mtui and Haji)
Special thanks go to the Kenyan community in ITC who stood by me during times of sorrow
and happiness. I would also like to thank my immediate family members especially my father
and mother, my father in law and mother in law who understood the need for me to be away
for studies.
Above all I wish to thank my dear wife Lillian for taking up the role of father and mother for
our family during my period of study and doing the best for the family despite the many ups
and downs she encountered and endured. I will always be indebted to you dear for this
sacrifice.
v
Table of contents
1. INTRODUCTION ....................................................................................................... 12
1.1. General objectives .............................................................................................. 13
1.1.1. Specific objectives........................................................................................ 13
1.2. Research Questions ........................................................................................... 13
1.3. Hypothesis ........................................................................................................... 13
1.4. Thesis Outline ...................................................................................................... 14
2. LITERATURE REVIEW............................................................................................. 15
2.1. Rainwater harvesting concepts .......................................................................... 15
2.2. Rainwater harvesting potential assessments ................................................... 15
2.2.1. Application of Remote sensing and GIS .................................................... 15
2.2.2. Analytic Hierarchy Process (AHP) .............................................................. 16
2.3. Rainfall-Runoff modelling ................................................................................... 17
2.4. Study area ............................................................................................................ 17
2.4.1. Climate .......................................................................................................... 18
2.4.2. Topography................................................................................................... 18
2.4.3. Drainage Characteristics ............................................................................. 19
2.4.4. Soils and Geology ........................................................................................ 19
3. METHODOLOGY ...................................................................................................... 20
3.1. Conceptual framework ........................................................................................ 20
3.2. Primary data sets and field work........................................................................ 21
3.2.1. Primary data sets ......................................................................................... 21
3.2.2. Field work ...................................................................................................... 21
3.3. Application of RS and GIS .................................................................................. 22
3.3.1. Digital Elevation Model (DEM) Hydro-Processing .................................... 22
3.3.2. Analysis of Rainfall Distribution from rain gauges network ...................... 23
3.3.3. EUMETSAT MPE Rainfall Product ............................................................. 25
3.3.4. Thematic maps ............................................................................................. 26
3.3.4.1. Land use/Landcover ........................................................................................ 26
3.3.4.2. Soil map ........................................................................................................... 28
3.3.4.3. Slope map ........................................................................................................ 29
3.4. Rainfall-runoff modeling ...................................................................................... 31
3.4.1. SCS Curve Number Method Description ................................................... 31
3.4.2. Evaluating Curve Number for the study area ............................................ 33
3.4.2.1. Reclassification of landcover and soli map to hydrologic conditions .......... 33
3.4.2.2. Reclassification of Soil map to Soil Group .................................................... 34
3.4.2.3. Building up CN map ........................................................................................ 36
3.4.3. Determination of Runoff using Curve Numbers (CN) ............................... 37
3.5. Decision making and RWH site selection ......................................................... 39
3.5.1. Problem definition ........................................................................................ 39
3.5.2. Criteria identification and selection ............................................................. 40
3.5.3. Calculation of the relative weights .............................................................. 41
3.5.4. Assessing Consistency of Pairwise comparison ....................................... 45
3.5.5. Evaluation of Results ................................................................................... 46
3.6. Sensitivity Analysis.............................................................................................. 46
3.7. Selection of technological choices..................................................................... 47
4. RESULTS ................................................................................................................... 51
vi
4.1. Rainfall Analysis .................................................................................................. 51
4.2. Rainfall Runoff modelling.................................................................................... 54
4.2.1. Annual Runoff and Runoff Coefficient ........................................................ 54
4.3. Rainwater harvesting potential........................................................................... 55
4.3.1. Roof catchment ............................................................................................ 55
4.3.2. Micro Catchment Rainwater Harvesting .................................................... 57
4.3.3. Macro Catchment Rainwater Harvesting ................................................... 60
4.4. Validation of results ............................................................................................. 62
4.5. Sensitivity analysis .............................................................................................. 63
5. Discussions ................................................................................................................ 64
5.1. Rainfall ................................................................................................................. 64
5.2. Rainfall Runoff Modelling.................................................................................... 64
5.3. Rainwater Harvesting.......................................................................................... 66
5.3.1. Roof Catchment ........................................................................................... 66
5.3.2. Micro-Catchment RWH................................................................................ 67
5.3.3. Macro-Catchment RWH .............................................................................. 67
6. Conclusions and Reccommendations ..................................................................... 69
References......................................................................................................................... 71
Appendencies .................................................................................................................... 75
Appendix 1: Aerial photos of study area and points sampled during field work ....................... 75
Appendix 2: Accuracy assessments of land cover classification ............................................... 76
Appendix 3: Long term mean monthly rainfall ............................................................................. 76
Appendix 4: Table of Runoff Curve Numbers (SCS, 1986) ........................................................ 77
Appendix 5: Validation points Micro-Catchment RWH................................................................ 78
Appendix 6: Validation Points Macro-catchment RWH............................................................... 79
Appendix 7: Macro-Catchment RWH System selection (FAO, 1994) ...................................... 80
Appendix 8: Random Indices (RI) for n = 1, 2... 15 (Saaty, 1980) ............................................. 80
vii
List of figures
Figure 2.4-1: Map of Unguja Island .............................................................................................. 18
Figure 3.1-1: Conceptual framework for generating runoff coefficient and suitable RWH site 20
Figure 3.3-1: Interpolated DEM, Sink Filled DEM and Sink Areas ............................................ 22
Figure 3.3-2: Drainage and catchment areas derived from DEM .............................................. 23
Figure 3.3-3: Spatial distribution of rainfall stations .................................................................... 24
Figure 3.3-4: Spatial distribution long term mean annual rainfall............................................... 25
Figure 3.3-5: MPE Annual Rainfall Totals .................................................................................... 26
Figure 3.3-6: Classified Land cover/land use .............................................................................. 27
Figure 3.3-7: Soil texture map of the study area ......................................................................... 28
Figure 3.3-8: Classified slope map of the study area ................................................................. 30
Figure 3.4-1: Reclassified soil map to soil groups....................................................................... 36
Figure 3.4-2: Curve Numbers per 30 meters pixel size .............................................................. 37
Figure 3.4-3: Initial maximum storage values per 30 meters pixel size .................................... 38
Figure 3.5-1: Slope Standardization (Cost Function) for Micro-Catchment RWH .................... 44
Figure 3.5-2: Runoff Coefficient Standardization (Benefit function) .......................................... 44
Figure 3.5-3: Slope Standardization (Benefit function) for Macro-RWH ................................... 45
Figure 4.1-1: Plot of measured and MPE derived rainfall ........................................................... 52
Figure 4.1-2: Comparison of MPE rainfall retrieval and gauge data ......................................... 52
Figure 4.1-3: Adjusted MPE derived rainfall ................................................................................ 53
Figure 4.2-1: Annual runoff Depth (mm/year) .............................................................................. 54
Figure 4.2-2: Annual Runoff coefficient ........................................................................................ 54
Figure 4.2-3: Comparison of monthly rainfall and runoff volumes (MCM) ................................ 55
Figure 4.3-1: Roof Catchments suitability Map ........................................................................... 56
Figure 4.3-2: Micro-Catchment RWH Suitability.......................................................................... 57
Figure 4.3-3: Percentage area under different suitability classes .............................................. 58
Figure 4.3-4: Macro Catchment RWH suitability ......................................................................... 60
Figure 4.3-5: Percentage of area covered by different suitability .............................................. 61
Figure 5.2-1: Monthly Rainfall runoff relationships ...................................................................... 65
viii
List of tables
Table 3.2-1: Primary Data sets ..................................................................................................... 21

Table 3.3-1: Areas covered by different soil classes................................................................... 29
Table 3.3-2: Extents of land area under different terrain classes .............................................. 30
Table 3.4-1: Curve Number for hydrological soil cover complexes ........................................... 33
Table 3.4-2: Generic conditions for soil classification (according to the CN method). ............. 34
Table 3.4-3: Reclassification of soils to soil groups .................................................................... 35
Table 3.4-4: Percentage of areas under different soil groups .................................................... 36
Table 3.5-1: General criteria and constrains for Micro RWH...................................................... 40
Table 3.5-2: General criteria and constrains for Macro RWH .................................................... 41
Table 3.5-3: The fundamental scale of absolute numbers (Saaty, 2008) ................................. 42
Table 3.5-4: Weighting for Micro-Catchment RWH ..................................................................... 43
Table 3.5-5: Weighting for Macro-Catchment RWH.................................................................... 45
Table 3.7-1: Micro-catchment RWH systems .............................................................................. 47
Table 3.7-2: Typical daily rates of water outflows and seasonal water input in lowland rice:
Adopted from (Tuong et al., 2003) ................................................................................................ 48
Table 4.1-1: Correlation of measured rainfall and MPE product ................................................ 51
Table 4.3-1: Recommended basic water requirements for human needs (Gleick, 1996)........ 56
Table 4.3-2: Areas suitable for different Micro-catchment RWH technologies ......................... 58
Table 4.3-3: Possible Impoundment site and contributing catchment areas ............................ 61
Table 4.3-4: Storage required covering the dry season irrigation water demand and maximum
irrigable area (based of available runoff)...................................................................................... 62
ix
Plates
Plate 1: Level Bunds for Rice Farming (Paddy Rice) ................................................................. 48
Plate 2: In-stream water harvesting using a storage weir .......................................................... 49
Plate 3: Earth pan to harvest runoff for dry season irrigation ..................................................... 50
x
Abbreviations and Acronyms
AHP Analytic Hierarchy Process

ARC Antecedent Runoff Conditions
CI Consistency Index
CN Curve Number
CR Consistency Ratio
DEM Digital Elevation Model
EUMETSAT European Organisation for the Exploitation of Meteorological Satellites
FAO Food and Agricultural Organisation of the United States
GIS Geographic Information Systems
ICRAF International Centre for Research in Agro-forestry
ILWIS Integrated Land and Water Information System
KARS Kizimbani Agricultural Research Station
MCDM Multi-criteria Decision Making Methods
MDGs Millennium Development Goals
MPE Multi-sensor Precipitation Estimate
NDVI Normalised Difference Vegetation Index
NGO Non Governmental Organisation
RC Roof Catchment
RI Random Index
RS Remote Sensing
RWH Rainwater Harvesting
SCS Soil Conservation Services
SMCE Spatial Multi Criteria Evaluation
SMOLE Sustainable Management of Land and Environment
UNEP United Nations Environmental Programme
ZAWA Zanzibar Water Authority
xi
USE OF SATELLITE PRODUCTS TO ASSESS WATER HARVESTING POTENTIAL IN REMOTE AREAS OF AFRICA
1. INTRODUCTION
“Water is at the heart of Millennium Development Goals (MDGs) numbers 1, 3 and 7,
and is indirectly associated with the success or otherwise of all the other Goals. But for
Africa to meet the MDGs, bold and targeted actions are required in the water sector. To
address this, the African Water Vision for 2025 has set to develop the full potential of
Africa’s water resources for sustainable growth in the region’s economic and social
development, of which rainwater harvesting (RWH) and storage forms a major
component” (ICRAF, 2005).
In general terms RWH can be defined as the harnessing of rainwater that will normally
runoff for beneficial use in areas of water scarcity. Various methods exist that can be
used to harness rainfall key of which that have been applied include:-
• Micro-catchment rainwater harvesting

• Macro-catchment rainwater harvesting
• Large catchment flood water harvesting with storage in dams, pans and sub-
surface dams
RWH can be a measure to increase access to water for the vulnerable sections of the
society in arid and semi-arid parts in countries where water resources are scarce or
inaccessible.
The water harvested can be used for various purposes ranging from domestic,
livestock, agricultural production, industrial and groundwater recharge. A successful
implementation of RWH should integrate social-economic and environmental issues to
ensure sustainability and protect fragile ecosystems.
RWH may lead to increased food production through minimizing the risk of crop failure
during droughts and floods; avail more water for domestic and industrial use. At
watershed level anticipated benefits include recharge to groundwater systems and
improvement of environment. The results of rainwater harvesting in modification of the
ecosystems is clearly demonstrated by Vohland et al., (2009).
According to a report by Millennium Development Goals - MDG Centre, Nairobi Kenya

(2007) the following key issues that contribute to poor water access need to be
addressed in Zanzibar.
• Poor access to and availability of water due to inadequate water harvesting

infrastructure – water storage falls below 1700 m3/capita/year (international
accepted minimum).
• Extremely low agricultural production – averaging less than one ton per hectare due
to intra-seasonal dry spells and drought; this has been made more critical by
climate change and weather risk; these could be mitigated through supplementary
irrigation and in-situ RWH
• Poor management of rainwater e.g. flooding, erosion etc.
This research seeks to address, the key contributing factor which as outlined by ICRAF
and UNEP (2005) is the lack of tangible scientifically verified information that can be
12
used to identify areas where RWH can be applied. This was achieved by developing a
user friendly database with formats that can be easily updated, queried, managed and
utilized based on Remote sensing (RS) and Geographic information systems (GIS).
1.1. General objectives

This study explored the potential of data integration (use of historical and near real time
RS data, GIS and hydrological modelling) to assess the potential rainwater harvesting
sites in remote and data scarce areas.
1.1.1. Specific objectives

The followings specific issues are addressed:-
• Identify and map out the potential rainwater harvesting sites for Unguja Island.
• Determine effectiveness of integrating RS and GIS (data preparation and model
parameterization) with hydrological modelling to identify potential rainwater
harvesting site
• Identify data requirements (bio-physical and socio-economic) and structure of a GIS
based RWH potential identification model that can be applied locally.
1.2. Research Questions

