University of Southern California Dissertations and Theses

Developing and implementing a GIS-based framework to identify optimal locations for clean water wells in sub-Saharan Africa

(USC Thesis Other)

Developing and implementing a GIS-based framework to identify optimal locations for clean water wells in sub-Saharan Africa

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content DEVELOPING AND IMPLEMENTING A GIS-BASED FRAMEWORK TO IDENTIFY
OPTIMAL LOCATIONS FOR CLEAN WATER WELLS IN SUB-SAHARAN AFRICA

by

Jared Charles Hyneman

A Thesis Presented to the
FACULTY OF THE USC GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
MASTER OF SCIENCE
(GEOGRAPHIC INFORMATION SCIENCE AND TECHNOLOGY)

December 2014

Copyright 2014 Jared Charles Hyneman

ii

ACKNOWLEDGMENTS
I am indebted to a number of people whose insight and support have made this such a rewarding
experience. I want to thank my thesis committee at the University of Southern California, Dr.
Karen Kemp, for her continuous guidance, expert advice and patience through all stages of this
thesis and to Dr. Su Jin Lee and Dr. Daniel Warshawsky for sharing their knowledge and passion
for geography. I also want to thank the professors of the Spatial Sciences Institute: Dr. John
Wilson, Dr. Darren Ruddell, Dr. Yao-Yi Chiang and Dr. Jennifer Swift, for shaping my
understanding of geographic information science and technology.
And finally, I want to thank my parents and sisters for their love, support and encouragement for
all these years.

iii

For Sean

iv

TABLE OF CONTENTS
ACKNOWLEDGMENTS .............................................................................................................. ii
LIST OF FIGURES ...................................................................................................................... vii
LIST OF TABLES ......................................................................................................................... ix
LIST OF ABBREVIATIONS ......................................................................................................... x
ABSTRACT ................................................................................................................................... xi
CHAPTER 1: INTRODUCTION ................................................................................................... 1
1.1 The need for safe drinking water…………………………………………………………... 2
1.1.1 Helping Provide Clean Water……………………………………………………… 4
1.1.2 Social Context of Clean Water…………………………………………………….. 4
1.2 Goal of this Project………………………………………………………………………… 5
1.3 Scope of the Framework…………………………………………………………………… 6
1.4 Project Workflow………………………………………………………………………….. 8
1.5 Report Organization……………………………………………………………………….. 8
CHAPTER 2: RELEVANT BACKGROUND RESEARCH ....................................................... 10
2.1 Potable Water Situation in Rural Africa…………………………………………………. 10
2.2 Site Selection for Siting Water Wells……………………………………………………. 12
2.3 Previous Research on Risks to Drinking Water…………………………………………. 15
2.4 Methods for Site Suitability Analysis…………………………………………………… 16

v

CHAPTER 3: A GENERAL FRAMEWORK FOR SITING CLEAN WATER WELLS IN
SUB-SAHARAN AFRICA .......................................................................................................... 19
3.1 Fuzzy Overlay…………………………………………………………………………….. 19
3.2 Modeling Risks Where Groundwater Data is Limited…………………………………… 21
3.3 Suitability Criteria for the General Framework……………………………………..…… 23
3.3.1 Locations to Avoid and the Implied Constraints………………………………… 23
3.3.2 Locations to Favor……………………………………………………………….. 26
3.4 Appropriate Data for Analysis of Criteria………………………………………………. 34
CHAPTER 4: IMPLEMENTATION STUDY ............................................................................. 40
4.1 Study Area……………………………………………………………………………….. 40
4.2 Data……………………………………………………………………………………….. 44
4.3 Methodology……………………………………………………………………………… 51
4.3.1 Surface Flow of a Contaminant…………………………………………………… 53
4.3.2 Wetlands and Depressions………………………………………………………… 59
4.3.4 Slope………………………………………………………………………………. 67
4.3.5 Major Roads and Schools…………………………………………………………. 68
4.3.6 Populated Places…………………………………………………………………... 70
4.4 Fuzzy Suitability Overlay Results………………………………………………………... 74
4.5 Iterative Model Fitting…………………………………………………………………... 76
4.6 Evaluating Results of Refined Model…………………………………………………….. 80

vi

CHAPTER 5: SUMMARY AND CONCLUSION ...................................................................... 90
5.1 Assessment of model success……………………………………………………………. 88
5.2 Future Work………………………………………………………………………...……. 91
5.3 General applicability of research……………………………………………………….... 92
REFERENCES ............................................................................................................................. 97
APPENDIX 1: PRELIMINARY MODEL EXISTING WELL SUITABILITY VALUES….. 104
APPENDIX 2: REFINED MODEL EXISTING WELL SUITABILITY VALUES ................. 107
APPENDIX 3: REFINED MODEL OVERALL SUITABILITY. ............................................. 110

vii

LIST OF FIGURES
Figure 1 Girl collecting water from an unprotected natural spring ................................................ 3
Figure 2 Overview of project workflow ......................................................................................... 8
Figure 3 Components of current well site selection...................................................................... 12
Figure 4 Overview of Study Area ................................................................................................. 41
Figure 5 Overview of Study Area with Lake Kyogo .................................................................... 42
Figure 6 Weather Station in Soroti, Uganda ................................................................................. 43
Figure 7 Overview of Average Rainfall in Soroti ......................................................................... 44
Figure 8 Map of Community Infrastructure.................................................................................. 46
Figure 9 Map of Hygiene and Sanitation Infrastructure ............................................................... 46
Figure 10 Flow Path Example Map .............................................................................................. 55
Figure 11 Overview of Contaminant Flow Analysis .................................................................... 56
Figure 12 Fuzzy Linear Function .................................................................................................. 58
Figure 13 Criteria 1a Membership Layer Overlaid on Satellite Image ........................................ 59
Figure 14 ATP Class Names ......................................................................................................... 60
Figure 15 WRI Wetlands Polygon overlaid on ATP Layer .......................................................... 62
Figure 16 Overview of Process to Create Depressions Layer ...................................................... 64
Figure 17 Map of Membership for Criterion 2a ........................................................................... 65
Figure 18 Map of Membership for Criterion 3a ........................................................................... 66
Figure 19 Map of Membership of Criterion 4a ............................................................................. 67
Figure 20 Map of Membership of Criterion 4b ............................................................................ 69
Figure 21 Map of Membership for Criterion 6a ........................................................................... 70

viii

Figure 22 Map of Membership for Criterion 7a ........................................................................... 72
Figure 23 Map of Membership of Criterion 7b ............................................................................ 74
Figure 24 Overall Suitability Map based on Refined Model ........................................................ 75
Figure 25 Membership based on Refined Model and Infrastructure ............................................ 76
Figure 26 Preliminary Suitability Map ......................................................................................... 77
Figure 27 Refined Suitability Map with Existing Wells ............................................................... 78
Figure 28 Existing Wells with Contaminant Flow Membership Value ........................................ 82
Figure 29 Existing Wells with Wetlands Membership Value ...................................................... 83
Figure 30 Existing Wells with Stagnant Water Membership Value ............................................. 84
Figure 31 Existing Wells with Slope Membership Value ............................................................ 85
Figure 32 Existing Wells with Major Roads Membership Value ................................................. 86
Figure 33 Existing Wells with the Schools Membership Value ................................................... 87
Figure 34 Existing Wells with the Populated Places Membership Value .................................... 88
Figure 35 Suitability Map with Existing Wells Criteria ............................................................... 93

ix

LIST OF TABLES
Table 1 Minimum distance from sources in Illinois ..................................................................... 15
Table 2 General Criteria................................................................................................................ 29
Table 3 General Data Sources....................................................................................................... 31
Table 4 Recommendations for Data ............................................................................................. 38
Table 5 Suitability Criteria in Implementation Study ................................................................... 47
Table 6 Implementation Data Requirements ................................................................................ 48
Table 7 Data Specifics .................................................................................................................. 49
Table 8 Overview of Fuzzy Methodology .................................................................................... 52
Table 9 Contamination Source Criterion Fuzzy Assignment ....................................................... 57
Table 10 Wetland Criterion Analysis ........................................................................................... 65
Table 11 Depressions Criterion Analysis ..................................................................................... 66
Table 12 Slope Criterion Analysis ................................................................................................ 67
Table 13 Analysis of Major Road Criterion ................................................................................. 68
Table 14 Analysis of School Criterion.......................................................................................... 69
Table 15 Analysis of Populated Places Criterion ......................................................................... 72
Table 16 Analysis of Existing Wells Criterion ............................................................................. 73
Table 17 Model Refinements and Improvements ......................................................................... 79
Table 18 Suitability Classification ................................................................................................ 81
Table 19 Existing Well Distribution in Suitable Zone .................................................................. 81
Table 20 Distribution of Existing Wells in Membership Zones by Criteria ................................. 89
Table 21 Percent Distribution of Existing Wells in Membership Ranges by Criteria .................. 89

x

LIST OF ABBREVIATIONS
ATP Africa Topographic Position

DEM Digital Elevation Model

MDG Millennium Development Goal

NGO Non-Government Organization

OSM Open Street Map

RWSN Rural Water Supply Network
SRTM Shuttle Radar Topography Mission
SSI Spatial Sciences Institute
UN United Nations

UNDP United Nations Development Programme
UNEP United Nations Environment Programme
USGS United States Geological Survey
WHO World Health Organization

WMO World Meteorological Organization

xi

ABSTRACT
In 2000 the United Nations (UN) created the Millennium Development Goals (MDGs) to focus
on addressing major issues like poverty, education, children’s health, sustainable environment,
disease prevention, and economic development. One of the targets (7C) of the MDGs is to halve
the portion of the population that does not currently have sustainable access to safe drinking
water and basic sanitation by 2015. As a region, sub-Saharan Africa is not on track to meet the
goal. In fact, the region has the lowest clean drinking water coverage of any region in the world.
This project develops a general framework to improve water resource planning in sub-Saharan
Africa. The project defines criteria, data and methods to improve planning for clean drinking
water wells. The result is a general framework for 1) finding locations where contamination of
water wells is least likely to occur, and 2) ensuring the benefits of clean water support overall
community health and education. This is all with the aim to increase efficient water resource
planning to support the MDG to increase safe drinking water coverage. The general framework
is implemented as a model which is the functional component of the framework. The general
framework was refined through the implementation of the model in a model fitting study in rural
Uganda. The result of the implementation is a suitability map identifying locations where (1)
risks to drinking water are minimized and (2) benefits to people living in the study area are
maximized. The success of the model was evaluated by assessing the locations of existing wells
against what the model identified as suitable well locations. The framework and model fitting
process can be used as a tool by governments and non-government organizations (NGOs) to
improve current water site suitability workflows.

1

CHAPTER 1: INTRODUCTION
The pressing global issues of poverty, inequitable education, and children’s health have
long spanned the world, affecting countless populations. In 2000 the United Nations (UN)
created the Millennium Development Goals (MDGs) to focus on critical humanitarian issues
including poverty, education, children’s health, sustainable environment, disease prevention, and
economic development. The MDGs serve as a global framework for development and also serve
to guide the efforts of primarily developing countries to fight against obstacles that stand in the
way of realizing and harnessing the full human potential. Additionally, from 2000 through 2014
there has been progress towards many of the targets of the MDGs, but other areas still require
much attention.
One of the targets (7C) of the MDGs is to halve the proportion of the population that does
not currently have sustainable access to safe drinking water and basic sanitation by 2015 (UN
2000). This effort does appear to have been successful in helping provide access to improved
drinking water sources: the UN reports “more than 2.1 billion people have gained access to
improved drinking water sources since 1990, exceeding the MDG target” (UN MDG Report
2013, 1). Between 1990 and 2010, the proportion of the global population using improved
drinking water sources increased to 89 percent, up from 76 percent.
However, progress has not been universal, and problems remain. According to the UN’s
definition, a rural household is considered to have safe drinking water coverage if there is a safe
water source within 1.5 km from the household (Adam et al. 2009). By this measure, sub-
Saharan Africa has the lowest clean drinking water coverage of any region in the world
(UNICEF and WHO 2012). For example, compared to the global average of 89 percent, in Latin

2

America coverage is 90 percent, but in sub-Saharan Africa, coverage is 61 percent. This means
that 61 percent have safe water within 1.5km, but the remaining 39 percent have to go further to
find safe water. Furthermore, countries that still have less than 50 percent coverage in water
supply are almost all in sub-Saharan Africa (UNICEF and WHO 2012).
Additionally, the Democratic Republic of the Congo only has improved water sources for
16 percent of its population since 1995 (UNICEF and WHO 2012). This is not enough progress
to be on track to meet the MDG in the region. It is widely recognized that there is a strong need
for more effective planning and better decision-making if the MDGs of the UN are going to be
met. Therefore, this project focuses on rural sub-Saharan Africa, the region of the continent of
Africa to the south of the Sahara Desert, because it is the region with the least coverage of clean
water wells.
1.1 The need for safe drinking water
Access to safe drinking water is a foundational first step to poverty reduction and
decreased mortality rates. It can prevent the spread of waterborne and sanitation-related diseases.
The World Health Organization (WHO) reports that around 2.2 million people die annually from
water-related diseases (2014). When clean water is available, there are lower mortality rates due
to water-related diseases including cholera, diarrhea and malaria (UNDP-UNEP, 2004).
As described above, clean drinking water can be a scare resource in parts of Africa and
can require people to travel long distances to access a safe drinking source. It is common for
people living in the sub-Saharan Africa to wake up early in the morning to travel to get clean
water. One village chief describes the situation, saying “If you want clean water, then you rise up

3

early at 4 a.m. and go to a borehole in the neighboring village” (Ariet village chief, 19 September
2008, conversation).
Women and girls are primarily responsible for the collection of water in sub-Saharan
Africa (UNICEF and WHO 2012). Figure 1 shows a girl collecting water from an unprotected
hand dug well in Ariet village, 315 km northeast Kampala, Uganda. Because of this
responsibility, they are often put at greater risk for violence as a result of the distance that must
be traveled to obtain clean water. This also limits their ability to attend school. Improving access
to safe drinking water can serve as a way to keep girls in schools (Faeth and Weinthal 2012).
Another of the MDGs, Goal 5, is to achieve gender equality. This goal includes addressing goals
of eliminating violence against girls and women. Making progress in the availability of clean
water will very likely provide benefits in this critical area as well.