This research will seek to answer the following questions:-
• Which historical and near real time satellite data products can be used to map out
RWH potential sites?
• How can integrating remote sensing, GIS and hydrological modelling be optimally
utilized to identify suitable RWH sites?
• Which is the best approach in assessing RWH potential site?
• How appropriate are the identified RWH sites for the specific technology?
1.3. Hypothesis
The validity of the following hypothesis is tested:-
• Available historical and near real time satellite data sets can be used to identify
potential RWH sites.
• Runoff available for storage can appropriately be modelled using available rainfall
runoff models in remote and data scarce areas.
• RWH sites and appropriate technologies can be optimally determined by integrating
RS, GIS and rainfall-runoff models.
13
1.4. Thesis Outline
The general purpose of this research is to develop methodology that can be easily
applied to identify RWH site using freely available RS data and GIS for data scarce and
remote area of Africa.
This thesis is presented in six chapters as outlined below:-
Chapter 1 gives an overview of the study area and outlines the key problem that forms
the basis of this research. The Research objectives, research questions, the hypothesis
and the thesis outline.
Chapter 2 reviews related works conducted in this field to gain insights on key
methodologies used that may be applicable to this research. A brief description of the
study area is also highlighted.
Chapter 3 presents the conceptual framework used in conducting the research, the
methodology used, field work data collection and analysis. It forms the basis of all the
other chapters.
Chapter 4 presents the results of analysis of the RWH potential using the analytic
hierarchy process (AHP). The suitability for both micro and macro catchment RWH is
presented.
Chapter 5 discusses the results obtained and their relevance to the study area.
Chapter6 outlines the conclusions and recommendations arising from this research.
14
2. LITERATURE REVIEW
2.1. Rainwater harvesting concepts
Zanzibar has experienced an increasing water demand in all sectors since early 1980s
according to Halcrow (1994). RWH can be used as a measure to increase water
availability for all sectors.
RWH in various forms has been traditionally practised throughout the centuries.
Diversions using spate flow from normally dry water courses (wadi) into agricultural
area in the Middle East form some of the earliest examples. Other examples include
the Negev desert (Evenari et al., 1971), the desert areas of Arizona and Northwest
Mexico (Zaunderer et al., 1988) and Southern Tunisia (Arnold et al., 1986).
The importance of traditional, small scale systems of rainwater harvesting in sub-

Sahara Africa has recently been recognised (Critchley et al., 1989). Simple stone lines
are used, e.g. Burkina Faso and Mali; earth bunding systems in eastern Sudan, Kenya
and the central rangelands of Somalia.
Rainwater harvesting for improved crop production has received great attention in the
1970s and 1980s mainly due to the widespread variability of rainfall with the associated
effects of crop failure or reduced yield and threat to livestock and human life in semi
arid and arid regions of Africa (Hatibu et al., 1999)
It is advocated that RWH holds the opportunity to contribute to the equitable, efficient
and sustainable use of water resources by alleviating temporal and spatial water
scarcity, providing water beyond the basic human needs and, hence enabling small-
scale productive activities (Kahinda et al., 2007). More emphasis is made on the
importance of social, economic, and environmental considerations when planning and
implementing RWH projects (Arnold et al., 1986) to ensure sustainability.
RWH technologies are flexible and can be adjusted to local circumstances and should
therefore be built according to the ecological characteristics of a particular region or
locality (Bancy et al., 2007).
2.2. Rainwater harvesting potential assessments

2.2.1. Application of Remote sensing and GIS
Diverse research methodologies using RS and GIS have been applied by different
authors to identify potential rainwater harvestings in remote and data scarce areas; in
most of these methods, thematic maps are derived from remote sensing data and
integrated in GIS to evaluate suitable sites for rainwater harvesting.
Remote sensing is of immense use for natural resources mapping and generating
necessary spatial database required as an input for GIS analysis. GIS is a tool for
collecting, storing and analyzing spatial and non - spatial data, and developing a model
based on local factors can be used to evaluate appropriate natural resources
development and management action plans. Both these techniques can complement
each other to be used as an effective tool for selecting suitable sites for water
harvesting structures (ICRAF, 2005).
15
In assessment of rainwater harvesting potential using GIS and RS, FAO(2003) outlines
six key factors that require to be integrated into a GIS framework in order to
successfully develop a suitable model for RWH. This include; rainfall, hydrology
(rainfall-runoff relationships), slope, land cover, soils (texture, structure, depth) and
socio-economics of the area under consideration.
Identifications of potential sites for construction of rainwater harvesting structures for

recharging groundwater in Bakhar watershed of Mirzapur District, Uttar Pradesh, India,
was conducted by Kumar, Agarwal and Bali (2008) through deriving various thematic
maps such as Landuse/Landcover, geomorphology and lineaments, etc, using remote
sensing. These layers along with geology and drainage were integrated using GIS with
some weighting using expert knowledge to identify sites for rainwater harvesting.
The application of GIS as an integrating tool to store, analyse and manage spatial
information and linking it to hydrological response models, to facilitate decision making
by providing catchment level identification, planning and assessment of runoff
harvesting sites has been applied by de Winnaar et al.,(2007).
Kahinda et al.,(2008) presented a methodology that enables water managers to assess

the suitability of RWH for any given area which incorporated social economic factors
which previous methodologies did not consider. These came out of the realisation that
the non-integration of socio-economic factors leads to failure of rainwater harvesting
projects. Using a combination of physical, ecological and socio-economic factors in-
field RWH and ex-field RWH suitability maps were developed.
2.2.2. Analytic Hierarchy Process (AHP)

The integration of Multi-criteria decision making methods (MCDM) with GIS has
considerably advanced the conventional map overlay approaches to the land-use
suitability analysis (Malczewski, 2004).
Analytic hierarchy process (AHP) is one of a GIS-based MCDM that combines and
transforms spatial data (input) into a resultant decision (output). The procedures involve
the utilization of geographical data, the decision maker’s preferences and the
manipulation of the data and preferences according to specified decision rules referred
to as factors and constrains.
Key considerations that are of critical importance in decision making as outlined by

Malczewski (2004) are; (i) the GIS capabilities of data acquisition, storage, retrieval,
manipulation and analysis, and (ii) the MCDM capabilities for combining the
geographical data and the decision maker’s preferences into uni-dimensional values of
alternative decisions.
AHP is a key decision making tool that was used in this study to assist in obtaining an
appropriate solution over suitability assessment for RWH. The process involved the
structuring of factors that are selected in a hierarchy starting from the overall goal to
criteria, sub-criteria, and alternatives in successive levels (Saaty, 1990).
Four steps are outlined by Saaty (2008) that are key in undertaking AHP in an
organized way in order to make a decision over alternatives. These are; definition of
the problem or issue to be considered, identify the goal which is the criteria that the
other elements usually the alternatives will depend on which should be at the top of the
decision making tree, develop a pairwise comparison matrix, weigh priorities for each
element with priorities obtained in the comparison matrix to obtain a global priority that
will form the basis of decision making for the alternatives at the bottom of the hierarchy.
16
Kinoti et al., (2006) used expert knowledge based multi-criteria evaluation process to
identify water harvesting systems in Tanzania. This study integrated various data such
as meteorological, terrain parameters and remote sensing to simulate runoff
generation. The runoff potential is determined by assigning weights and AHP is applied
as a decision support system to arrive at the final decision.
Integrating AHP in a GIS environment can be used to make decisions based both on
expert and indigenous knowledge and choose between alternatives. The weighting
assigned to the thematic layer vary from one site to the other hence may not be
replicated.
2.3. Rainfall-Runoff modelling

Rainwater harvesting is a hydrological intervention which can best be depicted through
hydrological models that are able to show directions of flow, runoff and run-on areas
and identify locations for impounding structures.
This can be achieved through appropriate extraction of the key hydrological parameters
in GIS based environment. The data required for input in the hydrological models are
currently obtainable through remote sensing techniques.
Gupta et al.,(1997) suggested the use of land cover information derived from remote
sensing satellite data in the form of the normalized difference vegetation index (NDVI)
to derive maps that are used as input to derive a modified Soil Conservation Service
(SCS) runoff curve number (CN). The derived CN is then used to model rainfall- runoff
relationships for a watershed/catchment.
The Soil Conservation Services (SCS) method has been widely applied to estimate the
surface runoff from a given rainfall event. This method is usually acceptable where the
rainfall amount from a given rainfall even exceeds 40 mm. This method has been
applied by de Winnaar et al., (2007) to determine the runoff available for in determining
the potential RWH potential sites for Thukela River Basin, South Africa.
The key parameters that can be used to derive the CN are the reclassified soil
categories based on the soil texture units and landcover to derive the final curve
number. The derived CN is then used to derived the runoff expected from a given
rainfall amount and hence the runoff index is developed (Senay et al., 2004).
2.4. Study area

The study area is Unguja which is the main Island of the two that form Zanzibar. The
Island has a total area of 1658 km2 and is located 40 kilometres off mainland Tanzania;
approximately bounded by co-ordinates 5 degrees and 6 degrees south latitudes and
along 39 degrees east longitude. The north- south extent of Unguja is approximately 85
kilometres with the east-west extent varying from 9 kilometres in the northern end to
about 35 kilometres in the south (Hettige, 1990). Figure 2.4-1 shows the extent of the
study area.
17
Figure 2.4-1: Map of Unguja Island
2.4.1. Climate
The Island is characterised by a bimodal rainfall pattern. The average annual rainfall of
the Island varies from about 1,200 mm along the east coast to more than 2,000 mm in
the central hilly part of the Island.
There are two distinct rain seasons locally known as Masika (long rains) and the Vuli
(short rains). The main rainfall season (Masika) starts in March with a peak in April
extending to June. The second rain period (Vuli) is between October to December.
Majority of rain falls during Masika rainfall with April-May accounting for 49% of the total
rainfall; the driest months known locally as Mchoo are July and August; this period
though, also receive some precipitation (Hettige, 1990).
Annual average temperature of the Island is 26oC, with maximum temperatures of 27 oC

occurring in January and minimum of 24 oC in July. Evapotranspiration varies between
4.4 mm in May to 5.8 mm per day in February with a mean of about 5mm per day.
2.4.2. Topography
Unguja is characterised by wide valley corridors, fault structures and residual hills
reaching a maximum of about 117 meters in the central parts of the Island.
Four main topographic systems are identified by Hettige M.L (1990) namely: marine,
ridge, coralline reef and alluvial systems. The ridge system has a varying elevation
with low elevation system ranging between 0-45 meters, low to medium (30 -70 meters)
18
and medium (45 -117 meters). The fourth system of the ridge system is defined by
isolated wedge shaped limestone outcrops in the coral rag regions.
The alluvial system is composed of open and closed corridors; plains; depressions and
basins. This system can be differentiates by their drainage patterns with open corridors
having unrestricted drainage while the closed corridors have a subsurface drainage.
The depressions and basins are characterised by flat areas with a blocked drainage
towards the sea.
2.4.3. Drainage Characteristics
Drainage is mainly westerly but predominantly subsurface apart from areas with heavy
clay soils.
In the ridges system, with underlying slowly permeable clay soils drainage channels
have developed with time along the slopes of Miocene limestone ridges, draining
directly to the sea and some minor rivers within the corridor valleys (FINNIDA, 1991;
Hettige, 1990) giving rise to dissected landscapes and a dendritic drainage pattern.
2.4.4. Soils and Geology
Surface geology of Unguja is characterised by a sequence of recent deposits (Q1);
quaternary formations (Q2); early quaternary deposits and Miocene limestone. Recent
deposits are found within the corridor zone and are composed of colluvial and alluvials.
The quaternary system consists mainly of terraced coralline reef formation.
Miocene limestone’s are in three classes differentiated by age and stratigraphy as M1,
M2 and M3. M1 is the most recent and consist of crystalline, reef and detrital limestone.
M2 is composed mainly of grey to white limestone with hard siliceous bands. M3 are
greyish to bluish green limestone's consisting mainly of marls clays and sandy clays
and can be found underlying the weathered M2 system (Hettige, 1990)
Soil types of the island largely depend on geological formation and variations are
associated with the parent material. Sandy Mchanga (sandy soils) is mainly found in
the Q1 formation. Within the M2 and M3 systems the Uwanda and Maweni kinongo
soils (loamy soils) are dominant. Kinongo soils (loamy soils) are a product of M1
weathering while the Kinamo (clay soils) are formed from the M3 system.
19
3. METHODOLOGY
3.1. Conceptual framework
The methodology used to determine the potential RWH site for the study area using RS
and GIS is as indicated in the flow chart figure 3.1-1.
Geological Aerial GeonetCAST

Soil map DEM
map photos MPE data
DEM hydro- Extraction of MPE

processing rainfall
Evaluation of
Extraction
Reclassification Extraction of Extraction of rainfall
of slope
soil map for CN landcover class catchment areas distribution
map
using GIS
Reclassification
Classification
of soil and land
of slope map
cover map
Building up CN
map for the study
area
Runoff Modelling
using the SCS
Curve Number
Method
Classified Classified
Classfied Runoff
land cover slope
Soil map coefficient
map map
Constrain(Buffer
Thematic layers Social economic
for settlement
Weighted overlays factors
and infrastructure
RWH potential
Maps
Figure 3.1-1: Conceptual framework for generating runoff coefficient and suitable
RWH site
20
3.2. Primary data sets and field work

3.2.1. Primary data sets
The primary data sets used in this study are presented in table 3.2-1.The data giving
the scale, spatial extent, data type and source of data was collected from different
sources during field work.
Table 3.2-1: Primary Data sets
Data set Year Scale Spatial Extent Type Source

Soil Map 1990 1:100000 Unguja Island Scanned map (Hettige, 1990)
Meteorological various Point 9 weather Daily totals for 2 Tanzania
data Measurements stations on the stations and Meteorological
Island monthly average Agency and
totals ZAWA
Otho-rectified July 2004 0.5 metres Unguja Island Geo-referenced SMOLE
aerial to may spatial
photographs 2005 resolution
Layer digitized 2009 Varying Unguja Island Geo-referenced SMOLE

from aerial vector format
photos in layers for
vector format Roads,
buildings,
streams ,
Unguja outline
and land use
Hydro- 1987 1:125000 Unguja Island scanned map ZAWA
geological
map
3.2.2. Field work