Figure 1 Girl collecting water from an unprotected natural spring
Photographed by Simon Peter Esaku

4

1.1.1 Helping Provide Clean Water
While access to clean water is improving, there is still a great need for more access.
Governments and non-governmental organizations (NGOs) are partnering to provide clean
drinking water to communities in sub-Saharan Africa. GIS offers great opportunities to help
agencies improve clean water resource planning.
While GIS-based models generally require large amounts of high-resolution data, the
methods developed here are designed to be applied in data-deficient areas in sub-Saharan Africa.
Because aid and development organizations working in these areas often lack funds to acquire
commercial satellite data, the methods created here focus on using data that is available to the
public and is available at no cost.
The framework created in this study is designed to address risks to clean water, improve
sanitation at health facilities, and improve children’s attendance at school through limiting the
time required to travel to get safe drinking water. To increase performance in achieving greater
access to safe, clean drinking water, the analytical methods developed here can be used to easily
identify locations where demand for water is high (e.g. near schools) and there is a lower risk of
contamination. The result is an affordable solution to help NGOs improve planning for water
wells and thereby provide a better opportunity for communities to enjoy a higher quality of life.
1.1.2 Social Context of Clean Water
There are many perspectives on the effectiveness of the MDGs. This study uses the MDGs
as a point of reference for the need, scope and general benefits of clean drinking water to a
community. The discussion surrounding the MDGs is widely published. For an overview of the
discussion, Shobha Raghuram, an independent researcher, responded to several articles

5

published by the UN. In the article, there are both critical and positive appreciations of the efforts
of the MDGs (National Institute of Advanced Studies 2008). Additionally, for a discussion on
the MDGs in sub-Saharan Africa, retired professor of Education at the University of Zambia, Dr.
Michael Kelly, describes several limitations and positive ways forward in his article for the
Institute of Development Studies (2013).
Additionally, because of its scarcity, clean water can be a source of contention. There have
been clashes between people over access to water, including incidents of violent conflicts in sub-
Saharan Africa (BBC 2006). This highlights the social, cultural and political issues involved with
water resources. However, it is beyond the scope of this project to consider all factors and it
focuses on key environmental and social factors such as health and education.
1.2 Goal of this Project
The research goal of this project is to create a site suitability framework that, when
implemented, will produce recommendations for optimal water well placement that minimize the
risks and maximize the benefits for people. This is the study portion of the project. The
framework is then applied to a specific context as a model. This implementation of the
framework is the model fitting study. In the implementation of the framework, the model is
refined to improve performance for when a user applies the framework to his or her context. In
summary, this project has two components: (1) the development of the framework (study), and
(2) the implementation of the framework (model fitting).
The model created in this project is in a GIS format that can be applied in data-deficient,
developing regions. The intent of this effort is to produce information that can be used to
improve water resource planning. The suitability map created can be used by NGOs as a

6

precursor to well siting surveys, thus limiting the extent of the area that needs to be examined in
costly and time-consuming detail.
An aim for the project is that the framework be easily implemented by interested NGOs.
This project is intended for use by NGO practitioners and is written to be understood by
technical experts in both GIS and non-GIS related fields. Therefore, Chapter 2 provides an
introduction to relevant GIS concepts. Then in Chapter 3 and 4 the methods are described
progressively in more detail. The methods are discussed in each of the chapters at increasing
levels of specificity.
1.3 Scope of the Framework
Certainly, using data about groundwater is preferable when assessing locations for water
wells. However, quality data on groundwater and sub-surficial geology are rarely available in
this region. Rather than develop a model suitable for a data-rich context such as the United States
and then apply it to data-deficient regions, a constraint was placed on the design to use data
readily available for sub-Sahara Africa. This helps to ensure the methods used are suitable for the
region. To this end, it was necessary to design a model that does not use groundwater, geology or
high-resolution commercial satellite data. To ensure that the framework could be replicated
globally, only publicly available datasets with global coverage were selected. Given these data
limitations, rather than focusing on where clean water is likely to be found, the model identifies
areas that are more likely to be free of contamination and easily accessible to the local
population. Therefore, in order to find areas with minimal contaminants to drinking water and
where benefits are maximal in terms of access, criteria were developed based on (1) contextual

7

risks, (2) availability of data and (3) proximities to community infrastructure. How the criteria
were selected is described in Chapter 3.
Moderate resolution digital elevation data was obtained through United States Geological
Survey’s (USGS) EarthExplorer program (http://earthexplorer.usgs.gov/). In addition to the
globally available data sets, point locations of community infrastructure were used. It is an
assumption in this study that an implementing agency will have this type of data. This includes
data elements such as the location of schools, communities, existing water sources, latrines, and
other facilities.
The general framework is designed to be applied throughout sub-Saharan Africa. To
demonstrate its use and to evaluate its effectiveness, an area in the Nakasongla District of
Uganda was selected as an example study where the generic framework can be applied. The area
is typical of the region in that it demonstrates what data are available for contexts for which the
generic framework is designed. The data about communities and schools was captured through a
partnership between the community and a 501(c)3 non-profit humanitarian NGO, World Vision,
whose goal is to improve the quality of life for children and families by tackling the causes of
poverty and injustice.
Given the remoteness and data-deficient nature of the study region, the effectiveness of
the general framework was evaluated by comparing the areas identified as having minimal risks
and maximal benefits with the locations of existing water well sites. While this assumes that the
existing wells are located at ideal sites, an assumption that is not likely to be universally correct,
it does provide a way to refine the model and serves as one measure of success.

8

This study is the first part of a larger project. Future research will include a field
evaluation of the results from the application of the framework. Areas that are identified as the
most suitable locations for a clean water well will be tested in the field to determine if the
locations are, in fact, the most suitable locations.
1.4 Project Workflow
GIS-based suitability projects often have similar workflows, and this project follows a
basic approach, beginning with a detailed literature and methods review and the careful selection
of the criteria to be analyzed. An outline of the project workflow is provided in Figure 2.

Figure 2 Overview of project workflow
1.5 Report Organization
This report continues on to Chapter 2 with a review of relevant published literature on the
current water situation is sub-Saharan Africa, site suitability analysis, and modeling of risks to
clean water. The third chapter describes the development of the general framework in terms of
methodology, suitability criteria and data. The fourth chapter describes the implementation of the

9

general framework. This includes sections describing the study area, the implementation, the
evaluation of results, and model refining. This report concludes with a final chapter discussing
research findings and areas for future work.

10

CHAPTER 2: RELEVANT BACKGROUND RESEARCH
This section presents a description of the potable water situation in rural Africa, followed
by a review of current water siting practices. Then a report outlining guidelines for siting surveys
for water wells is evaluated for its applicability in the sub-Saharan context. Next, previous
research on risks to safe drinking water is summarized. Finally, the methods used for site
suitability are reviewed.
2.1 Potable Water Situation in Rural Africa
Groundwater provides a high proportion of sub-Saharan Africa’s population with a
drinking water source. Groundwater, rather than surface water, is the preferred source of
drinking water because aquifers and water wells have a degree of natural protection from
contamination and drought (MacDonald, Davies and Dochartaigh 2002). Groundwater has the
benefit of being naturally protected from bacterial contamination and is a reliable source during
droughts (Lewis no date). Alternatively, surface water is often polluted, and infrastructure for
water pipes is costly. Therefore, groundwater is likely to remain a reliable source of drinking
water.
Nevertheless, groundwater aquifers can also become contaminated from sources such as
latrines, garbage dumps, corrals, cemeteries, and through poorly constructed wells (UNICEF
1999). Additionally, water wells can serve as a channel to transmit contaminated surface water
into an aquifer. Therefore, there is an increasing need to install wells in areas that are least likely
to be contaminated.
The basic principle of a clean water well is that a hole is drilled into a groundwater
source and water is then extracted with the help of a pump. Three types of wells are common to

11

access groundwater in Africa. The most common type is a hand dug well (WaterAid 2008).
These are normally uncovered and the most easily contaminated of the three types. While hand
dug wells are not ideal, when installed correctly they can be effective. Because these wells are
typically open, they require daily cleaning, which unfortunately is often not regulated (Awuah et
al. 2008).
The second type of well is the shallow well. These wells cost more than hand dug wells
but provide a more protected drinking water option. This type of well is created by drilling
through dirt and installing a pump. Pumps can be manual or mechanized. In sub-Saharan Africa
and other rural contexts, the diameter of a drilled well is usually 50 mm. The length of a drilled
well can range between 35 m to 300 m. A concrete slab is normally used to cover the well. These
wells are normally sealed and have higher protection against contaminants getting into the water
supply. The wells are susceptible to contamination, however, if a seal breaks or if there is a crack
in the pipes.
The third type of well, the deep well, is the most expensive. These wells can be over 270
m deep. These wells typically have a mechanized pump or pump house and can cost up to
$30,000 USD (http://thewaterproject.org/). Therefore, since drilled wells (shallow and deep) are
less susceptible to contamination compared to hand dug wells, the framework identifies suitable
locations of shallow and deep wells to encourage the use of improved drinking wells. However,
the framework developed in the study can also be used to identify suitable locations for hand dug
wells.

12

2.2 Site Selection for Siting Water Wells
In 2010, the Rural Water Supply Network of Illinois published guidance on best practices
for siting surveys for water wells. Even though this report is specific to Illinois and the geology
and infrastructure referenced are different from the context where the general framework will be
used, the workflow and many issues it discusses are still applicable. While the network describes
the process as siting surveying, it is one form of site selection. The current standard workflow is
outlined in Figure 3. There are four components to their standard workflow: (1) groundwater
availability, (2) impacts and risks of a new well, (3) water use (requirements of a well) and (4)
access to a source.

Figure 3 Components of current well site selection
Source: Rural Water Supply Network (2010)

The general framework developed in this study is intended to be a precursor to the site
surveying work. While it does not replace any of the existing techniques, it does supplement

13

them and can increase their effectiveness. When the general framework is implemented, it
provides suggestions to siting survey teams of areas that have minimal risk and maximum
potential benefits. Even with this information, geophysical data about groundwater depth, water
quality and geologic formations is still required.
The first component of the site selection process addresses groundwater availability. In
sub-Saharan Africa there are several challenges to effectively complete this first step. The reason
for this challenge is a lack of universal information about groundwater. In fact, Adelana and
Macdonald observe that in many areas of sub-Saharan Africa, there is relatively little attention
paid to the systematic information gathering about groundwater resources (as cited in Danert
2014). Where there are maps available that identify groundwater sources, the accuracy of the
information is not sufficiently reliable (MacDonald, Davies and Dochartaigh 2002). A model
developed from the framework described in this study, however, can be used to identify sites that
are likely to have minimal risks to the well. The siting team can then use the model to reduce the
number of areas in which to do a physical siting survey. The model will, even in this first step,
result in a more efficient use of resources in terms of time and money.
The second component of well site selection identifies the potential impact of a well on a
groundwater source. It assesses the potential risks to groundwater that a new well might cause. It
does not address actual risks to drinking water. Since this component does not address risks to
drinking water, the framework developed in this study does not improve this component.
The third component evaluates the water at a site. The value added by the framework is
the ability to evaluate a site for the contamination risk. The general framework improves this
evaluation by pre-identifying risks to the clean water and identifying community needs for clean

14

water (population and other factors). Water quality sampling is still required because it can
determine whether it has high natural contents of arsenic, fluoride and iron. Arsenic and fluoride
are toxic in high concentration, and their occurrence is related to hydrogeological conditions
(Rural Water Supply Network 2010).
The Rural Water Supply Network’s guidance also identifies the need to avoid point
contamination sources such as pit latrines, septic tanks, livestock pens, and solid waste dumps
identifying risks on a micro level, one site at a time. Therefore, there is a need for a macro level
decision support tool (i.e., the application of the framework developed here) that is able to
identify high-risk areas more broadly. This ensures new wells not only avoid contaminants, but
are also installed to achieve maximum coverage (i.e. within 1.5 km of households).
The general framework developed here improves upon their method by using GIS to add
multiple criteria (e.g. locations of schools and health clinics) to ensure the investment in clean
water benefits the highest feasible number of community facilities. The map generated as a result
of implementation of the framework can be used to engage community members and supplement
discussions by providing quantifiable suggestions for where a well is most suitable.
As discussed, the framework does not replace the site suitability process identified by the
Rural Water Supply Network—it improves upon it. The framework, when applied in a specific
context, can supplement local knowledge and help drilling contractors reduce the number of
siting surveys that must be conducted. This improvement in efficiency is possible because the
framework pre-identifies the specific locations where there may be risks to water sources.

15

2.3 Previous Research on Risks to Drinking Water
To ensure a water source is protected, it should be upslope from a source of
contamination (UNICEF 1999). As mentioned above, water wells are susceptible to
contamination at three areas: the opening on the surface, the piping from groundwater to surface,
and the groundwater source. As described above, groundwater is a good resource to help provide
more clean water coverage. However, groundwater is not a fail-safe resource when it comes to
providing clean drinking water. Groundwater can be contaminated when it is underneath areas
containing pathogens and chemicals derived from fecal and other waste (Rural Water Supply
Network 2010).
Furthermore, a report from the University of Illinois at Urbana – Champaign Water
Resources Center identifies common sources of contaminants of drinking water. The list
includes: septic tanks, road salt, underground gas storage tanks, manure piles, fertilized cropland,
and solid waste disposal sites (1990). The report also provides recommended minimum distances
between a well and a potential contamination source. The list of minimum distances is prefaced
with the guidance that a well should be as far removed as possible from potential contamination
sources. Table 1 shows the minimum setback for common sources of contamination using
guidelines from the Illinois Water Resources Center.
Table 1 Minimum distance from sources in Illinois
Common Contamination Source Minimum Distance
Sources and Routes of Contamination 61 m
Existing Cesspools 46 m
Leaching pits 31 m
Septic Subsurface seepage tile or manure pile 23 m
Sewer line and Septic Tank 15 m
Lakes, Ponds or Streams 8 m
Source: University of Illinois at Urbana – Champaign Water Resources Center (1990)

16

These distance recommendations are helpful in understanding the relative potency of a
particular pollutant source, but because the distance recommendations are designed for a person
installing a small well on personal property, the value of the specific recommendations is limited.
Finally, the University of Illinois Water Resources Center recommends that a well should
be placed on the side of the contaminant source opposite the flow of groundwater. For example if
groundwater flows to the south, a well should be placed as far north of the contaminant source as
possible. This type of general consideration is included in the general framework.
Additionally, the U.S. Environmental Protection Agency (EPA) recommends that well
owners have a zone of protection around a well to prevent contamination. The recommended
distance designated to limit risk of groundwater contamination is 30 m (California State Water
Resources Control Board 2011). Both the guidance from University of Illinois Water Resources
Center and the EPA are recommendations to prevent groundwater contamination as opposed to
surface water contamination. Since surface water can travel faster than groundwater, a
contaminant can travel further than the minimum distances recommended by both agencies.
Therefore, the framework described in the next chapter was built on the foundational principle
that the greater the distance a well is located from a contamination source, the more suitable is
the location.
2.4 Methods for Site Suitability Analysis
To find suitable locations, a user can overlay different layers in a GIS. Ian McHarg
pioneered the use of overlay for suitability evaluation. In 1969, in his seminal work, Design With
Nature, he showed how a user could superimpose a set of transparent layers, one for each

17

criterion, to create an overall suitability map. This technique is regarded as a precursor of
modern GIS overlay (Qiu et al. 2014).
There are several methods for modeling suitability. One method for modeling suitability
divides locations into two groups, or sets: those that are suitable and those that are not. This
method is known as Boolean overlay and it evaluates whether a location meets each criteria, on a
yes/no basis (Mitchell 2012). This is a particularly useful method when boundaries or attributes
of a criterion are crisp. Alternatively, where these are not crisp, there are two common methods
that allow a user to rate locations on a scale from more suitable to less suitable. These two
methods are weighted overlay and fuzzy overlay.
Weighted overlay allows a user to assign importance to a specific criterion. When a user
assigns importance, a weight is assigned to the layer (Mitchell 2012). This is also known as the
percentage of influence for each layer. When weighted overlay is applied in a raster context, cell
values of each input layer are multiplied by their percentage influence, and the results are added
together to create the output raster. For reasons described in section 3.1, weighted overlay was
less suitable for this analysis.
Fuzzy overlay allows a user to rate suitable locations when criteria are hard to quantify or
when the relationship between specific criteria and suitability are not well defined. Additionally,
Mitchell points out, “fuzzy overlay is particularly good for creating a suitability model that
attempts to capture the knowledge of experts in a particular field” (93). Fuzzy overlay was used
in the general framework developed in this study. Fuzzy logic is built upon the concept of fuzzy
sets which allow partial membership within a range of 0 to 1 to represent the extent to which an

18

entity belongs to a certain class. In site suitability analysis, the degree of belonging is with
respect to the set of suitable locations.
Fuzzy overlay has been used for suitability analysis for a wide range of applications
including finding the best locations for wind power systems, rice growing areas, and solid waste
landfill (Demesouka et al. 2014; Kihoro et al. 2013). Examples of the use of fuzzy logic to model
risks include groundwater vulnerability risk mapping (Nobre et at. 2007) and landslide
susceptibility modeling (Chalkias et al. 2014). In both studies, the use of fuzzy logic highlights
how it can handle uncertainty where boundaries and attributes are difficult to define.
The application of fuzzy logic to model water resources was accomplished by Tsiko and
Haile in 2011. The authors used fuzzy logic to model optimal sites for locating water reservoirs
in Eritrea, in the Horn of Africa. That project is similar to this water well suitability project in
that it develops a framework and then applies it to a real world context. One reason the authors
selected fuzzy logic for water resource planning was because their decisions regarding criteria
were accompanied by ambiguities and vagueness. This meant there was a lack of certainty about
the measurement of the criteria. The authors note, “This makes fuzzy logic a more natural
approach to this kind of Multi-criteria Decision Analysis (MCDA) problems” (Tsiko and Haile
2011, 257).

19

CHAPTER 3: A GENERAL FRAMEWORK FOR SITING CLEAN WATER WELLS IN
SUB-SAHARAN AFRICA
This chapter describes the general fuzzy suitability framework developed here to identify
locations where risks to clean water wells are minimized and benefits to people are maximized. It
begins with sections that describe why fuzzy overlay was used. Next, the argument for modeling
only surface and subsurface contaminant flows using surface slope derived from free, readily
available global data is presented. Finally, the suitability criteria in the general framework, and
the data limitations and availability for sub-Saharan Africa are discussed.
3.1 Fuzzy Overlay
The framework developed in the study relies on fuzzy membership functions to assign
suitability to individual layers. If Boolean logic were used throughout the framework, it would
not be able to capture the vagueness or continuous nature of the data required for the framework.
Additionally, the use of fuzzy membership captures the nuances of partial goodness or badness
of a location with respect to suitability. When evaluating suitability of a location, a user
determines the method by which locations are evaluated for membership. Depending on the
user’s decision, the strength of membership in the suitability set changes. For example, low
membership would be assigned in the suitability set to locations that are near a risk.
Weighted overlay was not used for this project because information that can be used to
assign the weights is not known for each specific context in which it might be applied.
Additionally, due to the contextual nature of risks to clean water, the general framework
developed in this study does not lend itself to setting weights that can be applied universally.

20

Fuzzy overlay was selected for the general framework for several reasons:
(1) Uncertainty of the available data to accurately represent a criterion.
(2) Uncertainty of a crisp break point between values for suitable areas and unsuitable
areas.
(3) Continuous nature of some environmental phenomena which are represented as
classified data. Wetlands and vegetation data are two examples of this in which the
phenomena represented by the data can often have non-crisp boundaries where the
change from one class to the next is gradual rather than abrupt (i.e. crisp). Crisp
boundaries, on the other hand are well defined, it is evident when one phenomenon or
class stops and another starts.
(4) Uncertainty whether the resolution of data can capture elevation heterogeneity. The
suggested base data is a 90 m elevation layer and it is likely the resolution is unable to
capture accurately the elevation change within a grid especially when the framework
is applied to contexts with high frequency of elevation change.
In this project, fuzzy logic was used in two steps: assigning fuzzy membership and
performing fuzzy overlay. The fuzzy membership codifies each of the values of each data layer
on a scale of 0 to 1. This was done for each criterion. Most often in the framework the fuzzy
aspect is the distance from a contaminant at which it is safe to install a well. The assignment of
fuzzy membership for each criterion as it is applied in the implementation study is described in
Chapter 4.
Furthermore, once all layers have been assigned individual fuzzy membership, then fuzzy
overlay is used to combine multiple criteria into a single fuzzy membership layer. Each location

21

is given a value representing the strength of membership in the set of suitable locations. As a
result of the fuzzy overlay process, each cell is given an aggregate fuzzy membership value.
Implementation of the fuzzy overlay process is discussed in Chapter Four.
3.2 Modeling Risks Where Groundwater Data is Limited
The goal of this project was to create a general framework that can be applied throughout
sub-Saharan Africa to identify locations of minimal risks to water wells. The framework was
designed to create a decision support suitability map to be used as a tool by NGOs and
governments. Many NGOs and African governments have limited financial resources, so cost
was an important factor in the framework design. By excluding areas that are unsuitable, NGOs
and governments can save money by not undertaking water well site surveys in locations that are
at high risk, or have minimal positive impacts on the community.
To be useable, the data of the framework must be openly accessible and affordable to
users in sub-Saharan Africa. With an emphasis on low cost, and wide applicability, data sets
were selected to fulfill this requirement. This excluded higher resolution data available through
commercial companies from being used as input for the framework. Instead, lower resolution
data was used to ensure the framework is usable, replicable and feasible for governments and
NGOs in sub-Saharan Africa to implement.
As described above, there are three areas where a water source can be contaminated:
surface opening of a well, the underground piping, and the groundwater itself. Groundwater
contaminants usually form a concentrated plume underground that flows along the same path as
the groundwater. Among the factors that determine the size, form and rate of movement of a
contaminant plume are the number and type of contaminants and the speed of groundwater

22

movement (Jakhrani et al. 2009). When choosing locations to drill for water, it is ideal to know
where (1) the groundwater is least likely to be contaminated and (2) water most likely is to be
found. In order to do this, a map of groundwater is required. Unfortunately, groundwater data is
not available in many areas in rural sub-Saharan and therefore it could not be included as an
input for the framework. This meant that an alternative method to model the movement of a
contaminant was needed.
Elevation data that is now globally available through the Shuttle Radar Topography
Mission provides an alternative. Surface slope does not always reveal the direction a pollutant
might flow once it infiltrates the ground, but it can reveal the direction a pollutant might flow on
surface or near-surface. Therefore, slope, derived from elevation data can be used to determine
the surface and near surface flow of water and likely flow of a surface or near surface
contaminant.
In this framework, to avoid contamination, a water well should be on the side of a
contaminant source opposite the flow of surface or near-surface water. This decision is in line
with the instruction manual for drilling of water wells commissioned by the PRACTICA
Foundation, USAID, and UNICEF. The manual states, “It would not be good to place the well
down-slope (downstream) of the latrine, but rather on the same level or higher up (up-slope) of
the latrine” (Van der Wal 2010, 9).
Having presented the argument for modeling only surface and subsurface contaminant
flows using surface slope, we now turn to a consideration of the general risks to clean water
wells.

23

3.3 Suitability Criteria for the General Framework
This section describes the suitability criteria identified for use in the framework. There
are two categories of criteria: places to avoid (risks) and places to favor, which includes locations
near facilities providing public health and education and useful infrastructure locations which
would maximize clean water coverage. There are thirteen criteria that comprise seven categories.
3.3.1 Locations to Avoid and the Implied Constraints
This section describes the risks to a safe drinking water supply in sub-Saharan Africa and
implied constraints for the selection of a location for a well. In this section, eight factors are
described: human waste, animal waste, solid waste dumps, agricultural fields, burial sites,
wetlands, stagnant water and existing wells.
Human waste
Human waste can spread disease to a water supply. Bacteria from human feces are one of
the most serious threats to clean water (UNICEF 1999). Bacteria from a latrine can contaminate
a well when surface water or near surface water gets into the well (Korab 1999). Additionally, a
well can be contaminated from a nearby latrine if the groundwater flows from a latrine toward a
well. Bacteria, viruses and parasites, originating from a latrine will flow together with
groundwater to a well (Van der Wal 2010). Other sources of contamination related to human
waste include leaking septic systems or contaminated wells elsewhere in the groundwater system
(EPA no date).
There are several types of latrines used in sub-Saharan Africa. The most common is the
traditional pit latrine. This is a simple pit sometimes covered with logs (Farmer no date). Some
latrines are lined to contain the waste while others are not.