Field work was conducted during the months of September and October 2009. Random
sampling was undertaken considering soil variability, landscape, landcover and
topography to identify this relationship between the variable. Aerial photo for the study
area were used to aid in sampling and enable coverage of diverse landcover and soil
types.
Landcover/ land use and soil type (texture) were recorded for 191 sample points. The
key issues considered were; co-ordinates, dominant landcover/ land use, soil type and
texture; and the infiltration properties of the soils. Specific notes were made on sites
that were considered suitable for both micro and macro catchment rainwater
harvesting.
Soil types and texture were identified in the field with help from a soil scientist from
Kizimabani Agricultural Research Station (KARS). Social economic information was
also gathered to establish the community perception on RWH. Aerial photo used during
field work and points sampled are presented in appendix 1.
21
3.3. Application of RS and GIS
Application of the methods explored in the literature review is tested in deriving the
thematic layers that are the key inputs used to determine the potential sites for both
micro-catchment and macro-catchment RWH. The layers are processed using ILWIS a
freeware GIS/RS package that is accessible to most organisations.
3.3.1. Digital Elevation Model (DEM) Hydro-Processing
Aster Dem with 30m resolution tile number ASTGTM_S06E039 and
ASTGTM_S06E039 were downloaded (http://www.gdem.aster.ersdac.or.jp/.) and were
used for this study to derive the key hydrological parameters.
Digital image processing was performed to extract the DEM that is used for
hydrological processing using Integrated Land and Water Information System (ILWIS),
a GIS/RS package based on the approach developed by Maathuis et al., (2006).
Pre-processing of the DEM was performed to interpolate for undefined area using an
average filter of kernel size 5 by 5 before further processing to derive the catchments,
drainage and slope maps. The DEM was further analysed to remove pits (sinks) and
flat areas to maintain continuity of flow to the catchment outlets .Figure 3.3-1 shows the
Final Interpolated DEM, filled DEM (sinks free DEM) and the sinks area maps.
Figure 3.3-1: Interpolated DEM, Sink Filled DEM and Sink Areas
Steps used to delineate the catchment areas after the fill sinks operation are as
outlined below:
• Flow direction determination using the deterministic 8 model which determines

into which neighbouring pixel any water in a central pixel will flow. Parallel flow
22
correction is achieved through increasing the elevation of the flat area cell in
order to attain the desired drainage pattern (Garbrecht J et al., 1997)
• Flow accumulation to obtain the drainage pattern of the terrain which represents
the number of pixels contributing water to any outlet within the basin. The
outlets of the largest streams, rivers etc, acquire the highest values.
• The drainage network determined using a threshold of 750 contributing pixels;
otherwise a lower number of pixels indicate overland flow.
• Catchment extraction based on the derived drainage network and the flow
direction map and a minimum drainage length of 1000 metres.
• The final catchment map determined by merging the minor catchment extracted
using outlet point map based on the Strahler stream ordering. This operation
generates 26 catchments.
The final catchment and drainage maps generated are presented in figure 3.2-
2.
Figure 3.3-2: Drainage and catchment areas derived from DEM
In the eastern part of the island dominated by limestone outcrops (known as coral rag
region), subsurface flow is dominant. This region therefore has an undefined drainage
pattern due to absence of surface flow and high infiltration rates.
3.3.2. Analysis of Rainfall Distribution from rain gauges network

The rainfall gauges network in the island is sparse and has not been operational
continuously over the years. Rainfall point measurements have traditionally been used
to estimate rainfall for most regions in Africa. A dense network is required to estimate
accurately the spatial rainfall distribution for a given area.
23
Interpolation of point measurement is therefore necessary to estimate rainfall for areas

that are not covered by rain gauges (Goovaerts, 2000). Eight rainfall stations are used
for interpolation to surrounding areas that are not covered by the network. Figure 3.3-3
shows the location and distribution of the rainfall stations used.
Figure 3.3-3: Spatial distribution of rainfall stations
The eight rainfall stations area assigned attributes based on the long term annual
averages rainfall for interpolation using the moving average method. Due to the sparse
nature of the stations and a few years of continuous data, this method is preferred over
the kriging method that performs better when the data density is sufficient (Eischeid J
et al., 2000).
Moving average performs a weighted averaging on point values based on a specified

weight function and a limiting distance (ITC-ILWIS, 2001). The inverse distance
weighting method based on equation 3-1 was used.
Weight= ( )-1 3-1
Where:
d = = relative distance of point to limiting distance point.
D = Euclidean distance of point to limiting distance
24
Do = Limiting distance
n = weight exponent
The weight functions ensure that points close to the measurement receive higher
weight value than points which are farther away. The final rainfall map is developed
using equation 3-2.
Estimated value = 3-2
Where
Wi = Weight value for point i
Vali = Point value of point i
The interpolated rainfall map figure 3.3-4 of the study area is based on the long term
annual average measured rainfall, a limiting distance of 4 kilometres and a weighting
exponent of 1 to ensure a smooth interpolation. Appendix 2 gives the rainfall data used
for interpolation to spatially distribute point measurements over the entire study area.
Figure 3.3-4: Spatial distribution long term mean annual rainfall
More rainfall is received in the western part of the Island characterised by high
elevation than the east. There is a variation in the amount of rainfall received despite
the small extent of the Island with some areas receiving rainfall amounts as low as 700
mm. A maximum of about 1600 mm per annum is received in the central part of the
Island.
3.3.3. EUMETSAT MPE Rainfall Product
The Multi-sensor Precipitation Estimate (MPE) is a real-time instantaneous rain-rate
product which is derived every 15 minutes from the EUMETSAT’s geo-stationary
satellites. The product provides real rainfall rates and daily average precipitation mostly
for convective rainfall (Kidd et al., 2008). The product is suitable for use in Africa where
25
real time rainfall information is not readily available and the rainfall monitoring gauge for
meteorological and short-range hydrological applications is sparse or lacking.
MPE data with a spatial resolution of 3 kilometers for the period January 2007 to
December 2009 was downloaded and resampled to match the spatial resolution of the
base maps used in this study (30 meters). Bi-cubic resampling was applied since it
gives more reliable results compared to other methods.
MPE data is selected for use due to its ability to retrieve rainfall intensities for remote
area where no rainfall gauges exists hence enabling more representative retrievals for
hydrological modeling. Figure 3.3-5 outlines the annual total rainfall as derived from
MPE product for the years 2007 to 2009.
Figure 3.3-5: MPE Annual Rainfall Totals
3.3.4. Thematic maps

3.3.4.1. Land use/Landcover
Landcover/land use map was derived from aerial photos taken between March 2004
and May 2005 with a spatial resolution of 0.5 metres. Thematic mapping of the different
land cover/ land use classes was achieved through unsupervised classification and
visual interpretation owing to the high resolution of the RS data used. Automated
classification resulted to a high number of mixed pixels hence the choice of the two
methods applied.
Landcover class were determined based on land use and landcover classification
system for remote sensed data by James et al., (2001).This system was selected since
it uses the features of existing widely used classification systems that are amenable to
data derived from RS sources. To enhance the accuracy of classification 98 GCPs
(ground control points) are used. The accuracy of classification tested using 93
reference points collected during the field work survey.
26
Accuracy assessment of classification was performed using a confusion matrix which

compares the classification results with ground truth information. This is a simple cross-
tabulation of the mapped class label against what is observed on the ground or
reference data for a sample of cases at specified locations (Canters, 1997). Accuracy
or the degree of correctness of a map classification is considered unbiased if it gives a
accurate representation of the landcover (Foody, 2002) indicating the degree to which
the derived image classification agrees with reality.
This comparison gave a user accuracy of 82%, reliability accuracy of 85% and an
overall accuracy of 80%. The results of accuracy assessment are presented in the
appendix 3.
The classified landcover/use map figure 3.2.3 was resampled to 30 metres resolution
using the nearest neighbour algorithm to match all the other layers.
Figure 3.3-6: Classified Land cover/land use
Mixed shrubs is the dominant land cover class covering mostly the eastern part of the
Island which is mainly under flat coralline limestone referred to as the Coral rag region.
The western and central parts are occupied by different landcover classes and is
characterised by undulating terrain; it also represents different land use practices.
27
3.3.4.2. Soil map

The study area lacks an elaborate soil map that shows the variability of soil properties
among different land use systems and scales. The soil map of the study area was
digitised from national level soil map produced by Carton W.E,(1955) and improved
using the soil studies carried out by Hettige (1990) coupled with extensive sampling
during the field work.
Soil map figure 3.3-7 classified is based on the works by Carton (1955) and explained
Hettige (1990). They are the Kinongo soils which are mainly loamy soils; Mchanga
soils mainly sandy soils and Kinamo soils that are clayey soils. The eastern part of the
Island referred to as coral rag region is mainly covered by Uwanda and Maweni soils
that overlay the porous coralline limestone and are a sub-group of the Kinongo soils
(Hettige, 1990).
Figure 3.3-7: Soil texture map of the study area
28
The area covered by each soil class based on texture is presented in table 3.3.1.
Table 3.3-1: Areas covered by different soil classes
Local Class FAO classification Texture Area (Km2) % of total

Name Area
Shallow Kinongo Haplic and humic Loams, fine clay loam to 67 4
ferralsols clay
Deep Kinongo Rhodic ferralsols Heavy loams to clay, 139 9

sandy clay loams
Uwanda Mollic and Rendzic Sandy Loam To loam 310 20
Leptosols
Kinamo Eutric and calcic Humic clay loams to 64 4
vertisols,cambisol clays
Reddish Rhodic nitisols and Sandy clay loams to clay 99 6
Mchanga Haplic Acrisols loams
Gleyish Ferric and Gleyic Sandy to silty soils, 142 9
Mchanga acrisols, gleysol and Loams
Fluvisols
Sandy Mchanga Dystric Cambisols and Loamy sands to sandy 59 4
cambic Arenosols clay loams
Maweni Rendzic Leptosol Sandy Loam To loam 681 44
Total 1561 100
Maweni and Uwanda soils types cover about 60% of the total area and are
characterised by high infiltration rate since the overlay the parent coralline limestone
formation. Kinamo soils mainly derived from M3 geological formation and composed of
marls clays and sandy clays represent areas that are expected to generate more
runoff.
3.3.4.3. Slope map

The slope of a given area influences recharge and infiltration hence the amount of
runoff that is expected from the terrain. Technology suitability for different RWH options
highly depends on the slope of a given area.
Slope map was derived based on 30 metres pixel size Aster-DEM. A linear 5 by 5
gradient filter (DFDX and DFDY) was applied in the X and Y direction with a gain factor
of 0.083 (ITC-ILWIS, 2001). Filtering to resolve for undefined area was performed using
a 5 by 5 majority filter.
29
The derived slope map figure 3.3-8 is classified into 5 slope percentage classes based
on the FAO slope classification following guidelines by Allen et al., (1998). The FAO
slope class indicates the dominant relief or slope of a soil association.
Figure 3.3-8: Classified slope map of the study area
Areas under different slope classes are presented in table 3.3-2. The Island is mainly
covered by flat and flat to gently undulating slope classes representing 86.2 % of the
total area. For analysis of RWH harvesting potential the dominant slope percentage is
used which is based on FAO slope classification guidelines
Table 3.3-2: Extents of land area under different terrain classes
Dominant Fraction of
Area
Slope Definition Slope Total area
(km2)
Percentage %
Generally Flat 0-2% 996 65.0
Flat to Gently Undulating 2-8% 325 21.2
Rolling to hilly 8 - 30 % 83 5.4
Steeply dissected to Mountainous > 30 % 129 8.4

Total 1532 100
30
3.4. Rainfall-runoff modeling

Rainfall runoff relationships for the basin are considered using the SCS curve number
method. In undertaking hydrological modelling using remote sensing data in GIS
environment the SCS curve runoff model is largely suitable due to its reliance on land
cover parameters which can be extracted from RS (Senay et al., 2004).
This method has several advantages mostly based on its simplicity to apply and
acceptability; however the method is also associated with several disadvantages. This
method nevertheless is found to be more appropriate in the absence of accurate
hydrological and topographical data that is essential for runoff estimation (Senay et al.,
2004).
3.4.1. SCS Curve Number Method Description

SCS runoff curve number is a conceptual model whose main objective is to estimate
runoff depth from a rainfall storm based on the Curve Number parameter. The method
holds several advantages due to it’s; simplicity, predictability, stability, dependence on
one parameter and responsiveness to runoff producing watershed properties.
Associated disadvantages are; its marked sensitivity to CN, unclear description on how
to vary antecedent conditions, varying accuracy due to variation in biomass, lack of
provisions to account for spatial scale effects and the fixed initial abstraction ratio at 0.2
(Victor et al., 1996).
Runoff generation from a watershed is mainly due to both surface and near sub-surface
flow process key of which include ; Horton overland flow, overland flow, through-flow
processes, partial-area runoff direct channel interception. The curve number method
estimates direct runoff that combines channel runoff, surface runoff, and subsurface
flow (USDA, 2004).
Runoff curve number equation estimates total storm runoff from total storm rainfall and
this relationship excludes time as a variable and rainfall intensity. Its stability is ensured
by the fact that runoff depth (Q) is bounded between 0 and the maximum rainfall depth
(P). This implies that as rainfall amount increase the actual retention (P-Q) approaches
a constant value; the maximum potential retention (USDA, 2004; Victor et al., 1996)
The runoff equation relates runoff (Q) to precipitation (P) and the Curve Number (CN)
which is in turn related to storage (S). CN is based on the following parameters;
hydrologic soil group, land use and treatment classes, hydrologic surface conditions
and the antecedent moisture conditions.
31
Equation 3-3 known as the runoff curve number gives the relationship between the
parameters described above.
3-3
Where:
Q = depth of runoff, in inches
P = depth of rainfall, in inches
Ia = initial abstraction, in inches
S = maximum potential retention, in inches
3-4
Initial abstraction consists mainly of interception, infiltration during early parts of the
storm, and surface depression storage. Its determination is not easy due to the
variability of infiltration during the early parts of the storm since it depends on
conditions of the watershed at the start of a storm such as the land cover, surface
conditions and rainfall intensity; thus it is assumed to be a function of the maximum
potential retention as related in equation 3-5 (USDA, 2004)
3-5
Causes of variability of the CN are collectively called the Antecedent Runoff Condition
(ARC) and are divided into three classes: II for average conditions, I for dry conditions,
and III for wetter conditions. These are mainly due to rainfall intensity and duration,
total rainfall, soil moisture conditions, cover density, stage of growth, and temperature.
Attempts to explain the variability have been focused on antecedent soil moisture,
usually as indicated by 5-day antecedent precipitation (USDA, 2004).
Various studies have shown that there exists no relationship between the antecedent
precipitation and the CN hence it should be treated as random variable (Cronshey,
1983; Hjelmfelt, 1987, 1991; Van Mullem, 1992).
32
3.4.2. Evaluating Curve Number for the study area

CN is evaluated for the study area on pixel basis using the landcover/land use and soil
map that are reclassified to hydrologic conditions and hydrologic soil group.
3.4.2.1. Reclassification of landcover and soli map to hydrologic conditions

Land cover or land use represents the surface conditions in a watershed and plays key
role in determination on the amount of initial abstraction. The landcover/land use map
was reclassified to hydrologic conditions based on the USGS land use and land cover
classification system. The table of runoff curve number (SCS, 1986) following Chow et
al.,(1988) appendix 4 is used to assign codes to the various land cover/land use
classes. Table 3.4-1 gives CN for hydrological soil cover complexes for ARC II and
Ia=0.2s for the study area.
Table 3.4-1: Curve Number for hydrological soil cover complexes
Hydrological soil group
A B C D
Agricultural 72 81 88 91
Coastal Sand 50 50 50 0
Forest 36 60 73 79
Mangroves 0 0 0 0
Mixed Shrubs 30 48 65 73
Mixed Vegetation 39 61 74 80
Paved Roads 98 98 98 98
Plantations 36 60 73 79
Quarry 77 86 91 94
Recreation Parks 49 69 79 84
Settlements 61 75 83 87
Streams/Watercourse 0 0 0 0
Unpaved Roads 76 85 89 91
Wetlands/Water bodies 0 0 0 0
33
3.4.2.2. Reclassification of Soil map to Soil Group