24

One health problem heightened by poor sanitation in sub-Saharan Africa is cholera. In
May 2013 there was a cholera outbreak in Uganda. Cholera is an infection in the small intestine
that causes symptoms of diarrhea and vomiting. It is transmitted primarily through drinking
water that has been contaminated by feces of an infected person.
As a result of all of the above, a suitable location for a well will avoid areas such as
latrines and other sites of human waste.
Animal waste
Animal waste can contaminate a water source just as human waste can. It can carry
several harmful pathogens; studies have linked waterborne transmission of diseases to animal
waste (Dufour et al. 2012). Animal waste is a high-risk factor because it is typically not
contained in the same way as human waste. Latrines for human waste can be cleaned and filtered
to prevent the transmission of a pathogen into a water supply. However, animal waste is often
perceived as less dangerous to human health and therefore monitored less. Additionally, animal
waste is on the surface and more likely to travel with surface water. Therefore it is
recommended that a well be placed away from an area with concentrations of animals waste.
Solid waste dumps
Waste dumps can contain metals, acids, and other chemicals that can pollute a water
source. Dumps can pollute groundwater and surface water and therefore it is recommended that a
well be placed opposite of the downslope flow path of water from a dump.
Agricultural fields
In order to increase crop yields, farmers use chemicals in many areas in sub-Saharan
Africa. These on-farm chemicals can potentially contaminate the water supply. The nitrates in

25

fertilizers, pesticides and herbicides can cause an imbalance in the natural environment and alter
the quality of drinking water (Ongley 1996). To avoid farm chemical pollution, wells should
avoid farms where chemicals are used.
Burial sites
The decay of human and animal corpses can have a negative impact on water (Üçisik and
Rushbrook 1998). Contamination occurs as a result of germs and viruses that are generated
during the process of decay. Water can carry the germs and viruses through soil to a water
source. To avoid pollution, wells should avoid graves and burial grounds.
Wetlands
Flooding increases the spread of a pollutant. If a well is located in an area prone to
flooding, the clean water in the well can be contaminated when water flows into the well. To
avoid contamination, wells should avoid depressions in the surface and areas that experience
seasonal flooding (Rural Water Supply Network 2010).
Stagnant water
Stagnant water is also a threat to community health. For children living in Uganda,
malaria is the primary cause of death. Malaria is transmitted by female mosquitoes of the genus
Anopheles which, for breeding, prefer permanent, stagnant water bodies such as shores of rivers
and creeks or fish ponds (Uganda Ministry of Health 2014). Therefore, a clean water well should
not be in an area with increased likelihood to have stagnant water.
Existing wells
The locations of existing water wells can be used to identify areas to avoid. This ensures
that new wells will be installed at locations that improve clean water coverage and thus support

26

the MDG goal to halve the proportion of the population without access to sustainable safe
drinking water.
3.3.2 Locations to Favor
This section describes locations to favor. In these areas, when clean water is near, there
are significant benefits to people living in the area. This section highlights the social aspect of
the study and promotes health and sanitation by favoring locations near facilities that can benefit
the general public by having safe drinking water. Installing a clean water source near a school or
health facility can improve the sanitation practices at those locations. Therefore, locations to
favor include low slope, hospitals/clinics, schools, roads and villages/ populated places.
Slope
A well location must be accessible to drilling equipment. A steep slope (greater than 20
percent) can cause problems when drilling and installing a well. Additionally, an area of low
slope decreases the mechanical wear on the well equipment. Areas with 10 to 20 percent slope
are feasible for installing a well, but will require increased earth moving and grading (Building
Advisor no date). This increases labor and equipment costs. A user will need to determine an
acceptable slope in his or her context, but due to the increased cost and risks, it is not
recommended to install a well on greater than 20 percent slope. Therefore, to avoid additional
installation and maintenance costs, a well should be in an area of low slope.
Hospitals/ Clinics
Clean water wells should be placed in areas that support public health. Therefore
locations near hospitals are desired. This ensures patients are hydrated with clean drinking water
and are less likely to drink unclean water that can delay recovery. Additionally, inadequate clean

27

water, sanitation and hygiene can play a substantive role in propagating infections (Bartram and
Platt 2010). Therefore locations near health facilities are desired.
Schools
Clean water wells should be placed in areas that support education. Children often miss
school because their schools do not have adequate drinking water and or sanitation facilities
(Hillie and Hlophe 2007). With clean water in proximity of a school, children are more likely to
drink safe water and are less at risk to be ill due to water transmitted bacteria or disease.
Therefore locations near schools are desired.
Road network
Major roads help determine accessibility. Some wells require a drilling rig be brought in.
If a location has access to a major road, then drilling is more feasible and cost effective. To
increase efficiency and drilling feasibility, a well should be near a major road.
Villages
Point locations of a village, typically captured at a city center or market, can ensure
improved water coverage planning. Locations within 1.5 km of a clustering of households should
be given high membership in the suitable set to improve clean water coverage as defined by the
UN. Population data can also be associated with a village so that high population locations can
be more strongly suitable.
Now that all criteria haven been discussed, they are organized into categories. Table 2
organizes these risks and benefits into selection criteria categories and provides more detail on
how these criteria can be operationalized. Table 3 provides an overview of the general data
sources. It describes potential data sources that can be used to analyze a criterion in sub-Saharan

28

Africa. This table highlights the challenge of finding sufficient data to analyze a criterion. Many
of the criteria require data to be collected in the field to be provided by an NGO or government.
In the Possible data values/attributes column, a list of suggested attributes for data layers is
presented.

29

Table 2 General Criteria
Category Criteria
ID
Criteria Summary Statement for Selection of Criteria
1. Avoid travel path of a
contaminant from a
source
1a Should not be in the
flow path of human
waste
Human waste carries bacteria that can spread disease and cause
illness if ingested.
1b Should not be in the
flow path of animal
waste
Animal waste, similar to human waste can spread disease and cause
illness.
1c Should not be in the
flow path of solid
waste
Harmful chemicals such as metal deposits, acids, and other sources
can be found in solid waste dumps and can cause illness if ingested.
1d Should not be in the
flow path of farm
chemicals
On-farm substances such as fertilizers, herbicides and pesticides
contain chemicals that can be harmful if ingested.
1e Should not be in the
flow path of
decomposing carcasses
Germs and viruses generated from the process of decay can cause
spread disease if ingested.
2. Avoid areas prone to
flooding
2a Should not be within a
wetland
Water from flooding can seep into a clean water well and then
contaminate the water in the well.
3. Avoid areas with an
increased likelihood of a
person contracting
malaria
3a Should not be within
an area with stagnant
water
Stagnant water is the preferred breeding area of female mosquitos
which transmit malaria. Areas mosquitoes prefer should be avoided
as a precaution to prevent exposure.
4. Select areas where
drilling is feasible
4a Should not be on a
steep slope
The greater the slope, the higher the costs incurred, both in terms of
access to drilling a well and drinking from a well. Additionally, a
steeper slope correlates with increased mechanical wear on the well
equipment.

30

Category Criteria
ID
Criteria Summary Statement for Selection of Criteria
4b Should be near a major
road
A major road in the general area ensures accessibility to a potential
well site for a drilling rig.
5. Promote sanitation at
hospitals
5a Should be near a
hospital
Clean drinking water promotes safe hydration for people recovering
from illnesses.
6. Promote children’s
health
6a Should be near a
school
A clean water well near a school decreases the likelihood a child
will contract an illness from drinking water from an unclean source
and be absent from school.
7. Maximize clean water
coverage
7a Should be near a
populated place
A water well near a populated place supports the MDG goal to
improve coverage of clean water.
7b Should not be near an
existing water well
Installing a new well near an existing one creates redundancy and is
unlikely to increase water coverage.

31

Table 3 General Data Sources
Criteria
ID
Criteria Suggested Map
Layer
Justification for Selection
of Data Layer
Possible
Data
Types
Possible data
values/attributes
1a Should not be in the
flow path of human
waste
Latrines Latrines contain high levels
of human waste. The types
vary as well as their ability
to contain and sanitize
waste.
Point,
Polygon
Location of latrine, type of
latrine, # of pits,
1b Should not be in the
flow path of animal
waste
Livestock pens,
grazing areas
Areas with high
concentration of animals,
such as livestock pens and
grazing areas, correlates to
the amount of animal
waste.
Point,
Polygon
Location of pen, type of
animals, size of the area, # of
animals
1c Should not be in the
flow path of solid waste
Trash dumps Solid waste accumulates at
trash dumps. When these
sites are not properly
located they are a risk to
surface and groundwater.
Point,
Polygon
Location of dump, size of the
dump
1d Should not be in the
flow path of farm
chemicals
Agricultural Fields In many cases farmers use
chemicals to improve crop
yields. When farmers are
not trained to mix
chemicals properly, there is
an increased risk to water
contamination. Even when
used as directed, water can
be contaminated.
Point,
Polygon
Location of areas likely
using farming chemicals,
types of chemicals used,
frequency of use, type of
crops

32

Criteria
ID
Criteria Suggested Map
Layer
Justification for Selection
of Data Layer
Possible
Data
Types
Possible data
values/attributes
1e Should not be in the
flow path of
decomposing carcasses
Cemetery Cemeteries or large burial
sites have high
concentrations of
decomposing carcasses.
Point,
Polygon
Location of burial grounds, #
of carcasses, depth of
carcasses, age of burial
grounds
2a Should not be within a
wetland
Wetlands Wetlands are more likely to
flood than highlands.
Raster,
Polygons
Elevation can be used to
identify wetlands.
3a Should not be within an
area with stagnant water
Depressions Areas with little drainage or
no drainage accumulate
water and cause water to
pool.
Raster Elevation can be used to
identify areas where pooling
of water can occur in
depressions on the surface.
4a Should be on a slope of
less than 16 percent
Slope Slope affects the direction
of flow.
Raster Percent slope
4b Should be within 4000
m of a major road
Road Network Access component to
determine feasibility. It is
probable costs will increase
if no road exists in the
general area of a potential
well site.
Line Location of a road
5a Should be within 3000
m of a hospital
Hospitals/ Clinics Clean water can help a
person stay hydrated and
recover from an illness
faster. The high number of
people at a hospital who are
recovering from an illness
or injury make it an ideal
area to make a positive
impact.
Point Location of a hospital

33

Criteria
ID
Criteria Suggested Map
Layer
Justification for Selection
of Data Layer
Possible
Data
Types
Possible data
values/attributes
6a Should be within 3000
m of a school
Schools Illness can prevent a child
from attending school
and/or prevent him or her
from actively engaging
with the educational
material.
Point Location of a school
7a Should be within 1500
m of a populated place
Populated Places To maximize coverage to
the most people, villages or
a clustering of households
should be targeted for clean
water wells.
Point Location of a village
7b Should not be within
1500 m of a water well
Existing Wells To prevent installing a well
where clean water exists,
existing wells should be
avoided.
Point Location of a well

34

3.4 Appropriate Data for Analysis of Criteria
This section describes the different data layers that are appropriate for the analysis of
each of the criteria. As mentioned above, the data suggested were chosen due to their availability
and no cost. In some cases, there are several viable data options for a specific criterion. In those
cases, a user can choose the layer that is best for his or her context.
The data for analysis of the criteria of contamination risks must be collected locally by
field crews since data is site specific and is generally not available through public sources.
Usually, locations of these facilities will be collected with GPS devices. This is also true for
criteria of schools, hospitals, villages and existing water wells. There is no need for survey grade
GPS equipment due to the scope and resolution of the general framework. Therefore, GPS-
enabled smart phones and tablets can be used for mobile data collection. Where available it is
suggested to supplement data with data from other NGOs or government ministries.
Regarding village data, it is likely an NGO’s dataset will not include all populated places.
Thus it may be helpful to enhance it. A suggested source is GeoNames, an international, database
with global coverage of places (http://www.geonames.org/). Over 55 data sources are used by
GeoNames including National Geospatial-Intelligence Agency (NGA), U.S. Geological Survey
Geographic Names Information System (GNIS) and World Gazetteer. It is supported by over 60
national mapping agencies and is updated daily. Another method is to assess satellite imagery to
identify clustering of households. Landsat 8 imagery at 30 m resolution can be used for this and
is globally available.
For data on areas with an increased likelihood of flooding, there are three layers that can
potentially be used. The first is the Africa Topographic Position (ATP) layer from USGS which

35

is a binary raster layer where “lowland/depressions” indicates locations where water is likely to
accumulate due to the elevation and slope in the area. These areas are given a value of 1. The
other areas are categorized as “uplands” and are given a value of 2. This layer is different than a
layer such as the normalized difference moisture index that uses climate or soil attributes to
calculate moisture potential (USGS 2008). The ATP layer was created using a geographically
derived measure of slope for each raster cell and the contributing area from "upstream" raster
cells which expresses the potential for water flow to a location. To quality check the resulting
surface, independent estimates of water accumulation from existing wetland boundaries were
used (USGS 2008). For more technical details about how the data was created see Appendix 1.
A second viable option to identify lowlands is to use elevation data. A void-filled digital
elevation model (DEM) is recommended for implementing the framework. A good source is the
Shuttle Radar Topography Mission (SRTM). The original release of SRTM elevation data had
many NoData cells known as voids, which caused errors during data analysis. To help in the use
of this data, the USGS has produced void-filled SRTMs in which interpolation methods have
been used to fill in the data gaps. There are several methods for this depending on the purpose of
use. Several void filling interpolation methods are evaluated by Reuter et al. 2007. A void-filled
DEM can also be used to identify slope for criterion 4a.
Another viable option to identify lowlands is a HydroSHEDS layer. HydroSHEDs
provides hydrographic information in a consistent format for large portions of the earth. It is
derived from the SRTM elevation data at 90 m resolution. Hydrologically conditioned
HydroSHEDS are created using a sequence of automated procedures. Existing methods of data
improvement and newly developed algorithms have been applied, including void-filling,

36

filtering, stream burning, and upscaling techniques (USGS 2008). It was created to support
regional and global watershed analyses, hydrological modeling, and freshwater conservation
planning. More information about the data development of HydroSHEDS is in HydroSHEDS
Technical Documentation v1.0. It is available for download at
http://hydrosheds.cr.usgs.gov/datasets.php.
There are two sources for data on the locations of depressions. One method is to generate
a sinks layer from a DEM. A sinks and depressions source layer can be created from a void-filled
DEM, but not from a hydrologically conditioned DEM since sinks have been removed from
them to improve the hydrological modeling process. Sinks can be a manifestation of error, but
there can be sinks that are real. Real sinks are areas with no external drainage meaning they are
depressions in the surface. Another option is the Africa ATP layer described above. This layer
has data about lowlands and depressions. Since these are two different concepts, a user may
desire to separate the ATP layer to analyze the lowlands and depressions separately. The
decision to separate the layer, and the process used to do so, is described in Chapter 4.
Open Street Map (OSM) is a good source for data about public infrastructure. Data about
roads, water boundaries, populated places, and administrative boundaries can be downloaded at
no cost. OSM is a collaborative project to create a free, editable map of the world. It is
maintained by contributors, often local knowledge holders who update the map. The rate at
which an area is mapped is dependent upon what a user community uploads into the system.
Even where communities are not active there is usually reliable data on roads, shorelines and
administrative boundaries. These layers are mainly provided by governments who released their
data into the public domain. The data can be downloaded for free and imported to ArcGIS

37

through a simple process. Esri provides a free toolbox for importing and working with OSM data
in ArcGIS 10.x. It can be found at
http://www.arcgis.com/home/item.html?id=16970017f81349548d0a9eead0ebba39.
Table 4 below provides an overview of possible data sources to use when the model is
implemented. The table summarizes information described above and describes how the data
was created, the source, and resolution.
In summary, in this chapter, the general fuzzy suitability framework was presented. The
selection of fuzzy overlay was explained, the rationale for modeling contaminant flow was
described, and the general criteria to identify risks to clean water were discussed. In the next
chapter, the implementation of the framework is described. As part of the implementation, the
initial results are evaluated and the process used to refine the framework is described.

38

Table 4 Recommendations for Data
Criteria
ID
Criteria Suggested
Map Layer
How Data was Created Source Resolu-
tion
Extent of
Coverage
Accessibility
1a Should not be in the
flow path of human
waste
Latrines Data should be collected in
the field with GPS devices.
Government
/ NGO
6 m Organization
Dependent
Organization
1b Should not be in the
flow path of animal
waste
Livestock
pens
Data should be collected in
the field with GPS devices.
Government
/ NGO
6 m Organization
Dependent
Organization
1c Should not be in the
flow path of solid
waste
Trash
dumps
Data should be collected in
the field with GPS devices.
Government
/ NGO
6 m Organization
Dependent
Organization
1d Should not be in the
flow path of farm
chemicals
Farms Data should be collected in
the field with GPS devices.
Government
/ NGO
6 m Organization
Dependent
Organization
1e Should not be in the
flow path of
decomposing
carcasses
Cemetery Data should be collected in
the field with GPS devices.
Government
/ NGO
6 m Organization
Dependent
Organization
2a Should not be within a
wetland
Wetlands There are three options for
this criteria: (1) Africa
Topographic Position, (2)
SRTM DEM or (3)
Conditioned DEM through
HydroSHEDS.
USGS,
USGS,
USGS
100 m,
90 m, 90
m
Africa Public
3a Should not within an
area with stagnant
water
Sinks/
Depressions
There are two options for
this criteria: (1) Africa
Topographic Position, or
(2) SRTM DEM.
USGS,
USGS,
USGS
100 m,
90 m, 90
m
Africa Public

39

Criteria
ID
Criteria Suggested
Map Layer
How Data was Created Source Resolu-
tion
Extent of
Coverage
Accessibility
4a Should be on a slope
of less than 16 percent
Slope SRTM 90 m or better DEM USGS,
USGS,
USGS
90 m Global Public
4b Should be within 4000
m of a major road
Road
Network
Community mappers based
on local knowledge where
an active community exists
Open Street
Map or
Digital
Chart of the
World
6 m Global Public
5a Should be within 1500
m of a hospital
Hospitals/
Clinics
Data should be collected in
the field with GPS devices.
Government
/ NGO
6 m Organization
Dependent
Organization
6a Should be within 3000
m of a school
Schools Data should be collected in
the field with GPS devices.
Government
/ NGO
6 m Organization
Dependent
Organization
7a Should be within 1500
m of a populated place
Populated
Places
Data should be collected in
the field with GPS devices.
Government
/ NGO
6 m Organization
Dependent
Organization
7b Should not within
1500 m of a water
well
Existing
Wells
Data should be collected in
the field with GPS devices.
Government
/ NGO
6 m Organization
Dependent
Organization

40

CHAPTER 4: IMPLEMENTATION STUDY
In this chapter the general framework is implemented as a GIS model for a specific
location in order to assess its feasibility. As explained in this chapter, the implementation
identified weaknesses of the framework which led to modification of the model through an
iterative model fitting process. This chapter begins with a discussion of the study area, the data
used in the implementation and the fuzzy methods used. Then, the results are assessed and the
model refinement is discussed. This includes a presentation of the performance for both the
preliminary and refined model.
4.1 Study Area
A study area in central Uganda was selected to act a test case of an implementation of the
framework. This area is representative of other contexts in sub-Saharan Africa where the
framework might be applied. The area was also chosen because data to test the general
framework was made available from an NGO. The study area is located about 160 km north of
Kampala in the Nakasongola district (see Figure 4).

41

Figure 4 Overview of Study Area

The study area is approximately 630 sq km and is near the southwestern edge of Lake Kyoga
(see Figure 5).

42

Figure 5 Overview of Study Area with Lake Kyogo

The region is one of Uganda’s driest and is recognized as an area experiencing the effects
of climate change, notably an increased frequency and severity of floods and droughts (Egeru
and Majaliwa 2009). Historically, the region experiences two rainy seasons a year. The first
season is March to May and consists of regular heavy rains. The second season is August to
December and consists of irregular light rains. Inconsistent rainfall and intense rains make the
area susceptible to flooding. Information about the weather near the study area was found using a
weather station in Soroti, about 86 km northeast of the study area (see Figure 6).

43

Figure 6 Weather Station in Soroti, Uganda
At the Soroti weather station, the recorded 30 year rainfall average is 53.7 inches.
Additionally, the average number of rainy days is 126 and the average temperature is 76.9
degrees Fahrenheit. Figure 7 shows the average precipitation for the area by month.

44

Figure 7 Overview of Average Rainfall in Soroti
Source: World Weather Online (2014)

The elevation in the study area ranges from 1028 m to 1271 m. The study area is
generally flat with a few isolated areas with a slope greater than 6 percent. There are wetlands on
the east and west of the study area (see Figure 5).
For their food and livelihood needs, residents rely on the natural resources from the land
and especially Lake Kyoga. Livestock production is a major source of livelihoods and only one-
third of the people are crop farmers in the district (Muruli 2009). Major food crops include
cassava, maize, sweet potato, banana, potato, and millet. The area has suffered considerably from
soil degradation and deforestation which has caused erosion in many areas.
4.2 Data
As with any implementation of a general framework, the implementation is limited to the
data available. In this implementation study, due to limited availability of data, only seven
criteria of the fourteen were used in the suitability model. An existing wells layer (for criteria 7b)

45

was available but, as explained above, it is kept out of the analysis to use as a means of
validation of the model. In the tables below, the Unique Criteria ID that were established in
Chapter 3 are used throughout Chapter 4.
When the general framework is implemented, a user will also need make decisions about
the specific distances used to analyze the criteria. In this study, different distances were tested
until an acceptable result was identified. The preliminary model, along with the refining process,
is discussed in the model fitting section (4.5) below. As mentioned above, the model was refined
by comparing the proposed suitable sites against the location of existing wells in the study area.
The distances reported below are from the final set of distances.
The figures below provide an overview of the data acquired through World Vision. The
distribution of schools and health facilities are shown in Figure 8. The distribution of hygiene
and sanitation related infrastructure is shown in Figure 9. The points are shown on two maps to
reduce the overlap since many of the points are located close to one another.

46

Figure 8 Map of Community Infrastructure

Figure 9 Map of Hygiene and Sanitation Infrastructure

47

The eight criteria (7 for the model plus wells) described in Table 5 were selected because
of the data available in the study area. Some of the World Vision data was supplemented with
public data as described in more detail below. The data used are explained in Tables 6 and 7.
Table 5 Suitability Criteria in Implementation Study
Category Criteria
ID
Criteria Measure
1. Avoid travel path of a
contaminant from a
latrine
1a Should not be in
the flow path of
human waste
Flow Distance
2. Avoid areas prone to
flooding
2a Should not be
within a wetland
Yes/No
3. Avoid areas with an
increased likelihood of a
person contracting
malaria
3a Should not be
within an area with
stagnant water
Yes/No
4. Permit areas where
drilling is feasible
4a Should be on a
slope of less than
16 percent
Percent rise
4b Should be within
4000 m of a major
road
Distance from a major road (m)
6. Promote children’s
health
6a Should be within
3000 m of a school
Distance from a school (m)
7. Should maximize
clean water coverage
7a Should be within
1500 m of a
populated place
Distance from a populated place
(m)
7b Should not be
within 1500 m of
an existing well
Distance from an existing well (m)

48

Table 6 Implementation Data Requirements
Criteria
ID
Criteria Specific Map
Layer
Data
Type
Possible data values/attributes
1a Should not be in
the flow path of
human waste
Latrines Point Location of latrine
2a Should not be
within a wetland
Africa
Topographic
Position
Raster The Africa Topographic Moisture
potential layer categorizes the
surface into two categories:
Lowlands/depressions and highlands.
Lowlands/depressions are given a
value of 1 and uplands are given a
value of 2.
3a Should not within
an area with
stagnant water
Africa
Topographic
Position
Raster The Africa Topographic Moisture
potential layer categorizes the
surface into two categories:
Lowlands/depressions and highlands.
Lowlands/depressions are given a
value of 1 and uplands are given a
value of 2.
4a Should be on a
slope of less than
16 percent
DEM Raster Percent slope
4b Should be within
4000 m of a
major road
Major Roads Line Location of road
6a Should be within
3000 m of a
school
Schools Point Location of a school
7a Should be within
1500 m of a
populated place
Populated
Places
Point Location of a village
7b Should not be
within 1500 m of
an existing well
Existing
Wells
Point Location of a well

49

Table 7 Data Specifics
Crite
ria
ID
Specific Map
Layer
How data was created Resolution/
Horizontal
Accuracy
Quantity Source Extent of
Availability
1a Latrines Data was collected in the field by World Vision
staff with handheld GPS devices.
6 m 18 World
Vision
Study Area
2a Africa
Topographic
Position
This layer was derived from the Compound
Topographic Index (CTI) which is a
topographically derived measure of slope and the
areas of “upstream raster cells” (USGS, 2009).
Satellite Imagery was also used to create a
wetlands layer.
100 m grid USGS Africa
3a Wetlands
Polygons
Layer derived from satellite imagery. 30 m WRI Uganda
4a DEM,
SRTM3N01E03
2V1
SRTM elevation data was derived using two
radar images taken from slightly different
locations to calculate the surface elevation
(USGS, 2008).
90 m grid USGS Global
4b Major Roads Created by the community of Open Street Map
Contributors
Unknown 4 Open Street
Map
Global,
Community
Based
6a Schools The primary source was World Vision. These
points are World Vision schools only. There are
other schools not included in this layer because
they were unavailable. The performance of the
model is likely to improve with a more robust
school layer.
6 m 15 Government
/ NGO
Study Area

50

Crite
ria
ID
Specific Map
Layer
How data was created Resolution/
Horizontal
Accuracy
Quantity Source Extent of
Availability
7a Populated Places The primary source was World Vision. 6 m 99 World
Vision
Study Area
Landsat 8 satellite imagery was used to identify
areas with clustering of households.
USGS Global
GeoNames data was used to supplement data
where places seemed to be missing.
GeoNames Global
7b Existing Wells Data was collected in the field by World Vision
staff with handheld GPS units. Of the 76 wells
only 7 were not installed through World Vision.
Therefore the data is not likely to be complete.
6 m 76 World
Vision
Study Area

51

4.3 Methodology
This section describes how fuzzy methodology was implemented in the implementation
study. Table 8 provides an overview of the criteria used and describes the pre-processing of the
data that was necessary before the fuzzy membership layers could be created. Following the
table, several sub-sections describe the creation of each fuzzy membership layer.