Application of the CN method requires that the soils for the study area are reclassified
to fit in one of four categories (A, B, C, and D). The condition to fit the soils classes to
certain categories is subjective but depends highly on the infiltration rates and the
textural soil composition. These factors for the different soil classes were approximated
during the field work and are used to classify the soils of the study area following the
generic conditions for soil classification table 3.4-2.
Table 3.4-2: Generic conditions for soil classification (according to the CN

method).
A Low overland-flow potential. Minimum infiltration capacity when

wetted > 0.76 cm/hour. Deep well to excessively drained sands
and gravels
B Moderate minimum infiltration capacity when wetted 0.38 to 0.76
cm/hour. Moderately deep to deep, moderately to well drained,
moderately fine to moderately coarse grained (e.g. sandy loam)
C Low minimum infiltration capacity when thoroughly wetted 0.13
to 0.38 cm/hour. Or soils with impeding layer fragipan.
D High-overland flow potential. Very low minimum infiltration when
wetted < 0.13 cm/hour. Clay soils with swelling potential, soils
with permanent high water table, soils with clay near the surface,
or shallow soils over impervious bedrock.
Based on the conditions set above the soils for the study area are assigned groups has
shown in table 3.4-3. The final soil groups map figure 3.4-1 is based on table 3.4-3.
34
Table 3.4-3: Reclassification of soils to soil groups
Local Lithology FAO Texture Depth Infiltration Runoff Soil

class Classification Group
Name
Shallow Crystalline, reef Haplic and Loams, fine Shallow to Medium to Low B
Kinongo and detrital Humic clay loam thick High
Limestone Ferralsols to clay
Deep Crystalline, reef Rhodic Heavy Shallow to Medium to Medium C

Kinongo and detrital Ferralsol loams to Medium High
Limestone clay, sandy
clay loams
Kinamo Marls, Sandy Eutric and Humic clay Thick Low High D
clays and clayey calcic vertisols, loams to
sands cambisol clays
Uwanda Coralline and reef Mollic and Sandy Shallow High Low A
limestone Rendzic Loam To
Leptosols loam
Reddish Marls, sandy clays Rhodic nitisols Sandy clay Thick Medium to Low A
Mchanga and clayey sands and Haplic loams to High
Acrisols clay loams
Sandy Marls, Sandy Dystric Loamy Medium to Medium to Medium B

Mchanga clays and clayey Cambisols and sands to thick Low
sands, sands and cambic sandy clay
sandstones Arenosols loams
Maweni Sands and Rendzic Sandy Shallow High Low A

sandstones, Leptosol Loam To
marls, sandy clays loam
and clayey sands
35
Figure 3.4-1: Reclassified soil map to soil groups
Evaluation of area covered by different soil groups is presented in Table 3.4-4. Areas
under hydrologic soil group A are expected to generate less runoff and only about 32%
of the study area is expected to produces considerable prior runoff before the
landcover type is taken into consideration.
Table 3.4-4: Percentage of areas under different soil groups
Soil Group Area (Km2) % of total area
A 1016 68.2
B 271 18.2
C 139 9.3
D 64 4.3
1490 100
3.4.2.3. Building up CN map
CN map was generated using the reclassified landcover to hydrologic conditions and
the soil groups obtained earlier. The Values assigned to the landcover as hydrologic
conditions and the soil groups are reclassified to generate CN map using all the
possible combinations of the input classes. This procedure is performed in ILWIS using
the 2-Dimensional table operation.
36
Figure 3.4-2- shows the generated CN map per pixels for the study area. The map
gives an impression of the area that can generate more runoff based on the landcover
and soils in the study area.
Figure 3.4-2: Curve Numbers per 30 meters pixel size
High CN values indicate areas that have the lowest infiltration and more runoff is
expected from this areas since the initial abstraction and storage area minimal.
3.4.3. Determination of Runoff using Curve Numbers (CN)
Rainfall runoff relationships in this study are determined using pixel based curve
numbers and following the SCS curve number method. The formulation in equation 3-3
requires the determination of the initial abstraction (Ia) and the maximum potential
storage (S).
These are derived as input maps using equation 3-4 and 3-5 before the runoff can be
calculated. The maximum potential storage map is converted to mm from inches by
replacing 1000 and 10 with 25400 and 254 in equation 3-4 since the rainfall depth is
expressed in mm. Figure 3.4-3 shows the derived initial maximum storage per pixel.
37
Figure 3.4-3: Initial maximum storage values per 30 meters pixel size
The Initial storage is low in areas expected to generate more runoff which mainly
depends on the CN values as derived from the landcover and soils map. The coral rag
area covering mainly the eastern part of the Island and dominated by the limestone and
mixed shrubs have the highest initial storage and initial abstraction hence the least
runoff.
The runoff coefficient can be derived as either an event runoff coefficient or annual
runoff coefficient. Event runoff coefficient is defined as the portion of rainfall that
becomes direct runoff during an event. In hydrological modelling it represents the
lumped effect of a number of processes in a catchment which may include;
interception, evaporation, rainfall intensity, initial abstraction and hence runoff (Viglione
et al., 2009).
Annual runoff coefficient per pixel is derived for this study as opposed to event runoff
coefficient for individual storms since to establish the runoff amount that is available for
agricultural production this method takes into account rainfall events that do not
significantly contribute to any runoff (Zhu et al.).
Annual runoff depth is derived using the MPE data following equation 3-3: Annual
runoff coefficient per pixel is then derived using the formulation equation 3-5 which
gives an indication of the percentage rainfall that is transformed to runoff.
3-6
The annual runoff coefficient is based on runoff calculated using antecedent runoff
conditions II (ARCII); which in the median value, motivated by the fact that the
38
probability of occurrence of higher and lower values of the runoff coefficient would be
equal (Pilgrim et al., 1975, 1993).
3.5. Decision making and RWH site selection
GIS-based site suitability analysis has been applied in a wide variety of situations
including ecological approaches for defining land suitability/habitant for animal and
plant species (Pereira et al., 1993), land suitability for agricultural use (Cambell et al.,
1992); environmental impact assessment (Moreno et al., 1988), site selection for public
and private sector facilities (Church, 2002.; Eastman et al., 1993).
Site suitability analysis makes a distinction between the site selection problem and the
site search problem. The aim of site selection analysis is to identify the best site for an
activity from a set of potential (feasible) sites. In this type of analysis all the
characteristics (such as location, size, relevant attributes, etc.) of the candidate sites
are know. The problem is to rank or rate the alternative sites based on their
characteristics so that the best site can be identified. If there is not a pre-determined
set of candidate sites, the problem is referred to as site search analysis. The
characteristics of the sites (their boundaries) have to be defined by solving the problem.
The aim of the site search analysis is to explicitly identify the areal extent of the best
site (Malczewski, 2004).
The focus of this research was on site search analysis using thematic layers generated
in form of spatial raster layers and applies the analytical hierarchy process which takes
into account the spatial variability of all the input layers.
Analytic Hierarchy Process (AHP) to arrive at a decision on best sites RWH is

implemented using the spatial multi-criteria evaluation (SMCE) module embedded in
ILWIS. SMCE has been shown to support planning and decision making due to its
capability to perform spatial data analysis and is demonstrated to work appropriately for
waste disposal and park sites selection by Sharifi et al.,(2004) and Zucca et al.,(2008).
Biophysical and socio-economic criteria are considered in form of raster thematic layers
and integrated as either constrains or factors in RWH potential site search process.
The process is performed in two phases as documented by Sharifi et al.,(2004); the

identification (design) and comparison/evaluation (choice of solution). To achieve this
goal four steps as documented by Garfi et al., (2009) are implemented; problem
definition; criteria identification and selection; calculation of the relative weights; and
evaluation of results.
3.5.1. Problem definition

Currently, RWH is not practiced in most parts of the study area. Though RWH is
recognised to hold the potential to increase water availability for both domestic and
agricultural production, it is also considered core to achieving full potential of Africa
water resources for sustainable social and economic growth (ICRAF, 2005).
The problem is thus defined as the evaluation of appropriate sites that Micro and Macro
RWH can applied in the island in order to improve water availability for enhanced
agricultural production while preserving environmental integrity.
39
3.5.2. Criteria identification and selection

Criteria selection for site search analysis is based on an elaborate literature search,
indigenous and expert knowledge. The site selection process also takes into account
specific guidelines from Food and Agricultural Organisation (FAO) on conditions that
must be fulfilled both bio-physical and social economic to sustainably implement RWH
projects.
The criteria list considered for Micro and Macro RWH are outlined in tables 3.5-1 and
3.5-2.
Table 3.5-1: General criteria and constrains for Micro RWH
Group of Factors/ Thematic layer for

Spatial constrains and factors
Constrains evaluation
Constrains Distance to roads not less than 30 Distance to roads layer

metres
Distance to buildings not less than Distance to buildings layer

15 metres
Environmental Not within natural forests and Landcover / land use

protected areas or areas of
ecological importance
Not within water bodies, swamps and Landcover/land use

streams
Geomorphologic Soils with high water holding capacity Soil texture
Slope not more than 30 percent Slope classes
Socio-economic Recreational and historical sites Landcover/ land use

unsuitable
Hydrological Runoff index not less than 0.5 Runoff index layer
40
Table 3.5-2: General criteria and constrains for Macro RWH
Group of Factors/ Thematic layer for

Spatial constrains and factors
Constrains evaluation
Constrains Distance to roads not less than 30 Distance to roads layer

metres
Distance to buildings not less than 50 Distance to buildings layer

metres
Environmental Not within natural forests and Landcover / land use

Geomorphologic Soils with high water holding capacity Soil texture
Slope not less than 5 percent Slope classes
Socio- economic Recreational and historical sites Landcover/ land use

unsuitable
Hydrological Runoff index not less than 0.5 Runoff index layer
Constraints are criterion that determine in arriving at the main goal areas that should be
considered as absolutely not suitable and as opposed to the factors, a poor
performance of a constraint cannot be compensated by good performance of another
factor or constraint.
Factors are criteria that contribute to a certain degree towards the output and can be in
form of a benefit or cost. A benefit contributes positively while a cost contributes
negatively to the overall goal. As opposed to constraints, poor performance of a factor
can be compensated by good performance of another factor. This can still lead to good
overall performance towards the final goal (Sharifi et al., 2004; Zucca et al., 2008).
3.5.3. Calculation of the relative weights
Weighting of the factors and groups of factors is an important step since this
determines the relative contribution that a factor or group of factors will have towards
attaining the sub-goals and the overall goal. Three options that are available in ILWIS
SMCE module for assigning weights are; direct method, pair-wise comparison and rank
ordering. Use is made of the pair-wise comparison and the rank order methods in this
study.
In the pairwise comparison method, also known as the Analytical Hierarchy Process
(AHP) (Saaty, 1990, 2008), for each pair of factors an indication is made to which factor
is the most important using the fundamental scale of absolute numbers; Table 3.5-3.
Computation of weights in AHP for each decision element based on the pairwise
comparisons makes use of the eigenvalue technique (Saaty, 1980).
41
Table 3.5-3: The fundamental scale of absolute numbers (Saaty, 2008)
Intensity of Explanation
Definition
Importance
1 Equal Importance Two activities contribute equally to the
objective
2 Weak or slight
3 Moderate importance Experience and judgement slightly favour

one activity over another
4 Moderate plus
5 Strong importance Experience and judgement strongly favour

one activity over another
6 Strong plus
7 Very strong or demonstrated An activity is favoured very strongly over

importance another; its dominance demonstrated in
practice
8 Very, very strong
9 Extreme importance The evidence favouring one activity over

another is of the highest possible order of
affirmation
Reciprocals If activity i has one of the A reasonable assumption

of above above non-zero numbers
assigned to it when
compared with activity j,
then j has the reciprocal
value when compared
with i
1.1–1.9 If the activities are very May be difficult to assign the best value but
close when compared with other contrasting
activities
the size of the small numbers would not be
too
noticeable, yet they can still indicate the
relative importance of the activities
The qualitative terms to what extent a factor is more important than another is
subsequently indicated which is then used to convert these comparisons of all pairs of
factors to quantitative weights for all factors (Saaty, 2008).
In the rank order method, all factors and optional sub goals are placed in a rank-order
(most important item at the top) and numerical weights are calculated using either the
expected value method or the rank sum method.
The expected value method assumes equal probability for each set of weights that fits
the rank order of criteria. The weight vector is calculated as the expected value of the
feasible set and the result is a unique weight vector. The expected value method
calculates the weight, wk, for criterion k according to equation 3-7 (Janssen et al., 1994)
42
wk 3-7
Where n = the number of criteria

k = criterion
The rank sum method calculates the weight, W k, for criterion k according to equation 3-
8. This method combined with a multi-criteria method, always leads to complete
ranking (Janssen et al., 1994).
Wk= 3-8
Where n = the number of criteria

k = criterion
The weights assigned to the criteria’s by applying the pairwise ranking and rank sum
methods are presented in tables 3.5-4 and 3.5-5 for micro and macro RWH
respectively.
Slope plays a key role in determination of technological choice for both micro and
macro catchment RWH. The limit of application in the site analysis is set through
standardization which is set as a cost or benefits depending of the type of RWH system
analysed. Figures 3.5-1 and 3.5-2 show the standardization applied for both micro and
macro catchment RWH.
Table 3.5-4: Weighting for Micro-Catchment RWH
Weight Group of factors Weight Spatial factors

0.17 Environmental 0.67 Not within natural forests and
0.33 Not within water bodies, swamps
and streams
0.37 Geomorphologic 0.67 Soils with high water holding
capacity
0.33 Slope not more than 30 percent
0.10 Socio- economic 1.00 Recreational and historical sites

unsuitable
0.37 Hydrological 1.00 Runoff Index greater than 0.5
43
Figure 3.5-1: Slope Standardization (Cost Function) for Micro-Catchment RWH
A cost function with goal of 5% is applied for micro-catchment RWH. The implication is
that as the slope increases beyond the set goal the suitability decreases.
A similar standardisation is used for the runoff coefficient for both micro and macro-
catchment RWH. A benefit function with a goal of 0.40 is used to ensure that only those
areas that 40% of the rainfall is transformed to runoff under natural conditions attain
maximum suitability values in the final site selection process. Figure 3.5-2 below show
the standardisation applied.
Figure 3.5-2: Runoff Coefficient Standardization (Benefit function)
44
Table 3.5-5: Weighting for Macro-Catchment RWH
Weight Group of factors Weight Spatial factors