52

Table 8 Overview of Fuzzy Methodology
Category Criteria
ID
Criteria Source Layer Pre-processing
1. Avoid travel path of a
contaminant from a
source
1a Should not be in the flow
path of human waste
Latrines Path Distance with maximum distance
set to 1000 (cost), horizontal factor to
forward and vertical factor to linear.
Output grid snapped to 90m DEM
grid
2. Avoid areas prone to
flooding
2a Should not be within a
wetland
Africa Topographic
Position, WRI Wetlands
Obtained as 100m grid classified into
two classes only. Wetlands masking.
Required resampling after fuzzy
membership assigned to register with
other 90m grids.
3. Avoid areas with an
increased likelihood of a
person contracting
malaria
3a Should not within an area
with stagnant water
Africa Topographic
Position
Same as above
4. Permit areas where
drilling is feasible
4a Should be on a slope of less
than 16 percent
DEM,
SRTM3N01E032V1
Percent slope was derived from the 90
m DEM.
4b Should be within 4000 m
of a major road
Major Roads Polyline to Raster, Euclidean
Distance
6. Promote health of
children
6a Should be within 3000 m
of a school
Schools Euclidean Distance
7. Should maximize
clean water coverage
7a Should be within 1500 m
of a populated place
Villages, Geonames
data, Landsat 8 imagery
Euclidean Distance
7b Should not be within 1500
m of a water well
Existing Wells Euclidean Distance

In the following sub-sections, the creation of each fuzzy membership layer used to
evaluate each criterion is described. Following the description for each criterion, there is a
summary of how fuzzy membership was assigned in table form. Each of these tables is organized
into the following rows:
• Values: the process by which the fuzzy membership values are derived from the pre-
processed layer
• Source of Uncertainty: the elements of uncertainty reflected in the fuzzy measure used
• Definitely Suitable: the threshold value for areas that are most suitable
• Definitely Unsuitable: the threshold value for areas that are least suitable
• Suitability Range/Variation in Range: the ranges of values across which fuzzy
membership varies from 0 to 1 and an explanation of the nature of that variation
• Membership Function: the logic for the selection of the fuzzy type and description of the
specific algorithm used in the fuzzification of the input raster
4.3.1 Surface Flow of a Contaminant
A multi-step process was created to model the surface flow/ near surface flow of a
contaminant. In this implementation study, only human waste from a latrine point was available
for this criterion. To identify the flow path of human waste, the Path Distance tool was used.
This tool calculates for each cell, the least accumulative cost distance from a source location to
another. This tool is similar to Cost Distance, but it has added complexity because it is able to
account for surface distance and horizontal and vertical cost factors.
Several steps were required to use the Path Distance tool. First, a constant raster to be
used as the cost raster was created with all cell values set to one since the cost effect of slope is
accounted for by other parameters. Next an aspect layer was created using the 90 m DEM. This
is used to indicate the direction of downslope flow. Then, in the Path Distance tool, the
horizontal factor parameter was set to forward. This meant that only forward (downslope)
movement was allowed. The vertical parameter was set to linear so that the vertical factor of
travel cost was dependent upon slope angle between the FROM cell and the TO cell. The

54

maximum cost distance was set to 1000. Finally, the cell size and extent were set to the 90 m
void-filled DEM.
The output raster is the travel cost of human waste based on flow path. The values range
from 0 to 1000. Cells not in the calculated flow path were given a value of NoData by the Path
Distance tool. Since these are suitable locations, they were subsequently assigned a value of
1000, the limit of the calculated cost distance.
To confirm the results of the Path Distance analysis are valid for this purpose, Figure 10
shows a close up view of the fuzzy membership layer overlaid on an elevation layer. In the figure
there is no distance data (i.e. it is transparent) beyond the 1000 path distance. Note that the flow
from this latrine bends around the hill on the east where the elevation is lower than the latrine. It
also shows some depressions near the latrine sites where the flow is significantly restrained.
While it appears the Path Distance tool is successful at representing the flow, this analysis
highlights one of the challenges of using elevation data at a coarse resolution since the 90 m
scale does not capture small scale elevation changes.

55

Figure 10 Flow Path Example Map
To ensure the maximum cost distance used is appropriate as a limit, it is recommended a
user manually check the horizontal distance to this limit using Measure in ArcMap. In this study,
the 1000 cost distances were sufficiently further than the distances suggested by the Illinois DOH
and the EPA as described in Chapter Two. The shortest actual horizontal distance at which a cell
with a value of 1000 was from a latrine was 686 m. Thus it was concluded that the 1000 cost
distance limit is far enough to ensure that a well beyond this distance would not be located in the
surface or near surface flow path from a contamination source. The analysis process, including
the fuzzy membership process described in Table 9 below, is outlined in Figure 11.

56

Figure 11 Overview of Contaminant Flow Analysis

57

Table 9 Contamination Source Criterion Fuzzy Assignment
Criteria 1a - Should not be in the flow path of human waste
Source Layer Latrines
Values in source
layer
The cell values are the cost distance a contaminant will travel from a
latrine. The value of 1000 is assumed to represent the downslope
distance beyond which a contaminant could travel and pollute a water
well.
Source of
Uncertainty
Given the coarse resolution of the DEM, it is uncertain that the tool
accurately captures the specific path a contaminant will flow. It is
also uncertain how far it will flow based on the amount of water it
travels with and how much contaminant is moving. Therefore, the
specific travel distances of the contaminant are uncertain.
Definitely Suitable Areas greater than the cost distance of 1000 are definitely suitable.
Definitely
Unsuitable
Areas less than 300 cost distance of a contaminant source are
definitely unsuitable.
Suitability Range/
Variation in range
The acceptable suitability range is 300 to 1000. As described above,
areas not within the direct flow path of the contaminant were
assigned a value of 1000. Therefore, cell values range from 0 to
1000. Within the travel path the values are continuous from 0 to
1000.
Membership
Function
The linear function was used for this criterion because it provided the
ability to set a range of acceptable values. In this case, the range was
between 300 and 1000. This gives full membership (1) to 1000 and
higher and no membership (0) to 300 and below. The linear
algorithm decreases membership from 1000 to 300.

It is useful to take some time here to explain the process of assignment of fuzzy
membership values using the linear function that is used for several of the criteria. The fuzzy
linear function applies a linear function between the specified minimum and maximum
suitability values. Any value below the minimum is assigned a 0 indicating it is definitely not a
member of the suitability set. Any value above the maximum is assigned a 1 indicating it is
definitely a member of the suitability set. For example, in Figure 12, the blue line represents a
positive sloped linear transformation with the minimum set to 30 and the maximum set to 80.
Any value smaller than 30 was assigned a 0 and any value above 80 was assigned a 1. Values

58

between 30 and 80 are in the transition zone and are given membership based on their value
related to what is designated as suitable. The linear function uses a straight line for its
membership slope. Then when the fuzzy membership tool is used, it assigns a membership value
(between 0-1) based on the strength of membership in the suitability set. In Figure 12 there are
two lines, the red line represents a negative slope and is used when a higher value is suitable (e.g.
when a farness to a contaminant is preferred). The blue line represents a positive slope and is
used for when a low value is suitable (e.g. when nearness to a school is preferred).

Figure 12 Fuzzy Linear Function
Source: Esri 2014

In the contaminant surface flow criterion above, the further the distance, the higher the
value and the more suitable a location is. Therefore, the positive slope is used because the
maximum input is 1000 and the minimum is 300. As explained later, in the criterion analysis of
schools, higher membership is given to locations near a school (low distance value). Therefore a
slope is negative because the maximum value (given membership of 1) is set to 0.

59

Figure 13 shows the membership layer overlaid on satellite imagery. The membership
layer was clipped to the study area.

Figure 13 Criteria 1a Membership Layer Overlaid on Satellite Image

4.3.2 Wetlands and Depressions
The wetlands and depressions themes are discussed together because they are analyzed
using the same source, the ATP layer. As noted above, the ATP layer has one value to designate
areas that are either wetlands or depressions. In order to assign different fuzzy membership to
these two different concepts, it was necessary to create two layers: (1) wetlands which was used
to analyze the criterion about areas that are more likely to flood, and (2) depressions, which was

60

used to analyze the areas where water is more likely to be stagnant. The process to separate the
two layers is described below.
The ATP layer was used as a reference for the wetlands analysis and as a source for the
depressions layer because it has a broad extent and was created by USGS, a credible source. The
author tested several other data sources and methods but encountered problems of data
availability, insufficient metadata, and the methods used to derive the data were not explained.
Information about how the ATP layer was created is available through the USGS
http://rmgsc.cr.usgs.gov/ecosystems/africa.shtml). Figure 14 shows the ATP data in the study
area.

Figure 14 ATP Class Names

61

Separating the Wetlands
In order to separate lowland areas from depressions, each had to be distinguished from
one another. To do this, an additional dataset was required. A wetlands layer for the region was
acquired through the World Resources Institute (WRI). This layer consists of wetlands polygons
derived from satellite imagery. When the framework is applied to a different context, it may be
possible to find a similar wetlands layer for the region of interest because wetlands are areas of
particular ecological concern in many countries.
Since the metadata associated with the WRI wetlands layer is limited, there was some
question as to its accuracy. Therefore, the original ATP layer and the WRI wetlands polygons
layer were compared. Since the ATP lowlands/depressions layer was created using the CTI
threshold, the lowlands should be areas that have a steady state of wetness and therefore should
be similar to the wetlands polygon from the WRI. However, these are not identified as wetlands
but are classified as lowland/ depressions. The WRI wetlands layer is overlaid on the ATP layer
in Figure 15. The overlay reveals the two layers have a clear overlap of wetlands and lowlands,
though the coverage is not identical. This step validated the WRI wetlands polygons.

62

Figure 15 WRI Wetlands Polygon overlaid on ATP Layer
The Raster to Polygon tool was used on the WRI wetlands polygons layer. The cell size
was set to 90 m. Once the tool was used, the wetlands areas were assigned a value of 0
(unsuitable) and the non-wetlands areas were assigned a value of 1 (suitable).
While this study used WRI wetlands to analyze areas where flooding might occur, there
are other options that could have been used. One alternative is to use the WRI wetlands mask on
the ATP layer to extract the wetlands for fuzzification. However, because the ATP layer does not
actually classify wetlands, it was decided to use the WRI polygons as the source layer.
Therefore, where the ATP layer and WRI wetlands do not align, the data in the WRI wetlands is
used.

63

Separating the Depressions
Using the study area polygon layer, the Erase tool was used to erase wetlands polygons.
What was left were study area polygons that were not within wetlands polygons. This created a
polygon that was used to mask the original ATP layer. Then Extract by Mask was used to change
the unmasked wetlands areas to NoData. The result is a raster with NoData values where there
are wetlands. Therefore, the only cells left with the original class name of wetlands/depressions
(value = 1) are now only depressions.
The areas of depressions were assigned a fuzzy membership value of 0.4, rather than the
0 fuzzy value that is assigned to the wetlands. This increases the membership of the depressions
area in the suitability set because of the uncertainty about their exact location or severity of the
risk. However because the masked ATP layer was stored as integer data, it was necessary to
change the raster to a floating point type which would allow cell values to be either 1 or 0.4.
Thus the cells were reclassified initially to 10 or 4. Then Divide was used to divide each cell by
10 and thus a raster with values of 1 or 0.4 was created. The process used to create the
depressions layer is in Figure 16.

64

Figure 16 Overview of Process to Create Depressions Layer

Once the depressions layer was created, it was necessary to resample the original 100 m
grid to a 90 m. The method chosen to Resample the cells was bilinear interpolation. This method
uses the value of the four nearest input cell centers to determine the value on the output raster.
The new value for the output cell is a weighted average of these four values and is adjusted to
account for the distance from the four input cells. The smoothed values now ranging between 0.4
and 1 reflect the uncertainty of the location of the depressions within the grid cells given the
resampling.
Table 10 summarizes the analysis of the wetlands criterion and the fuzzy assignments.
Then Figure 17 shows the fuzzy membership values for the wetlands layer. Next, Table 11
shows the analysis of the depressions criteria and fuzzy assignments and Figure 18 shows the
fuzzy membership values for the depressions layer.

65

Table 10 Wetland Criterion Analysis
Criteria 2a - Should not be within a wetland
Source Layer WRI, Africa Topographic Position
Values in source layer The layer is divided into two classes: 0 is wetland and 1 is upland.
Source of Uncertainty There is uncertainty about how accurately the layer represents the
phenomena. The source layer was a polygon was converted to a
polygon and therefore the boundary is modified through the
conversion process.
Definitely Suitable Uplands areas are definitely suitable
Definitely Unsuitable Wetlands are definitely unsuitable
Suitability Range/
Variation in range
The suitability range is 1.
Membership Function No membership function was used. The source layer was used
directly by the fuzzy overlay tool since values already ranged
between 0 and 1.