0.17 Environmental 1.00 Not within natural forests and
0.37 Geomorphologic 0.67 Soils with high water holding
capacity
0.33 Slope not less than 5 percent
0.10 Socio-economic 1.00 Recreational and historical sites

unsuitable
0.37 Hydrological 1.00 Runoff index greater than 0.5
For macro-catchment RWH the slope is standardised using a benefit function whose
implication is that as the slope increases suitability increases and attains a maximum
value at 30% slope.
Figure 3.5-3: Slope Standardization (Benefit function) for Macro-RWH
3.5.4. Assessing Consistency of Pairwise comparison
The accuracy of pairwise comparison is assessed through the computation of the

consistency index (CI). This determines the inconsistent in the pairwise judgments
hence allows for re-evaluation of comparisons.
The consistency index which is a measure of departure from consistency based on the
comparison matrices is expressed as
CI= (-n) / (n-1) 3-9
Where  is the average value of consistency vector and n are the number of columns in
the matrix (Garfì et al., 2009; Saaty, 1990; Vahidnia et al., 2008)
45
The consistency ratio (CR) is the calculated as
CR=CI/RI 3-10
The random index (RI) is an index that depends on the number of elements that are
being compared (Garfì et al., 2009).The table of the random indexes of matrices of
order 1-15 as derived by Saaty,(1980) is presented in appendix 8.
Perfect consistency implies a value of zero for CR which may not be attainable due to
bias and inconsistencies in subjective judgments. Pairwise judgement is considered
acceptable if CR ≤ 0.1 otherwise the pairwise judgments may be revised before the
weights can be applied (Saaty, 1980).
3.5.5. Evaluation of Results

The areas identified for both Micro and Macro RWH must have the highest suitability
index that is derived based on the weight assigned to the factors and group of factors.
This study though, focuses mainly on site search analysis which is to explicitly identify
the boundary of the best sites for RWH.
The results of analysis are evaluated based on areas deemed suitable on the basis of
soil characteristics and key land use practices. This is depicted through areas that the
generated composite index maps show maximum suitability.
3.6. Sensitivity Analysis

Sensitivity analysis was performed to assess how a change in criteria weighting affects
the RWH potential site selection.
Parameter sensitivity evaluation was achieved by applying different weighting to the

main criteria in the SMCE decision making tree. The sensitivity of a factor or a group of
factors is shown by the change in the spatial extent of RWH suitability. The process
was used to determine:-
• The importance of a factor or group of factors is in the site selection process for
RWH.
• Establish the levels of uncertainties of the different thematic layers and identify
parameters that need to be more accurately determined to ensure more accuracy
of the RWH model.
46
3.7. Selection of technological choices

RWH technologies location and distribution can be identified using spatial mapping
based on biophysical characteristics (Ngigi et al., 2007). A number of factors may be
considered, however the key spatial factors considered for identification of the suitable
RWH systems are the percentage slope, the soil characteristics and to some extent the
annual runoff coefficient based on FAO recommendations.
The classification of RWH techniques as outlined in FAO (1994) is mainly adopted with
minor adjustments. Table 3.7.1 and appendix 7 outline the classification system used in
this research to derive areas that are suitable for application of different RWH
techniques.
Table 3.7-1: Micro-catchment RWH systems
RWH system Criteria Reference

Roof Catchment Presence of (Arnold et al., 1986)
settlements (Zhu et al.)
Runoff coefficient > 0.8
Ponds and pans Runoff Coefficient > 0.5 (FAO, 1994; Hatibu et al.,
Slope > 5% 1999; Hudson, 1987; Mbilinyi
et al., 2007)
Strip catchment tillage CBAR = 2:1 (FAO, 1994; Hatibu et al.,
Agricultural lands 1999)
Contour bunds CBAR of less than 3:1. (FAO, 1994; Hatibu et al.,
slope < 5% 1999; Hudson, 1987; Mbilinyi
et al., 2007)
Semi-circular bunds Slope < 3% (FAO, 1994; Hatibu et al.,
CBAR of at least 3:1 1999; Hudson, 1987; Mbilinyi
et al., 2007)
Water storage structure for Clay soils (FAO, 1994; Hatibu et al.,
crop production (ndiva) Sloping terrain > 8% 1999; Hudson, 1987; Mbilinyi
Near water sources e.g. et al., 2007)
stream
Conservation Bench terraces Slope < 2 % (FAO, 1994; Hatibu et al.,
Deep soils 1999; Hudson, 1987; Mbilinyi
CBAR 2:1 et al., 2007)
Borders Slope < 8% (FAO, 1994; Hatibu et al.,
Clays, silt clays and sandy 1999; Hudson, 1987; Mbilinyi
clays et al., 2007)
Stone terraces Slope > 30% (FAO, 1994; Hatibu et al.,
Unstable soils 1999; Hudson, 1987; Mbilinyi
et al., 2007)
The main micro catchment RWH currently practiced in the island is the level bunds
mainly for rice production (Plate 1). These have not been widely adapted but present a
window of expansion since the community has some experience on their
implementation and management.
47
Plate 1: Level Bunds for Rice Farming (Paddy Rice)
Assessment of the amount of runoff that can be harvested to meet the water
requirement for rice farming to enhance production and reduce risk of crop failure is
based on this system though other harvesting methods can still be adopted.
Consumptive water use for rice production is estimated to range between 1000 mm to
2000 mm depending on the efficiency of the systems applied by the farmers (Tuong et
al., 2003). Table 3.7-2 outlines the seasonal amount of water required for various
purposes during the entire growing period for rice.
Table 3.7-2: Typical daily rates of water outflows and seasonal water input in
lowland rice: Adopted from (Tuong et al., 2003)
Daily (mm day-1) Duration (days) Season (mm)

Land preparation
Land soaking 100 - 500
Evaporation 4-6 7 to 30 28 - 180
Seepage and percolation 5-30 7 to 30 35 - 900
Total land preparation 160 - 1580
Crop growth period
Evapotranspiration
Wet season 4-5 100 400-500
Dry season 6 -7 100 600-700
Seepage and percolation
Heavy clays 1- 5 100 100-500
Loamy/sandy soils 15-30 100 1500-300
Total crop growth 500-3700
Total seasonal water input 660-5280
Typical range of values for total seasonal water input 1000-2000
48
The extra water required for seasonal rice production during the long and short rains is
based on the potential area for rice production estimated at 8240 hectares by JICA
(2002). The current area utilised is 5400 Ha under rain-fed rice production and 400
hectares under irrigations which implies that the full potential is not exploited.
An average amount of 1500 mm is used to calculate the extra water required during for
the two rain seasons. Based on a design rainfall of 820 mm (long) and 340 mm (short)
rains for a normal year, the run-on and runoff areas ratios and the amount of water to
be harvested are determined based on equations 3-11 and 3-12 (Zhu et al.).
3-11
3-12
Where CWR – Crop water Requirement

RC – Runoff Coefficient
EF – Efficiency Factor
The adaption of runoff harvesting and storage in earth dams, ponds and small weirs is
slowly picking up in the study area due to the need for irrigation during the dry periods
when surface water resources are inadequate to meet irrigation water demands (Plate
2 and 3)
Plate 2: In-stream water harvesting using a storage weir
49
Plate 3: Earth pan to harvest runoff for dry season irrigation
Runoff harvesting and storage (Macro-Catchment RWH) viability is assessed through

selection of 15 possible impoundment sites aided by the use of aerial photos of the
study area. The potential runoff expected at impoundment site based on the suitability
criteria’s derived using AHP is applied. The water demand per hectare as determined in
Zanzibar irrigation master plan (JICA, 2002) of 11,000 m3/ha over the dry period is
used to determine the storage required. Evaporation and seepage loss are accounted
for in the above estimates since it is based on the analysis of dam operations which
also include inflow and withdraws.
50
4. RESULTS
4.1. Rainfall Analysis
The core purpose of this research is not to validate MPE product but to explore how
best it can be applied to derive the runoff index in data scarce areas where the spatial
distribution of rain gauge network is inadequate or lacking.
The comparison is based on total annual rainfall measured from two rainfall stations
that data was available for the years 2007 and 2008. The correlation coefficient (r) and
the coefficient of determination (r2) are determined for the two data sets. The
correlation coefficient is used to measure the strength of the relationship between the
two data sets while r2 is to test how well the MPE rainfall amounts can be used to
predict the actual measured rainfall. Table 4.1-1 gives the correlation values obtained
and figures 4.1-1 shows coefficient of determination obtained.
Table 4.1-1: Correlation of measured rainfall and MPE product
Airport Kizimbani Airport Kzimbani

2007 2007 2008 2008
MPE Airport 2007 0.87

MPE Kizimbani 2007 0.24
MPE Airport 2008 0.86
MPE Kizimbani 2008 0.86
The measured and MPE derived rainfall are highly correlated in both the year except
for Kizimbani in 2007. A similar result is obtained for the coefficient of determination.
Based on the result obtained, the MPE product for the year 2008 is used in all further
rainfall runoff modelling.
51
Figure 4.1-1: Plot of measured and MPE derived rainfall
MPE rainfall despite the high correlations obtained under or over estimates the rainfall
amounts (figures 3.3-5 and 4.1-2) in most cases. The retrieval trends and the error in
estimation of the rainfall amounts is as shown in figure 4.1-2. MPE though captures
appropriately all the daily rainfall events recorded over the period of analysis.
Figure 4.1-2: Comparison of MPE rainfall retrieval and gauge data
Underestimations are more during the wet season with an average underestimation of
18 % over these periods (March to June and October to December) and a 43 %
average overestimation over the dry months. The total measured rainfall over the
period of analysis is 1440 mm with the MPE retrieval being 1399 mm showing a total
average underestimation of 3 %.
Based on these results, and in order to exploit the MPE strength of its ability to derive
rainfall with a high temporal and spatial resolution, the MPE derived rainfall amounts
52
were adjusted based on the obtained linear regression relationships (figure 4.1-1)
before it is further utilised for rainfall runoff analysis.
Figure 4.1-3: Adjusted MPE derived rainfall
The adjusted rainfall amounts depict the spatial-temporal rainfall distribution depth
more appropriately and thus can be adequately applied to derive the runoff depth.
53
4.2. Rainfall Runoff modelling

4.2.1. Annual Runoff and Runoff Coefficient
Results of spatial distributions of modelled annual runoff depth in mm are shown in
figure 4.2-1. Daily runoff maps were aggregated to prepare expected annual runoff
map. A variation from as low as 97 mm in the coral rag region with an increase
westwards to a maximum of 542 mm was observed.
The pixel based runoff index which is the ratio of modelled runoff depth to the annual
rainfall, shows a wide variation over the Island. It ranges from 0 - 0.99 (99%) figure 4.2-
2.
Figure 4.2-1: Annual runoff Depth Figure 4.2-2: Annual Runoff coefficient
(mm/year)
The main agricultural areas in the Island fall within the range of 360 mm and 450 mm of
annual runoff. These are areas that require supplementary irrigation to mitigate inter-
seasonal crop failures and increase the crop yields.
Monthly runoff volume generated compared to the rainfall volume is presented in figure
4.2-3. The main rainfall season (March to May) contributes 64 % of the total annual
runoff while the short rain season (October to December) account for 21% of the
annual runoff volume. The dry months produce only 15% of the total runoff even though
about 20% of the total rainfall is received within this period.
54
Figure 4.2-3: Comparison of monthly rainfall and runoff volumes (MCM)
4.3. Rainwater harvesting potential

4.3.1. Roof catchment
Roof catchment (RC) rainwater harvesting can be implemented in all regions of Island
since the annual rainfall amount is above 200 mm. The roof catchment RWH potential
is determined based on the presence of suitable roofing system. Figure 4.3-1 shows
the areal extents of roof catchment RWH suitability for the Island.
55
Figure 4.3-1: Roof Catchments suitability Map
Currently only 10.18 km2 is built up with suitable roofing systems that can be used for
RWH and are depicted as the suitable areas in figure 4.3-1.This area can generate a
total annual runoff volume of 4.6 million cubic meters. Areas shown as unsuitable are
an indication of the absence of suitable roofing systems associated mainly with lack of
settlements. Despite the existing roof catchment RWH potential has identified, RC
rainwater harvesting has not been adopted and exploited to its full potential.
The runoff available for storage was further analysed to determine the proportion of
domestic water demand that it can meet. Analysis assumes the basic water
requirements for domestic use as recommended by Gleick (1996). Table 4.3-1 outlines
the different water uses that account for 50 litres/capita/day required to meet the basic
needs.
Table 4.3-1: Recommended basic water requirements for human needs (Gleick, 1996)
Range (litres
Recommended minimum
Purpose per person per
(litres per person per day)
day)
Drinking Water 5 2 to 5
Sanitation Services 35 5 to 145
Cooking and Kitchen 10 10 t0 50
Total 50
Daily water requirements of 0.05 m3/day translates to annual domestic water demand
of 18.25 m3 per capita. The runoff generated can meet an annual domestic water
56
demand of 252,000 persons which represents about 33% of the total population of the
island estimated at 0.76 million persons currently.
4.3.2. Micro Catchment Rainwater Harvesting

Based on AHP analysis that took in to account various physical layers, the spatial
extents of micro-catchment RWH suitability areas are identified. All the factors and
group of factors are integrated to produces five suitability classes figures 4.3-2 and 4.3-
3. The potential sites for Micro-catchment RWH as identified reflect specific suitability
levels of parameters and weight of factors applied in the analysis.
Figure 4.3-2: Micro-Catchment RWH Suitability
57
Figure 4.3-3: Percentage area under different suitability classes
Area considered suitable (suitable and highly suitable classes) cover a total of about
44,000 hectares’ that represent 30 % of the Island. Suitable areas are mainly located in
agricultural areas and in area with soils having high water holding capacities. Areas
dominated by limestone’s in the coral rag region; eastern part of the Island are least
suitable.
The most suitable RWH methods for different area are evaluated based on table 3.7-1
and appendix 7. Accordingly areas suitable for different harvesting techniques are
shown in table 4.3-1 below.
Table 4.3-2: Areas suitable for different Micro-catchment RWH technologies
Technology Percentage Slope Area (Ha)

Conservation Bench terraces,
Stone terraces, Semi-circular 0-2 24179
bunds
Contour bunds, Borders 2-8 11902
Strip catchment tillage 8-16 2048
Stone terraces 16-30 367
Water storage structure for crop

> 30 3260
production (ndiva1)
Most of the technologies options for micro-catchment RWH can be practised since the
landscape is mainly within flat to undulating slope classes in areas identified as
1
Small water reservoirs
58
suitable. The results indicate that integration of multiple RWH systems is possible for
the study area.
Analysis of rainfall trends indicate that in a normal year 815 mm of rainfall is received
during the main rain season (Masika2) and 340 mm in the short rain (Vuli3) period
representing 61% and 24% of the total rainfall. Over the same period a runoff volume of
40 MCM (69%) and 11.7 MCM (20%) respectively is produced by the suitable areas for
micro catchment RWH.
The additional water depth (mm) required seasonally based on design rainfall depth of
815 mm and 340 mm in a normal year is 685 mm and 1160 mm for the main and short
rain season respectively. Based on equations 3-11 and 3-12; an average runoff
coefficient of 0.2 as derived from figures 4.2-2 and 4.3-2, efficiency factor of 0.5 and the
current area under rain-fed rice production of 5400 hectares, the extra water required;
run-on and runoff ratios for level bunds was determined.
The extra water required in main season is 36. 6 MCM while in the short rains is 62.6
MCM. Catchment area and cultivated area ratios for the seasons are 1:2 and 3:1 which
represents a minimum catchment area of 4,500 ha and 18,200 ha for the both seasons
respectively.
2
Long rains
3
Short rains
59
4.3.3. Macro Catchment Rainwater Harvesting

The macro-catchment suitability map indicated that 23% of the study area is suitable
(with suitable and high suitable classes) for RWH Figures 4.3-4 and 4.3-5. Suitability
areas cover a total area of about 35,000 hectares.
Determination of suitability is based on natural catchments with no treatment hence

areas with soils with high water holding capacity present the more suitable area.
Unsuitable areas are dominated by limestone areas were the amount of runoff
generated is minimal and impoundment may not be feasible.
Figure 4.3-4: Macro Catchment RWH suitability
60
Figure 4.3-5: Percentage of area covered by different suitability
Possible impoundment sites were selected (figure 4.3-5) to evaluate the possible runoff
that can be captured and stored based on the catchment contributing area. The sites
are selected based on the derived drainage pattern figure 3.3.2 and are as presented in
table 4.3-1 showing the annual runoff volume that can be harvested at the point of
impoundment.
Table 4.3-3: Possible Impoundment site and contributing catchment areas
Proposed Catchment Drainage Longest Flow Annual

Point of Area (Ha) Density Path Length runoff
Impoundment (m/km2) (m) volume
(MCM)
1 161 1612 4469 0.3

2 148 398 2521 0.2
3 176 284 2132 0.1
4 169 747 2928 0.5
5 1314 690 9182 1.6
6 597 717 4044 2.2
7 1528 818 10417 5.0
8 1220 738 6667 3.0
9 411 840 3870 0.9
10 988 742 6236 1.6
11 1892 732 8170 3.7
12 1125 827 5091 2.5
13 2834 749 12368 5.1
14 3104 895 11122 6.4
15 5221 977 14639 11.2
These sites are selected for analysis purposes and do not represent all the possible
harvesting sites. The amount of water harvested and the size of the impounding
structure need to be considered based on maximum water demand and demand
fluctuation over the year, losses through evaporation and seepages must be taken into
consideration.
61
Downstream irrigable area based on land suitability for irrigation and location of the
impoundment site was determined to evaluate the storage required to meet dry season
irrigation water demand. The limiting factor to the possible irrigable area is the runoff
generated that can be harvested for storage at impoundment site. Table 4.3-4 show the
required storage volume to meet the dry period irrigation demand and the possible
maximum irrigable area.
Table 4.3-4: Storage required covering the dry season irrigation water demand
and maximum irrigable area (based of available runoff)
Storage Maximum
Annual
Target required % of water Irrigable
Proposed runoff
Catchment Irrigable to cover requirement area (Ha)
Point of volume
Area (Ha) area dry met by (based on
Impoundment generated
(Ha) Season runoff runoff
(MCM)
(MCM) generated)
1 161 0.3 12 0.1 227 12
2 148 0.2 15 0.2 121 15
3 176 0.1 10 0.1 91 9
4 169 0.5 800 8.8 6 45
5 1314 1.6 85 0.9 171 85
6 597 2.2 1400 15.4 14 200
7 1528 5 940 10.3 48 455
8 1220 3 560 6.2 49 273
9 411 0.9 200 2.2 41 82
10 988 1.6 125 1.4 116 125
11 1892 3.7 360 4.0 93 336
12 1125 2.5 540 5.9 42 227
13 2834 5.1 360 4.0 129 360
14 3104 6.4 115 1.3 506 115
15 5221 11.2 1360 15.0 75 1018
Total 44.3 6882 75.7 59 3357
The selected impoundment site can only be able to meet 48% of the dry season
irrigation water demand based on the target irrigable area. This clearly indicates the
need to develop other water sources to meet the supplementary irrigation water
demand in the area.
4.4. Validation of results
In order to validate the results of analysis 21 and 24 locations of existing and areas
considered suitable for both Micro and Macro catchment respectively are used. The
points were selected based on indigenous knowledge and expert assessment during
the field work survey assisted by personnel from the department of irrigation-Zanzibar.
The selected point’s appendix 5 and 6 are used to test and check the quality of
performance and reliability of the developed RWH assessment model.
Testing for Micro-catchment rainwater harvesting show that 10 % of the sites identified
suitable are unsuitable, 10 % marginally suitable and 80 % within suitable and highly
suitable areas. Validation for Macro catchment RWH indicates that 12 % of the points
are in unsuitable areas, 20 % in marginally suitable and 68 % within suitable and highly
suitable areas.
62
The results give an indication of the reliability of the developed rainwater harvesting
assessment (RWHA) models and owing to the fact that most of the sites are
appropriately located the accuracy of the model is found satisfactory.
4.5. Sensitivity analysis

Sensitivity analysis was performed to help identify the spatial layers that are critical in
accurately determining the spatial extents of RWH suitability. This was achieved by
assessing the effects of the spatial extent variability by changing the weights assigned
to the group of factors in the criteria tree.
The degree of suitability and areal extent was examined through variation of the
weights assigned to the group of factors starting with an equal weight assignment. A
variation of weights using the pairwise comparison method, (Saaty, 1990, 2008), was
then performed for each pair of factors to examine which factor was the most
important.
The results of this analysis revealed that the geo-morphological factors are more
important followed by the hydrological factors. The soils, landcover and slope layers
were the most sensitive layers and inaccuracies in this layer can lead to errors in RWH
site suitability assessment. This assessment also played a major role in deriving the
appropriate weights for the group of factors and individual factors (tables 3.5-4 and 3-5-
5) that are used in suitability assessment.
63
5. Discussions
5.1. Rainfall
There exists strong rainfall dependence, in Sub-Saharan Africa mainly for agricultural
production but the region ranks among the lowest in the world in the density of rainfall
monitoring stations. The reliability of the data collected is a major issue since even
where rainfall data are available, weeks can elapse between collection and accessibility
to users is poor (Bowden et al., 2007). Arising from the above, the need to explore the
possibility of using satellite derive rainfall is explored. The MPE product is preferred
due to its high temporal and spatial reliability within the tropics.
Rainfall is a major deriving force in the runoff generation in any watershed. An accurate
rainfall data input is therefore required in order to derive accurately the amount of
runoff. Owing to this fact, a comparison is made between the measured rainfall and the
MPE derived rainfall amounts through determination of the correlation coefficients (r)
and the coefficient of determination (r2) table 4.1-1 and figure 4.1-1. To ensure that the
MPE derived rainfall accurately fits to the measured rainfall, adjustments are made to
correct for time lag in between the measured and simulated rainfall. Correction for
underestimation and overestimation is based on the empirical relationships derived in
determination of r2. This reduced the overall underestimation of the annual rainfall
depth to 3 %.
A low correlation coefficient of r=0.24 is obtained at Kizimbani station for the year 2007
which maybe an indication of an error in the measured rainfall since all the other period
have a high agreement.
Results obtained give an indication of how the MPE derived rainfall amounts fit to the
gauge measured rainfall. An indication that in absence of reliable ground
measurements the MPE rainfall product can satisfactorily be applied to estimate the
spatial rainfall distribution based on values of r and r2 (0.721) obtained. Despite the high
correlations obtained, one of the key problems may be the accuracy of the measured
rainfall and its reliability.
With adjustments to correct for time lag between the measured and the derived rainfall
amounts the MPE product is well suited for the tropical and convection rainfall
simulation. Correction for underestimation and overestimation of the total rainfall
amount is necessary before the derived rainfall amounts can be applied to model
rainfall runoff relations or design of water infrastructure.
5.2. Rainfall Runoff Modelling

Analysis of rainfall–runoff relationships determines how much of the net precipitation is
partitioned into runoff after all the initial losses. The SCS curve number method as
applied to model the runoff has been show to adequately give accurate results by
different authors.
The main cause of uncertainties in runoff modelling using satellite derived rainfall as
identified by Senay et al.,(2004) may arise from inaccuracies of the derived satellite
rainfall which require to be quantified. Annual runoff depth as derived from the MPE
64
product is lower than the amount derived using the actual measured rainfall due to the
overall underestimation of the rainfall depth by 3 % (section 4.1). It is therefore
necessary to make adjustments to the MPE derived rainfall before it is applied to model
runoff.
Rainfall–runoff relationships results obtained in a spatial scale give a clear indication

the amount of runoff generated by various land uses in the study area. An indication of
the variation of runoff coefficient for different soil types and land slopes is demonstrated
by Ngigi et al.,(2007). This correctly agrees with the results obtained in this study as
depicted by figure 4.2-2.
Application of the spatially distributed MPE derived rainfall and the pixel based CN for
runoff modelling represents one of the key strengths of this study. In application of SCS
curve number method more reliable results are expected if rainfall runoff relationships
are determined for a small area instead of averaging over the entire watershed (USDA,
2004).
Monthly runoff generation from the study area ranged from 2% to 16% of the total
rainfall and is highly reliant on the ARC. This relationship is shown in figure 5.2-1. This
is greatly dependent of the soils and landcover conditions of the study area.
Figure 5.2-1: Monthly Rainfall runoff relationships
In the study area 84% of the total runoff is generated within flat and undulating slope
classes. Deep kinango, shallow kinango and Kinamo soil type areas produce 30.3 % of
the total runoff though they only cover 18.2 % of the total area of the island. Mawani
soils with the highest coverage of 46.2% produce only 11.3 % of the total runoff. The
built up areas generates 50.2% of the total runoff with the agricultural, mixed vegetation
areas contributing 38.6 %. This is an indication that only a small area of the island is
viable for rainwater harvesting due to the minimal extent of the runoff generating areas.
The total annual runoff estimated for the island of 531.5 million cubic metres JICA
(2002) agrees reasonably with the modelled runoff of 415.7 million cubic metres.
65
Runoff volumes generated in this study may also be underestimated mainly due to the
coarse nature of the soil map used to derive the CN values which plays a role in
determining the amount of rainfall that will translated to runoff and the MPE
underestimation of the rainfall amounts.
5.3. Rainwater Harvesting

Rainwater harvesting has a potential of addressing spatial and temporal water scarcity
for domestic, crop production, livestock development, environmental management and
overall water resources management (Ngigi et al., 2007). Despite this realisation the
existing potential for RWH in the study area has not been exploited.
Rainwater harvesting potential assessment requires accurate information on the

spatial-temporal information on run-off potential area. This study provides an integrated
approach to model the spatial-temporal pattern of run-off potential areas using the SCS
CN model with remote sensing-derived inputs and ancillary data in GIS.
To asses RWH potential sites for the study area an annual runoff coefficient (figure 4.2-
2) is derived and thresholds (figure 3.5-2) set for use in spatial modelling of suitable
RWH. A threshold value of 0.8 is used to assess suitable area for roof catchment RWH
while a value of 0.2 is used for both micro and macro RWH.
The cost of implementing RWH can be evaluated using general basic cost per cubic
metre of water harvested that are based on published sources, expert consultations
and experiences by SEARNET in East Africa (ICRAF, 2005). This can enable
assessment of the best and most economically viable option since the areas identified
for different harvesting methods overlap.
5.3.1. Roof Catchment

Roof catchment (RC) harvesting is not currently practised in the island mainly based on
the perceived low water quality of rainwater and cultural background of the community.
The low adoption rates may also be attributed to the long standing policy that assumed
water has a social good and hence provided free for all the citizens. With a policy swift
towards attaching an economic value to water supplied the rate of adoption is expected
to increase.
Roof catchment (RC) is considered in this study based on a runoff coefficient of 0.8
mainly for corrugated roof systems. Rural household though may have lower runoff
coefficients (0.5) since they may have grass thatched roofs hence not considered.
This study shows that rainwater harvesting can supplement other water sources by
supplying about 49 million cubic metres of water annually if full potential (all roofs) are
used. This water can satisfy an annual domestic demand of 33% of the current
population.
RC rainwater harvesting can provide adequate water supply for households to cover for
times of water shortage and also reduce expenditure on water. The main required
intervention and challenge is to produce a system within the means of every household
that can meet their demands.
During the field work survey it was established that most of the residents have interest
in harvesting rainwater but were concerned with its taste and perceived low quality. It
66
was also established that there exists adequate capacity for construction of the RWH
system chosen using local skills, materials and equipment.
In order to achieve and exploit the full potential, the implementation and adoption of RC
rainwater harvesting requires extensive advocacy for the community by the
government, private sector and NGO to educate the community on its benefits.
5.3.2. Micro-Catchment RWH
The total annual rainfall received in region may be enough to sustain crop production,
but its distribution and occurrence of intra-season dry spells and off-season dry spells
may affect crop production. Mitigation of the effects of reduced crop production can be
achieved through implementation of RWH systems (Ngigi et al., 2005)
Poor agricultural production can be associated with poor rainfall partitioning which
implies that only a small fraction of rainfall reaches the root zone and mid season dry
spells that result into to poor soil water availability during the growing season
(Rockstro¨m, 2000).
Micro-catchment rainwater harvesting is currently under implementation in the study

area following the development of the Zanzibar Irrigation Master plan (JICA, 2002). An
indication of increased cropping intensity with adoption of RWH harvesting from the
current 14% to 64% during the dry season is implied
The identified suitable sites for macro-catchment RWH covering an area of 44,000 Ha
can greatly enhance agricultural production. To mitigate water shortages during the
cropping period and especially during the short rain period when a shortfall of about
62.6 MCM is evident for the main crop grown (rice), level bunds with a CBAR of 3:1 are
considered appropriate (section 4.3.2 Para: 6). This technology will be possible to
implement since most of the runoff generating area fall within flat and undulating slope
classes (section 5.2, Para: 6) which are most suited for in-situ RWH. The soil
associated with this slope classes are also well suited for agricultural production.
Most of the harvesting techniques identified (table 4.3-2) are relatively cheap and can
therefore be a viable alternative where irrigation water from other sources is not readily
available or too costly. RWH has been shown to be more viable than pumping water
since it saves energy and maintenance costs (Prinz et al., 2000).
Use of rainwater harvesting is envisaged to reduce over reliance on groundwater for

irrigation which according to JICA (2002) is not performing well due to the high
operation and maintenance costs involved.
5.3.3. Macro-Catchment RWH