Figure 17 Map of Membership for Criterion 2a

66

Table 11 Depressions Criterion Analysis
Criteria 3a - Should not be within an area with stagnant water
Source Layer Africa Topographic Position
Values in source
layer
The values in the source layer are 1 (depressions/lowlands) or 2
(uplands)
Source
of Uncertainty
There is uncertainty about how accurately the layer represents the
phenomena. Additionally, the source layer was at 100 m grid and was
resampled to 90 m. This caused the boundary of a depression to be
more uncertain.
Definitely Suitable Areas of no depression are definitely suitable.
Definitely
Unsuitable
Depressions are definitely unsuitable, but the uncertainty of their
location warrants better than 0, therefore .4 is assigned.
Suitability Range/
Variation in range
The suitability range is .4 to 1.
Membership
Function
No membership function was used. The source layer was used
directly by the fuzzy overlay tool since values already ranged
between .4 and 1.

Figure 18 Map of Membership for Criterion 3a

67

4.3.4 Slope
For the slope analysis, a void-filled DEM was used to create a slope layer which was then
used to create the fuzzy layer. The Slope tool was used to derive slope percent values for
membership evaluation. Table 12 shows an overview of the criterion analysis. Figure 19 shows
the fuzzy membership values for the slope layer.
Table 12 Slope Criterion Analysis
Criteria 4a - Should be on a slope of less than 16 percent
Source Layer Void-Filled DEM
Values in source layer The values are percent slope. Within the study area the values
range from 0 to 22.
Source of Uncertainty The slope change within a 90 m cell was uncertain. Therefore
membership was given to cells that likely have less than 16 percent
slope throughout a cell.
Definitely Suitable Locations with 0 slope are definitely suitable.
Definitely Unsuitable Areas with 16 percent slope or higher are definitely unsuitable.
Suitability Range/
Variation in range
The acceptable range is 0 to just under 16. The majority of the
study area has less than 5 percent slope. There are a few areas
spread throughout the study area with approximately 20 percent
slope.
Membership Function The linear function was used. The unsuitable locations with a value
of 4000 were assigned a membership value of 0. Then a
membership rating of 1 was assigned to the value of 0.

Figure 19 Map of Membership of Criterion 4a

68

4.3.5 Major Roads and Schools
A roads layer was downloaded from OSM. The locations of schools and existing wells
were acquired through World Vision. The layer used for membership evaluation was created
using Euclidean Distance within the study area.
Table 13 shows an overview of the criterion analysis. Then Figure 20 shows the fuzzy
membership values for the major roads layer. Table 14 shows an overview of the criterion
analysis for schools and Figure 21 shows the fuzzy membership values for the schools layer.
Table 13 Analysis of Major Road Criterion
Criteria 4b - Should be within 4000 m of a major road
Source Layer Euclidean Distance from Major Roads
Values in source
layer
The values are the distance of a location from a major road in meters.
Within the study area the values range from 0 to 8596.
Source of
Uncertainty
The road was converted to raster to use the Euclidean Distance tool.
Each cell was given a value based on whether or not a road was
contained within a cell. This caused the road to be visualized as 90m
wide. Therefore, the boundary and distance of a location from the
road is uncertain.
Definitely Suitable Locations near a major road are definitely suitable.
Definitely
Unsuitable
Areas further than 4000 m are definitely unsuitable. These areas are
less accessible and increase costs.
Suitability Range/
Variation in range
The acceptable range is 0 to 4000 m. There are only 4 major roads in
the study area. Most of the study area is further than 4000 m from a
major road. The closer an area is to a road, the better, and therefore
the value of 0 is assigned a membership rating of 1.
Membership
Function
The linear function was used. The unsuitable locations with a value
of 4000 were assigned a membership value of 0. Then a membership
rating of 1 was assigned to the value of 0.

69

Figure 20 Map of Membership of Criterion 4b

Table 14 Analysis of School Criterion
Criteria 6a - Should be within 3000 m of a school
Source Layer Euclidean distance from Schools
Values in source
layer
The values are the distance, in meters, a location is from a school and
range from 0 to 9,172.
Source of
Uncertainty
The values are continuous and the breakpoint of a suitable and
unsuitable distance of a well in relation to a school is not well
defined.
Definitely Suitable Locations near a school are definitely suitable.
Definitely
Unsuitable
Areas further than 3000 m away from a school are definitely
unsuitable.
Suitability Range/
Variation in range
Therefore the acceptable range is 0 to 3000 m.
Membership
Function
The linear function was used. The unsuitable locations with a value
of 3000 were assigned a membership rating of 0. A membership
rating of 1 was assigned to the 0 value.

70

Figure 21 Map of Membership for Criterion 6a

4.3.6 Populated Places
The populated places data layer was enhanced by combining several sources. The
primary data of villages was from World Vision. On inspection of this data, it was evident that
some places were missing. As a result, GeoNames was used to add populated places that were
missing. The following process was developed to identify duplicates. First, the attribute tables
were imported into Excel for comparison. Four duplicate names were identified in both data sets.
Then these were visualized in ArcMap to determine if the duplicate names were in the same
general area of each other, which they all were. It was also confirmed that there was not a
consistent positional shift between the two source layers indicating the use of different datums.
The differences are likely because the locations of the points in the GeoNames dataset may have

71

less positional accuracy than the field collected World Vision data since GeoNames is a global
dataset. Therefore, the locations of these four duplicate places in the World Vision data were
used.
A process was developed to determine if any of the points might represent the same place
but have different names. This process was used to ensure no new places from GeoNames were
added to what was already present World Vision data. First, duplicate names that were
previously identified were selected on the map to determine the maximum distance between a
World Vision point and a GeoNames point. This showed the positional shift of locations that are
actually the same. The maximum distance was 2000 m. Therefore, no points were added if they
were located within 2000 m of an existing place in the WV dataset. This is because it is possible
that these are representing the same place at a different positional accuracy. Then, to supplement
this rule, satellite imagery was used to examine where there were two places within 2000 meters
and whether or not these were two distinct populated places. Where two distinct populated places
existed, the point from GeoNames was added. In the implementation study, two populated places
were identified within the 2000 m on satellite imagery, and were confirmed as separate places.
Table 15 shows an overview of the analysis of the populated places criterion and Figure
23 shows the fuzzy membership values for the populated places layer.

72

Table 15 Analysis of Populated Places Criterion
Criteria 7a - Should be within 1500 m of a populated place
Source Layer Villages, supplemented with satellite imagery and GeoNames data
Values in source
layer
Values are the distance in meters of a location from a populated
place.
Source
of Uncertainty
The values are continuous but the suitable break point is well-defined
by the UN. This is the definition of water coverage.
Definitely Suitable Locations near a village are definitely suitable.
Definitely
Unsuitable
Areas further than 1500 m are definitely unsuitable.
Suitability Range/
Variation in range
The suitability range is 0 to 1500 m. The cell values range from 0 to
6000 because the distance extends outside of the study area.
Membership
Function
The membership function used was linear. The maximum was set to
0 and the minimum was set to 1500.

Figure 22 Map of Membership for Criterion 7a

73

Existing Wells
As mentioned above, the existing wells layer was used as a means to verify and refine the
implemented model. Thus, during the development and implementation phase of the framework,
it was withheld as a criterion. In the final suitability map, it was included as an input. For
completeness in this section, Table 16 shows an overview of the analysis of the existing wells
criterion.
Table 16 Analysis of Existing Wells Criterion
Criteria 7b - Should not be within 1500 m of a water well
Source Layer Existing Wells
Values in source
layer
Values are the distance in meters of a location from an existing water
well.
Source
of Uncertainty
The values are continuous and the breakpoint between a suitable
distance for well in relation to an existing well is not well defined.
Definitely Suitable Locations furthest from an existing well are definitely suitable.
Definitely
Unsuitable
Areas within 1500 m from an existing well are definitely unsuitable.
Suitability Range/
Variation in range
The suitable range is 1500 m to the maximum distance in a study
area. The Euclidean Distance layer was clipped to the study area to
identify the furthest distance from a well within the study area. The
cell values range from 0 to 6680 because the distance extends outside
of the study area.
Membership
Function
The membership function used was linear. The maximum was set to
6680 and the minimum was set to 1500.

Figure 23 shows the fuzzy membership values for the existing wells layer.

74

Figure 23 Map of Membership of Criterion 7b

4.4 Fuzzy Suitability Overlay Results
To combine the fuzzy membership layers, the And operator was used, which assigns the
minimum values from all the input fuzzy membership layers to the output cell. This operator
identifies the least common denominator for the membership criteria, producing a more
conservative (or exclusive) result with smaller overall membership values. This allows cells with
membership of a specific minimum value of a criterion to be identified. Figure 24 shows the
suitability map resulting from the fuzzy overlay of all criteria except the existing wells.

75

Figure 24 Overall Suitability Map based on Refined Model
Due to the exclusive nature of the And overlay type, and because it assigns values based
on the lowest common denominator for all criteria, the majority of the study area has been
disqualified. This is good because it removes from consideration a large part of the region and
now the site survey teams can focus on a limited number of suitable areas.
The metric used to define suitable locations in this study is greater than or equal to.5.
The suitability range above the .5 cut-off values shows up in most of the graphics below as
values that are yellow to green.
While the study area is mostly unsuitable, it is not as exclusive at is appears. Many of the
unsuitable locations are in remote areas where there is low need for water well. In more
populated areas, where there is more need, the suitability map shows some range in suitability.

76

Figure 25 shows the distribution of populated places and schools and reveals the suitable
membership values are generally aligned with the distribution of infrastructure.

Figure 25 Membership based on Refined Model and Infrastructure

4.5 Iterative Model Fitting
As mentioned above, the distances and methods used to analyze a criterion were refined
through an iterative process. In this section, the process used to refine the model is described.
The results of the refined model are evaluated in the next section.
The first implementation using initial distance limits yielded poor results. When the
existing wells were used to evaluate the results, there were no existing wells in what the model
identified as suitable locations. Furthermore, the criteria and initial distances yielded a

77

membership scale 0 to .4. This meant that there were no areas above .4 overall membership when
all criteria are combined (see Figure 26).
By adjusting the distance limits and methods for the criteria, the refined implementation
yielded 10 existing wells in locations deemed suitable by the model. This is 13 percent of the
wells (see Figure 27). Additionally, the membership scale improved from .4 to .9. The iterative
modeling fitting process used to improve the model is described below.
.
Figure 26 Preliminary Suitability Map

78

Figure 27 Refined Suitability Map with Existing Wells

In addition to evaluating the combined suitability score, each criterion was evaluated
individually. Table 17 shows the performance of the preliminary and refined models for each
criterion and lists the changes made in the refined model.

79

Table 17 Model Refinements and Improvements
Criteria
# of Wells in suitable
areas
%
Improved
Change in Method
Preliminary
Model
Refined
Model
1a - Should
not be in the
flow path of
human waste
72 72 - No Change
2a - Should
not be within a
wetland
0 69 N/A The first implementation used the
ATP layer for both the wetlands and
stagnant water criteria. The refined
model separated the layer. For the
wetlands the grid was divided into
yes/no. The first implementation
used cost distance to buffer the
distance. This was removed because
it was identifying many existing
wells that were located on the edge
of a wetland.
3a - Should
not be within
an area with
stagnant water
0 71 N/A The refined model has higher
uncertainty of the depressions areas.
This is due to low confidence in the
methods used to derive depressions.
4a - Should be
on a slope of
less than 16
percent
76 76 0% No change
4b - Should be
within 4000 m
of a major
road
32 46 18% The first implementation set the
suitability range to 2000 m. This
was determined to be too exclusive
and the range was increased to 4000
m.
6a - Should be
within 3000 m
of a school
11 31 26% The suitability range was increased
to 3000 m from 1500 m.
7a - Should be
within 1500 m
of a populated
place
22 38 21% The range was not changed because
clean water coverage is defined by
the UN as 1500 m. The populated
places layer was enhanced with
more populated places through
methods described above.

80

In summary, after refinement, the count of existing wells located in areas identified as
suitable (greater than or equal to .5 membership) increased from 0 to 10. Therefore based on the
refined model, 13% of the 76 existing wells are in overall suitable areas.
The refined model clearly shows improvement with respect to the number of existing
wells that fall in suitable locations. Refinement to produce the model described in this chapter
was accomplished by modifying the suitability distances for individual criteria and analyzing the
results. It is suggested that a user perform a similar model fitting refining process when the
model is implemented.
4.6 Evaluating Results of Refined Model
In this section, the results of the refined model implementation are evaluated by assessing
the suitability of the location of existing wells. To evaluate the results, values were extracted
from all fuzzy layers for the 76 existing well point locations using the Extract Multi Values to
Points tool. The resulting attribute table was imported into Excel for evaluation for analysis.
When each of the 76 wells is considered individually on each of the 7 criteria (76x7 = 532 fuzzy
membership estimates), 76 percent of the estimates are suitable (greater than or equal to 0.5
membership). This is an improvement of the preliminary model which resulted in 36 percent of
the estimates being suitable. Table 18 shows the suitability classification used to evaluate the
suitability of existing well locations. The values were summarized and are shown in Table 19. In
the evaluation, areas with membership values of .5 or greater are considered suitable.