Macro-Catchment RWH is considered based on the available runoff that can be stored
for use during the dry season. Most of the rivers in the study area have high peak
discharges during the rainy season with no flows or very low flows during the dry
season.
The ability to successfully manage this runoff is an important aspect towards

sustainable agricultural production for the Island. RWH systems that can be
implemented in the area considered suitable are mainly in-stream weirs, ponds and
water pans. These systems have an advantage in that no major loss of agricultural land
will occur and can be implemented by individual farmers or by the community use
(Plates 2 and 3).
67
Studies carried out in the island in the past (Halcrow, 1994; ICRAF, 2007; JICA, 2002)
show a considerable amount of surface runoff about 24% of the total rainfall is lost
each year. The potential for runoff harvesting exists as identified in figure 4.3-4 that can
be utilised through construction of runoff harvesting structures.
In-Stream weirs or check dams could be constructed across the small streams that
cover the landscape of Unguja and used to increase the retention time of runoff flows
during flash floods. The stored water could then be harnessed by gravity through buried
pipe collectors laid at the bottom or adjacent to the streambed or drawn through canals
to feed agricultural field crops or domestic and livestock water supply systems.
Runoff harvested can be used for irrigation during the dry period covering the months
of July to September, and January to February that are generally drier than the rest of
the year. The main wet period March to April contributes about 69% of the total runoff
with the short rain period producing about 20%. This gives a clear indication of the intra
seasonal variation of water availability in the island.
In total, 35,000 hectares of land in the study area could benefit from increased
agricultural production through increasing the management of surface runoff generated
through rainwater harvesting and storage.
This is demonstrated through selection of 15 possible impoundment sites within the

macro-catchment RWH suitable areas and analysis made on the required storage
volumes and possible irrigable area that can be adequately irrigated over the dry
period. Based on these sites, this study clearly demonstrates that a considerable
amount of dry season irrigation demand (table 4.3-4) i.e. up to 50% can be met through
RWH.
The storage volumes analysed are based on the assumption that the storage structures
will have one filling during the main rain season and the extracted amounts can be
replenished by the short rains.
68
6. Conclusions and Reccommendations

The objective of this study was to explore potential of data integration (use of historical
and near real time RS data, GIS and hydrological modelling) to assess the potential
rainwater harvesting sites in remote and data scarce areas in GIS environment. The
developed GIS-based rainwater harvesting models combines through AHP using
SMCE process physical, ecological, socio-economic and constrains layers as derived
from remote sensing to generate RWH suitability maps.
This study presents a contribution to site search analysis for RWH potential using
satellite products with minimal field data.
The MPE rainfall product can be used to determine the potential site for RWH due to its
highly reliable temporal and spatial variability with correction for time lag and
over/under-estimations. Despite its strength it is recommended that in designing of
water harvesting structure accurate rainfall measurement should be used since MPE
underestimates amount rainfall. Use of spatial rainfall data (satellite-derived) for 2008
as compared to the long term mean rainfall (derived from stations) showed that there
was a large difference in the amount of runoff amounts generated.
Results of sensitivity analysis revealed that the most sensitivity layers were the soils,
landcover and the runoff index. The soil layer used for this study was very coarse due
to lack of detailed work on soil mapping for the study area. A more detailed soil map for
this area would greatly improve the results obtained. A detailed soil mapping of the
island is recommended.
The CN is shown to vary with land cover changes and its application should be
considered alongside the changes that are taking place in Island due to changes in
land tenure and social economic development. The evaluation of the impacts of land
use change on the overall hydrology of the Island is therefore necessary.
Despite the fact that the potential use of the AHP as decision making tool in this study
is well demonstrated in coming up with site suitability for RWH the thematic layers need
to be more accurately determined. The methodology developed can be applied to
assess most of the parameters important for water harvesting systems in GIS
environment with limited ground data.
Currently utilisation of rainwater in Zanzibar is too low (1%). It will take great effort and
investment on the part of the government, private sector and general public to fully
utilize the existing potential. Adoption and implementation of RWH should be
considered with knowledge that the assessment of the associated impacts to the
overall water balance at the local and national scales is necessary. Linkages between
surface and groundwater need to be fully investigated before decisions are made on
key water resource development options.
The nature of the soil in the study area and mainly in the coral rag regions of the island
point to close relationship between sub-surface flow and surface runoff hence a more
elaborate water balance approach is required in order to understand this linkage which
cannot be explained by the SCS curve number method applied in this study.
69
The capabilities of using RS, GIS and field data for identifying potential sites for RWH
technologies for decision making on development and management of rainwater
harvesting programmes is will demonstrated in this research.
RWH suitability maps generated can be the first step in determination of the most
viable water resources management options that is feasible for different areas of the
island since the spatial perspective is well captured.
Arising from the results of validation, the application of the developed models shows
that it works effectively to identify potential sites for RWH technologies. Due to its
flexibility, its application can be adjusted based on changing scenarios in the study
area. This means that the subjective numbers in the suitability levels and weights of the
criteria can be changed according to characteristic changes of the study area.
The main constraint to the adoption of RWH could be associated with lack of
knowledge among the decision makers and the community on existing potential for
RWH for the island. RWH suitability maps developed in this study that give a clear
indication of the spatial extents and the existing potential can be a starting point for
creating awareness among stakeholder at the local and national scale
70
References
Allen R.G, Pereira, L., Raes D and Smith, M., 1998. Crop Evapotransipiration guidelines for
computing crop water requirements, Food and Agricultural Organisation of the United
Nations.
Arnold, P. and C, A., 1986. Rainwater Harvesting: The collection of rainfall and run-off in rural
areas. London: Intermediate Technology Publications.
Bancy, M. et al., 2007. Mapping the Potential of Rainwater HarvestingTechnologies in Africa.
Bowden, J.H. and F. H.M. Semazzi, 2007. Empirical analysis of intraseasonal climate
variability over the Greater Horn of Africa. Journal of Climate, 20: 5715-5731.
Calton W.E, 1955. A study of the more important soils of the Zanzibar protectorate. European
Journal of Science, 10(2): 169-176.
Cambell, J.C., Radke, J., Gless, J.T. and Whirtshafter, R.M., 1992. An application of linear
programming and geographic information systems: cropland allocation in antigue.
Environment and Planning, A(24): 535–549.
Canters, F., 1997. Evaluating the uncertainty of area estimates derived from fuzzy land-cover
classification. Photogrammetric Engineering and Remote Sensing, 63: 403-414.
Chow, V.T., Maidment D.R and Mays L. W, 1988. Applied Hydrology. McGraw Hill Higher
Education, 588 pp.
Church, R.L., 2002. Geographical information systems and location science. Computers and
Operations Research, 29(6): 541-562.
Critchley, W.R.S. and Reij, C., 1989. Water harvesting for plant production, Part 2. Case studies
and conclusions from Sub Sahara Africa.
Cronshey, R.G., 1983. Antecedent moisture condition probabilities. Irrigation and Drainage
Enginnering, Vol. 109(No. 2): 296–298.
de Winnaar, G., Jewitt, G.P.W. and Horan, M., 2007. A GIS-based approach for identifying
potential runoff harvesting sites in the Thukela River basin, South Africa. Physics and
Chemistry of the Earth, Parts A/B/C, 32(15-18): 1058-1067.
Eastman, J.R., Kyem, P.A.K., Toledano, J. and Jin, W., 1993. GIS and Decision Making.
UNITAR, Geneva.
Eischeid J, Pasteris P.A, Diaz H.F, Plantico M.S and Lott N.J, 2000. Creating a serially
complete, national daily time series of temperature and precipitation for the Western
United states. Journal of Meteorology, 39: 1580-1591.
Evenari, M., Shana, L. and Tadmor, N.H., 1971. The Neger, the challenges of a desert. Havard
University Press,Cambridge, Mass.
FAO, 1994. Water Harvesting For Improved Agricultural Production. In: FAO (Editor),
Proceedings of the FAO expert consultation, Cairo, Egypt, pp. 424.
FAO, 2003. Land and Water Digital Media Series, 26. Training Course on RWH
(CDROM).Planning of Water Harvesting Schemes, Unit 22. Food and Agriculture
Organization of the United Nations, Rome, FAO.
FINNIDA, 1991. Zanzibar Urban Water Supply Development Plan. ZB/REP/4.
Foody, G.M., 2002. Status of land cover classification accuracy assessment. Remote Sensing of
Environment, 80(1): 185-201.
Garbrecht J and Martz L.W, 1997. The assignment of drainage direction over flat surfaces in
raster digital elevation models. Journal of Hydrology 193: 204-213.
Garfì, M., Tondelli, S. and Bonoli, A., 2009. Multi-criteria decision analysis for waste
management in Saharawi refugee camps. Waste Management, In Press, Corrected
Proof.
Gleick, P.H., 1996. Basic water requirements for human activities: Meeting basic needs. Water
International, 21(2): 83-92.
71
Goovaerts, P., 2000. Geostatistical approachs for incorparating elevation into spatial
interpolation of rainfall. Journal of hydrology 228, 1: 113-129.
Gupta, K.K., Deelstra, J. and Sharma, K.D., 1997. Estimation of water harvesting potential for a
semiarid area using GIS and remote sensing. Remote sensing and geographic
information systems for design and operation of water resources systems. Proc.
international symposium, Rabat, Morocco, 1997(242): 53-62.
Halcrow, W., 1994. The development of Water Resources in Zanzibar, pp. 92.
Hatibu, N. and H. Mahoo, 1999. Rainwater harvesting technologies for agricultural production:
A case for Dodoma, Tanzania. Kaumbutho P G and Simalenga T E (eds)(Conservation
tillage with animal traction): 161-171.
Hettige, M.L., 1990. Land evaluation and land sustainability classification: Unguja and Pemba
Islands.
Hjelmfelt, A.T., 1987. Curve numbers in urban hydrology. Topics in Urban Hydraulics and
Hydrology. In: I.t.A.f.H. Research (Editor), pp. 73–78.
Hjelmfelt, A.T., 1991. Investigation of curve number procedure. Journal of Hydrological
Engineering, Vol. 117(No. 6): 725–737.
Hudson, N., 1987. Soil and water conservation in semi-arid areas. FAO, Bedford, United
Kingdom.
ICRAF, 2005. Potential for Rainwater Harvesting in Africa.A GIS overview, ICRAF and
UNEP.
ICRAF, 2007. An Assessment of Rainwater harvesting Potential in Zanzibar, pp. 86.
ITC-ILWIS, 2001. ILWIS 3.0 Academic: User's Guide, Volume 3, 530 pp.
James, R.A., Ernest E. H, John T. R and Richard E.W, 2001. A Land Use And Land Cover
Classification System For Use With Remote Sensor Data. In: G. Survey (Editor). A
revision of the land use classification system as presented in U.S. Geological Survey
Circular 671. United States Government Printing Office, Washington, pp. 41.
Janssen, R. and Van Herwijnen, M., 1994. Multiobjective decision support for environmental
management. Kluwer Academic Publishers, Dordrecht (Netherlands), 232 pp.
JICA, 2002. The Study on the Zanzibar Irrigation Master Plan, The United Republic Of
Tanzania.
Kahinda, J.M., Lillie, E.S.B., Taigbenu, A.E., Taute, M. and Boroto, R.J., 2008. Developing
suitability maps for rainwater harvesting in South Africa. Physics and Chemistry of the
Earth, 33(8-13): 788-799.
Kahinda, J.M., R.J. Boroto, R. Dube and Taigbenu, A.E., 2007. Preliminary Conceptual Model
of the GIS based Rainwater Harvesting Decision Support System.
Kidd, C., Heinemann T, Levizzani V and D.R., K., 2008. International Precipitation Working
Group (IPWG): Inter-comparison of regional precipitation products.
Kinoti Mutiga, J., Woldai, T. and Su, Z., 2006. Application of expert knowledge based multi -
criteria evaluation MCE process to select suitable areas for water harvesting systems in
Sub-Saharan Africa, a case study of Tanzania. In: AARSE 2006 : Proceeding of the 6th
AARSE international conference on earth observation and geoinformation sciences in
support of Africa's development, 30 October - 2 November 2006, Cairo, Egypt. Cairo :
The National Authority for Remote Sensing and Space Science (NARSS), 2006. ISBN
1-920-01710-0. 5 p.
Kumar, M., Agarwal, A. and Bali, R., 2008. Delineation of potential sites for water harvesting
structures using remote sensing and GIS. Journal of the Indian Society of Remote
Sensing, 36(4): 323-334.
Maathuis B.H.P and Wang L, 2006. Digital Elevation Model based Hydro-processing.
Geocarto-International, Vol 21(No. 1): 15.
Malczewski, J., 2004. GIS-based land-use suitability analysis: a critical overview. Progress in
Planning, 62(1): 3-65.
Mbilinyi, B.P., Tumbo, S.D., Mahoo, H.F. and Mkiramwinyi, F.O., 2007. GIS-based decision
support system for identifying potential sites for rainwater harvesting. Physics and
Chemistry of the Earth, Parts A/B/C, 32(15-18): 1074-1081.
72
Moreno, D. and Seigel, M., 1988. A GIS approach for corridor siting and environmental impact
analysis, Proceedings from the third annual international conference,, San Antonio,
Texas 2, pp. 507–514.
Ngigi, S.N., Savenije, H.H.G. and Gichuki, F.N., 2007. Land use changes and hydrological
impacts related to up-scaling of rainwater harvesting and management in upper Ewaso
Ng'iro river basin, Kenya, pp. 129-140.
Ngigi, S.N., Savenije, H.H.G., Thome, J.N., Rockström, J. and de Vries, F.W.T.P., 2005. Agro-
hydrological evaluation of on-farm rainwater storage systems for supplemental
irrigation in Laikipia district, Kenya. Agricultural Water Management, 73(1): 21-41.
Pereira, J.M.C. and Duckstein, L., 1993. A multiple criteria decision-making approach to GIS-
based land suitability evaluation. International Journal of Geographical Information
Systems, 7(5): 407–424.
Pilgrim, D.H. and Cordery, I., 1975. Rainfall temporal patterns for design floods. Journal of
Hydrology, 101: 81-91.
Pilgrim, D.H. and Cordery, I., 1993. Flood Runoff. In: D.R. Maidment (Editor), HandBook of
Hydrology. McGraw-Hill Companies,, pp. 42.
Prinz, D. and Anupam S, 2000. Technological Potential for Improvements of Water Harvesting.
Rockstro¨m, J., 2000. Water resources management in smallholder farms in Eastern and
Southern Africa: An overview. Physics and Chemistry of the Earth 25(3): 275-283.
Saaty, T.L., 1980. The Analytic Hierarchy Process. McGraw-Hill, New York, NY. USA.
Saaty, T.L., 1990. How to make a decision: The Analytic Hierarchy Process. European Journal
of Operational Research (48): 9-26.
Saaty, T.L., 2008. Decision making with analytic hierarchy process. Int. J. services Sciences,
Vol. 1, No. 1: 16.
Senay, G.B. and Verdin, J.P., 2004. Developing Index Maps of Water-Harvest Potential in
Africa. Vol. 20(6): 789-799: 12.
Sharifi, M.A. and Retsios, V., 2004. Site selection for waste disposal through spatial multiple
creteria decision analysis. Journal of Telecommunications and Information Technology:
11.
Tuong, T.P. and B.A.M. Bouman, 2003. Rice Production in Water-scarce Environments. In:
J.W. Kijne, R. Barker and D. Molden (Editors), Water Productivity in Agriculture:
Limits and Opportunities for Improvement pp. 53-64.
USDA, 2004. Estimation of Direct Runoff from Storm Rainfall, National Engineering
Handbook, pp. 79.
Vahidnia, M.H., A. Alesheikh, A. Alimohammadi and A. Bassiri, 2008. Fuzzy analytical
hierarchy process in GIS application. The International Archives of the
Photogrammetry, Remote Sensing and Spatial Information Sciences. , Vol. XXXVII.
Part B2.
Van Mullem, J., 1992. Soil moisture and runoff- another look. Irrigation and Drainage
Enginnering(Proceedings for Water Forum): 372–377.
Victor, M.P. and H.W, R., 1996. Runoff Curve Number: Has it reached Maturity. Journal of
Hydrologic Engineering, 1(1): 9.
Viglione, A., R. Merz and G. Bl¨oschl, 2009. On the role of the runoff coefficient in the
mapping of rainfall to flood return periods. Hydrology and Earth System Sciences, 13.
Vohland, K. and Barry, B., 2009. A review of in situ rainwater harvesting (RWH) practices
modifying landscape functions in African drylands. Agriculture, Ecosystems &
Environment, 131(3-4): 119-127.
Zaunderer, J. and Hutchinson, C.F., 1988. A review of water harvesting techniques of the Arid
Southwestern US and North Mexico ,Working paper for the World Bank’s Sub-Sahara
Water Harvesting Study.
Zhu, Q., Li Yuanhong and Ma Chengxiang, Rainwater Harvesting experience from international
community. In: T. Xiaojuan (Editor), Rainwater Harvesting, pp. 423.
73
Zucca, A., Sharifi, A.M. and Fabbri, A.G., 2008. Application of spatial multi-criteria analysis to
site selection for a local park: A case study in the Bergamo Province, Italy. Journal of
Environmental Management, 88(4): 752-769.
74
Appendencies
Appendix 1: Aerial photos of study area and points sampled during field work
75
Appendix 2: Accuracy assessments of land cover classification
AG MG MS MV PR Pl Sett ST/WC UR Totals UA %
AG 13 0 2 5 0 0 1 0 0 21 62
MG 0 2 0 0 0 0 0 0 0 2 100
MS 0 0 17 0 0 0 0 0 0 17 100
MV 0 0 2 12 0 0 0 1 1 16 75
PR 0 0 0 0 9 0 0 0 0 9 100
PL 0 0 2 0 0 6 0 0 0 8 75
Sett 0 0 0 1 0 0 8 0 0 9 89
ST/WC 1 0 0 2 0 0 0 2 0 5 40
UR 0 0 0 0 0 0 0 0 4 4 100
Totals 14 2 23 20 9 6 9 3 5 91 82
RA % 93 100 74 60 100 100 89 67 80 85
Abbreviations: AG-Agricultural, MG-mangroves, MS- Mixed shrubs, MV- Mixed vegetation,

PR- Paved roads, PL- Plantations, Sett- Settlements, ST/WC- Streams/Watercourses, RA%-
Reliability accuracy, UA%- User accuracy
Appendix 3: Long term mean monthly rainfall
Donge
Airport Kizimbani Kilombero Mahonda Selem Victoria Paje Makunduchi Kipange
JAN 72.0 70.6 42.3 63.6 47.4 31.0 74.2 63.2 25.3
FEB 37.6 35.1 10.2 12.7 27.3 20.0 22.8 29.6 79.1
MAR 204.5 186.3 144.1 136.3 138.6 91.4 65.4 112.5 189.3
APR 347.8 352.4 304.5 317.2 288.1 266.3 289.8 370.8 306.8
MAY 197.3 241.2 323.5 246.3 224.2 189.9 197.8 215.8 194.9
JUN 48.7 74.9 27.0 48.5 50.6 37.6 95.9 78.6 93.4
JUL 26.4 69.2 48.9 54.4 49.4 25.4 24.4 92.0 52.9
AUG 26.1 50.4 28.2 37.9 40.8 17.0 15.8 3.7 128.5
SEP 22.3 51.9 25.1 52.2 46.1 20.1 4.7 4.2 87.2
OCT 102.5 104.5 39.7 100.3 100.8 121.0 72.6 6.5 87.2
NOV 222.7 187.1 50.0 156.3 140.5 71.2 43.7 28.5 155.0
DEC 171.7 170.3 45.8 129.8 131.3 143.5 155.0 113.3 165.7
Total 1479.6 1593.9 1089.3 1355.5 1285.1 1034.4 1062.1 1118.7 1565.3
76
Appendix 4: Table of Runoff Curve Numbers (SCS, 1986)

Description of Land Use Hydrologic Soil Group
A B C D
Paved parking lots, roofs, driveways 98 98 98 98
Streets and Roads:
Paved with curbs and storm sewers 98 98 98 98
Gravel 76 85 89 91
Dirt 72 82 87 89
Cultivated (Agricultural Crop) Land*:
Without conservation treatment (no terraces) 72 81 88 91

With conservation treatment (terraces,
62 71 78 81
contours)
Pasture or Range Land:
Poor (<50% ground cover or heavily grazed) 68 79 86 89
Good (50-75% ground cover; not heavily

39 61 74 80
grazed)
Meadow (grass, no grazing, mowed for hay) 30 58 71 78
Brush (good, >75% ground cover) 30 48 65 73
Woods and Forests:
Poor (small trees/brush destroyed by over-

45 66 77 83
grazing or burning)
Fair (grazing but not burned; some brush) 36 60 73 79
Good (no grazing; brush covers ground) 30 55 70 77
Open Spaces (lawns, parks, golf courses, cemeteries, etc.):
Fair (grass covers 50-75% of area) 49 69 79 84
Good (grass covers >75% of area) 39 61 74 80
Commercial and Business Districts (85%

89 92 94 95
impervious)
Industrial Districts (72% impervious) 81 88 91 93
Residential Areas:
1/8 Acre lots, about 65% impervious 77 85 90 92
1 Acre lots, about 20% impervious 51 68 79 84
77
Appendix 5: Validation points Micro-Catchment RWH
Point Easting Northing Land use Landcover Soil type

1 558525 9287536 Agricultural Dry swamp area Clays
Seasonal Riverine vegetation
2 539665 9312805 Clays and sand
stream and Guavas
Sweet potatoes and
3 537372 9309859 Agricultural Clays
bananas
Reverine vegetation
4 534360 9341992 stream valley Clay loam soils
and water
Dam site Embankment of earth
5 536560 9345009 Sandy Clays
Embankment dam
Weeds and grass
6 533391 9345009 Agricultural Sandy soils
(uncultivated)
Weeds and grass
7 534824 9344541 Agricultural Clays
(uncultivated)
Agricultural Uncultivated (grass
8 531196 9317153 Sandy clays
(irrigated) and weeds)
Reverine vegetation
9 527515 9347213 Riparian area Sandy clays
and water
Sweet potatoes within
10 523300 9334086 Agricultural Sandy clays
the river valley
Rice on planted area
Irrigated
11 528776 9330284 and weed and grass Sandy clays
agriculture
on unplanted areas
12 541062 9315940 Agricultural Rice farming Clays
Abandoned
13 534968 9326626 Swampy area Limestone and silt
quarry
Rice farms with

Irrigated
14 528060 9321552 mangos and palms in Clays and loams
agriculture
the surrounding area
Rice farms with and

Irrigated
15 538064 9322338 palms in the Clay and loams
agriculture
surrounding area
Rain fed rice and
16 537338 9337468 Agricultural Sandy loams to clay
cassavas
17 533094 9344754 River clay
River not
18 529989 9350588 sandy clay
flowing
19 524568 9329083 mash, ditch sandy clay
20 531434 9349608 River clay
21 528187 9350393 River sandy clay
78
Appendix 6: Validation Points Macro-catchment RWH
Point Easting Northing Land use Landcover Soil type

1 529790 9320640 Agricultural Irrigated rice and Clays
bananas
2 558525 9287536 Agricultural Dry swamp area Clays
3 525128 9317295 Water pond Water Sandy clays
4 539665 9312805 Seasonal stream Riverine vegetation and Clays and sand
Guavas
5 535750 9312974 Road Paved tarmac Tar
6 537372 9309859 Agricultural Sweet potatoes and Clays
bananas
7 534360 9341992 stream valley Reverine vegetation and Clay loam soils
water
8 536560 9345009 Dam site Embankment of earth Sandy Clays
Embankment dam
9 534824 9344541 Agricultural Weeds and grass Clays
(uncultivated)
10 530410 9350371 River riparian Riverine vegetation and Sandy silt soils
area palms
11 527515 9347213 Riparian area Reverine vegetation and Sandy clays
water
water
water
14 525121 9337389 Riparian area Mixed reverine Sandy bottoms
vegetation and clays along
the banks
15 523300 9334086 Agricultural Sweet potatoes within Sandy clays
the river valley
16 528776 9330284 Irrigated Rice on planted area and Sandy clays
agriculture weed and grass on
unplanted areas
17 541062 9315940 Agricultural Rice farming Clays
18 533129 9364770 Bush land Mixed shrubs silty loams
19 536050 9321377 Agricultural Oranges, Palms and Loam soils
Mangos
20 527770 9347195 River riparian Reverine vegetation and
area water
21 528060 9321552 Irrigated Rice farms with mangos Clays and loams
agriculture and palms in the
surrounding area
22 538064 9322338 Irrigated Rice farms with and Clay and loams
agriculture palms in the surrounding
area
23 537338 9337468 Agricultural Rain fed rice and Sandy loams to
cassavas clay
24 530417 9350370 River sandy clay
79
Appendix 7: Macro-Catchment RWH System selection (FAO, 1994)
Appendix 8: Random Indices (RI) for n = 1, 2... 15 (Saaty, 1980)
n RI n RI n RI
1 0.00 6 1.24 11 1.51
2 0.00 7 1.32 12 1.48
3 0.58 8 1.41 13 1.56
4 0.90 9 1.45 14 1.57
5 1.01 10 1.49 15 1.59
80

WWW - Itc.nl Library Papers 2010 MSC Wrem Munyao

Uploaded by

Copyright:

Available Formats

WWW - Itc.nl Library Papers 2010 MSC Wrem Munyao

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

WWW - Itc.nl Library Papers 2010 MSC Wrem Munyao

Uploaded by

Copyright:

Available Formats

USE OF SATELLITE PRODUCTS TO ASSESS WATER

HARVESTING POTENTIAL IN REMOTE AREAS OF AFRICA

A CASE STUDY OF UNGUJA ISLAND

John Ngila Munyao

John Ngila Munyao

Thesis submitted to the International Institute of Geo-information Science and Earth

Thesis Assessment Board

Dr. C.M.M Mannaerts Chairman

INTERNATIONAL INSTITUTE FOR GEO-INFORMATION SCIENCE AND EARTH OBSERVATION

This document describes work undertaken as part of a programme of study at the

Table 3.2-1: Primary Data sets ..................................................................................................... 21

AHP Analytic Hierarchy Process

• Micro-catchment rainwater harvesting

According to a report by Millennium Development Goals - MDG Centre, Nairobi Kenya

• Poor access to and availability of water due to inadequate water harvesting

1.1. General objectives

1.1.1. Specific objectives

1.2. Research Questions

1.4. Thesis Outline

This thesis is presented in six chapters as outlined below:-

The importance of traditional, small scale systems of rainwater harvesting in sub-

2.2. Rainwater harvesting potential assessments

Identifications of potential sites for construction of rainwater harvesting structures for

Kahinda et al.,(2008) presented a methodology that enables water managers to assess

2.2.2. Analytic Hierarchy Process (AHP)

Key considerations that are of critical importance in decision making as outlined by

2.3. Rainfall-Runoff modelling

2.4. Study area

Figure 2.4-1: Map of Unguja Island

Annual average temperature of the Island is 26oC, with maximum temperatures of 27 oC

Geological Aerial GeonetCAST

DEM hydro- Extraction of MPE

3.2. Primary data sets and field work

Table 3.2-1: Primary Data sets

Data set Year Scale Spatial Extent Type Source

Layer digitized 2009 Varying Unguja Island Geo-referenced SMOLE

3.2.2. Field work

3.3. Application of RS and GIS

• Flow direction determination using the deterministic 8 model which determines

Figure 3.3-2: Drainage and catchment areas derived from DEM

3.3.2. Analysis of Rainfall Distribution from rain gauges network

Interpolation of point measurement is therefore necessary to estimate rainfall for areas

Figure 3.3-3: Spatial distribution of rainfall stations

Moving average performs a weighted averaging on point values based on a specified

Weight= ( )-1 3-1

Estimated value = 3-2

Figure 3.3-4: Spatial distribution long term mean annual rainfall

Figure 3.3-5: MPE Annual Rainfall Totals

3.3.4. Thematic maps

Accuracy assessment of classification was performed using a confusion matrix which

Figure 3.3-6: Classified Land cover/land use

3.3.4.2. Soil map

Figure 3.3-7: Soil texture map of the study area

Table 3.3-1: Areas covered by different soil classes

Local Class FAO classification Texture Area (Km2) % of total

Deep Kinongo Rhodic ferralsols Heavy loams to clay, 139 9

Maweni Rendzic Leptosol Sandy Loam To loam 681 44

Total 1561 100

3.3.4.3. Slope map