81

Table 18 Suitability Classification
Membership
Range
Membership
Index
.75-100
Most
Suitable
.50 - 749 Suitable
.25 - .499
Not
Suggested
0 - .249 Not Suitable

Next, the spatial distribution of existing wells is discussed individually for each criterion.
Then following the figures and discussion, two summary tables are presented (Table 20 and 21).
A complete list of existing wells and their membership values for each criterion is available in
Appendix 1 and 2 for the preliminary model and refined model respectively.
Table 19 Existing Well Distribution in Suitable Zone
Criteria Count of Wells
in Suitable
Locations
Percent in
Suitable
Locations
1a - Should not be in the flow path of human
waste
72 95%
2a - Should not be within a wetland 69 91%
3a - Should not be within an area with
stagnant water
72 95%
4a - Should be on a slope of less than 16
percent
76 100%
4b - Should be within 4000 m of a major road 46 61%
6a - Should be within 3000 m of a school 31 41%
7a - Should be within 1500 m of a populated
place
38 50%
Total Suitability 10 13%

Figure 28 shows the spatial distribution of wells in the suitability zones for the human
waste contaminant flow criterion. There are three wells close together in unsuitable locations in
northern area of the study area. The reason for this grouping is unknown and should be

82

investigated in a field visit if feasible. Additionally there are only a few wells in unsuitable
locations dispersed throughout the study area in no apparent pattern. As the map shows, where
an existing well is red (in an unsuitable location), there is generally a latrine located to the south-
west of that well. This is consistent with the general slope of the area with a subtle decrease in
elevation moving from the southwest to northeast (toward Lake Kyoga). Therefore the results of
this membership criteria analysis are consistent with slope which was used to determine the
direction of contaminant flow.

Figure 28 Existing Wells with Contaminant Flow Membership Value

Figure 29 shows the spatial distribution of wells in the suitability zones for the wetlands
criterion. There are five wells located in unsuitable locations in the northern region of the study

83

area. Further investigation is needed to understand these locations better. Visual analysis of the
elevation data revealed non-suitable locations are located in areas that appear to be in lowlands.

Figure 29 Existing Wells with Wetlands Membership Value

Figure 30 shows the spatial distribution of wells in the suitability zones for the stagnant
water criterion. The membership scale ranges from .4 to 1 as described above. This is
represented in the legend of Figure 30. An analysis of elevation data revealed the existing wells
with low membership are in areas with lower elevations where slope is smaller.

84

Figure 30 Existing Wells with Stagnant Water Membership Value

Figure 31 shows the spatial distribution of wells in the suitability zones for the percent
slope criterion. The lowest membership for the slope criteria is .71. Therefore, there are only two
categories symbolized in the map, suitable and most suitable. Percent slope did not disqualify
any wells from membership in the study area. This is because the area is generally flat. However,
the topography in other areas of sub-Saharan Africa can have a higher frequency of elevation
change. Therefore it is likely to be more restrictive in other contexts.

85

Figure 31 Existing Wells with Slope Membership Value

Figure 32 shows the spatial distribution of wells in the suitability zones for the major
roads criterion. As discussed earlier there are four major roads in the study area. This criterion
helps to identify and then disqualify more remote areas where the need is less severe. Of course,
the extent this is true or valid depends on the number of roads in an area. The map shows the
expected results that membership values depreciate the further a well is from a major road.

86

Figure 32 Existing Wells with Major Roads Membership Value

Figure 33 shows the spatial distribution of wells in the suitability zones for the schools
criterion. The map shows the wells in proximity to schools are given membership into the
suitability set. It is also observed that there is a higher rate of depreciation of membership outside
of the 3000 m suitable area.

87

Figure 33 Existing Wells with the Schools Membership Value

Figure 34 shows the spatial distribution of wells in the suitability zones for the populated
place criterion. The populated places are dispersed throughout the study area.

88

Figure 34 Existing Wells with the Populated Places Membership Value

After the review of all criteria and a cross evaluation with satellite imagery, it was
confirmed the analysis of the criteria and the resulting membership values are congruent with
what is known about the study area. For a numerical summary of the membership values, Table
20 shows the number of existing wells in each suitability zone and Table 21 shows the
percentages. The second column shows the membership index created to assist in the
visualization of the distribution of wells by suitability zones. These color ranges were used in
Figures 28 to 34.

89

Table 20 Distribution of Existing Wells in Membership Zones by Criteria
Membership
Range
Membership
Index
Contaminant
Flow
Wetlands
Stagnant
Water
Slope Schools Roads
Populated
Places
.75-100
Most
Suitable
72 69 66 68 11 32 19
.50 - 749 Suitable 0 0 5 8 20 14 19
.25 - .499
Not
Suggested
1 0 4 0 15 10 14
0 - .249 Not Suitable 3 7 0 0 30 20 24

Table 21 Percent Distribution of Existing Wells in Membership Ranges by Criteria
Membership
Range
Membership
Index
Contaminant
Flow
Wetlands
Stagnant
Water
Slope Schools Roads
Populated
Places
.75-100
Most
Suitable
95% 91% 87% 89% 14% 42% 25%
.50 - 749 Suitable 0% 0% 7% 11% 26% 18% 25%
.25 - .499
Not
Suggested
1% 0% 5% 0% 20% 13% 18%
0 - .249 Not Suitable 4% 9% 0% 0% 39% 26% 32%

As Table 20 shows, there is a higher count of existing wells in the suitable zones for the
of wetlands, stagnant water and slope criteria, while there is a lower count for the schools,
populated places and roads criteria. The lower count is likely due to the restrictive distance
metrics used to evaluate the criteria. The distribution of existing wells in membership ranges for
the overlay suitability is in Appendix 3.
In this chapter, the general framework was implemented as a model in Uganda. The
methodology and data used to evaluate criteria was described. Then the results of the refined
model were discussed. Next, the iterative model fitting process was described and the resulting
modifications to the model were presented. Even with the improvements, there is future work
that can be done to improve the model. These opportunities are discussed in the next chapter.

90

CHAPTER 5: SUMMARY AND CONCLUSION
This study developed a general framework to identify potential water well sites where
risks are minimized and benefits are maximized in rural sub-Saharan Africa. In summary, the
research had a threefold purpose of: building upon current research on clean drinking water in
sub-Saharan Africa and fuzzy methods for suitability analysis; creating a suitability framework
that could be implemented by NGOs and governments throughout sub-Saharan Africa; and
developing a refined model that applies the framework in a specific context in sub-Saharan
Africa. The necessary research included reviewing literature on risks to clean water, developing
a means to evaluate criteria, and developing methods to analyze the criteria in a GIS.
5.1 Assessment of model success
Fuzzy logic was a useful method for this application because it prevented a false
appearance of confidence in the data. Fuzzy logic allowed for low confidence to be captured in
the method. For example, due to low confidence of how the stagnant water data was derived,
higher membership was assigned. Fuzzy methods were used to represent the fuzziness that was
inherent in the data. This was especially useful for data layers that needed to be resampled to
ensure co-registration.
In working to develop a suitability framework, a number of requirements were involved.
The first focus of the framework was to approach suitability in a manner that minimized risk.
This was done by identifying several potential sources of contamination in sub-Saharan Africa.
Research into risks to clean water in sub-Saharan Africa yielded seven general risks to drinking
water in the region. Several of these can be analyzed simply with widely available data
(elevation and lowlands). Others are more difficult to analyze (such as the flow of a

91

contaminant). Despite data limitations, in particular no data about groundwater, a method was
created to predict a probable path of a surface or near-surface contaminant. In the
implementation of the model, it was shown how globally available 90 m DEM can be used to
model the likely flow path of a surface or near surface contaminant based on slope.
The second focus of the framework was to maximize the benefits by identifying areas
where access to clean water would most benefit people living in an area of interest. To support
this, schools, health facilities, and populated places were identified as places near which clean
water would be a particular benefit. Using the standard definition of clean water access, an ideal
location for a clean water well will, thus, be within 1.5 km of a populated place.
The application of this data to the specific context of this remote area in Uganda revealed
some areas of success and some areas that needed adjustments. The improvements made to the
poorly performing preliminary model through a process of model fitting and evaluation using
locations of existing wells in the study area resulted in a better model. The preliminary model
used distances and methods that were revealed to be too exclusive to yield helpful results. The
criteria used to analyze the probable flow path of a surface/near surface contaminant yielded
good results and did not need to be modified. The buffer initially used around the wetlands was
removed because a significant number of existing wells are located near the edge of a wetland.
The membership scale for the stagnant water criteria was modified due to uncertain methods
used to derive the layer. There was no change to the slope layer. The suitability buffer was
extended for the major roads, and schools. There was no change to the populated places criteria
to ensure the model supports the 1.5 km definition of access to clean water.

92

The refined model more capably identifies suitable locations of existing wells. The model
fitting process led to the overall suitability membership scale increasing from .4 to .9 (using the
And operator). It also led to 10 existing wells being in locations deemed suitable by the model,
this was up from 0. Additionally, it led to 76 percent of the criteria estimates being suitable,
which is an increase from 36 percent. This 76 percent is helpful to evaluate the distance and
method to evaluate a criterion. When all of the layers are combined to one suitability map, the
percent goes down substantially from 76 percent average per criteria to 13 percent. This shows
that while the model operates at an acceptable rate (76 percent) on an individual criterion basis,
adding more criteria makes the model more exclusive.
An overall suitability analysis using all eight criteria, including proximity to existing
wells, creates an exclusive result. The membership scale ranges from 0 to 0.3 (see Figure 35),
with no locations achieving the 0.5 suitability cut-off value. The majority of the study area has
very low membership.

93

Figure 35 Suitability Map with Existing Wells Criteria

This demonstrates the value of including an evaluation mechanism in the model
development process. In this study, the locations of existing wells were used to evaluate the
model, but a better option for future work is to field test the results.
5.2 Future Work
While the refined model showed better performance, it can be further improved in several
ways. Field research for further investigation into the wells that scored low on the membership
scale is needed. This can provide more information about why the wells were installed in areas
deemed unsuitable by the model. It can also identify how the wells are constructed and protected,
and can confirm if they need to be relocated. Field research can also help identify inappropriate
criteria in the general framework.

94

Once a field check is completed, the results of a model, such as the suitability map
created in this implementation study, can be used as a decision support tool for water well
planning. The model was able to identify areas of high risks based on multiple criteria. As
higher resolution data and more types of data become available, the criteria can be modified and
improved. It is hoped the general framework and the implemented model outlined here can serve
as a foundational tool that NGO practitioners will continue to improve.
The framework was developed remotely and can therefore continue to be improved with
local knowledge. It is likely that as research continues more criteria will be added to the
framework. Additionally, a user in the field may also be able to acquire higher resolution data
that will likely improve the performance of the framework.
5.3 General applicability of this research
Since many of the risks to clean water also exist outside of sub-Saharan Africa, the
framework can be used in other regions. The thirteen criteria were developed by researching
universal risks to clean water and risks specific to sub-Saharan Africa. Using this general
framework, a user is easily able to select only the relevant risks and to add additional ones that
may be appropriate.
The use of the fuzzy membership function is an important part of the framework because
it allows a user to set membership ratings based on whether or not a criterion is about a risk or a
social benefit. The shape of the membership function can be chosen for each criterion based on
knowledge about the uncertainty. Here the linear function was deemed most appropriate. The
framework included suitability factors that are good and bad and therefore the ability for a user

95

to set each membership value uniquely with a positive or negative slope based on input values
was critical. When a user implements the framework, the choice of fuzzy membership function
will be a key component of the framework.
When the framework is implemented it can improve the workflow used by an NGO or
government to determine where a well should be located. The benefits of using the framework
include disqualifying for consideration locations that have higher risks or do not provide high
social benefits. The framework also demonstrates how a GIS based tool can be used to evaluate
several criteria with one suitability map.
The tool does not however capture all the dimensions involved in deciding where a clean
water well should be. As mentioned earlier, there are several cultural, political, and social
elements that factor into the decision. The framework can however, be used to supplement
decision making as a macro level tool.
At the beginning of this project, the goal was for the model to perform with better results.
As expected, data availability was a challenge, but this challenge was mitigated by creating
alternative methods and using different data to analyze risks to clean water. Thus, it is felt that
this general framework can be of use to an NGO despite poorer results than desired in the
implementation study.
The work in this project lays the foundation for future developments. The criteria in the
general framework and the measures used in the model can continue to be refined by users as
they customize the model for their context. While the model performed at an adequate level, the
work generally offers some suggestions for how such a GIS-based conceptual framework can be

96

used by NGOs and governments to improve water well planning to provide better coverage of
safe drinking water in sub-Saharan Africa.

97

REFERENCES
Adam, S., N. Henninger, and F. Landsberg. 2009. Mapping a Healthier Future: How Spatial
Analysis Can Guide Pro-Poor Water and Sanitation Planning in Uganda. Uganda
Ministry of Health, Uganda Ministry of Water and Environment, Uganda Bureau of
Statistics, International Livestock Research Institute and World Resources Institute.
http://www.wri.org/publication/mapping-healthier-future (last accessed 23 October
2014).
Awuah, E., K. Nyarko, P. Owus, and K. Osei-Bonsu. 2008. Small town water quality.
Desalination 248 (1): 453-459.
Bartram, J. and J. Platt. 2010. How health professionals can leverage health gains from improved
water, sanitation and hygiene practices. Perspectives in Public Health 130(5): 215-221.
BBC. 2006. Somalis clash over scarce water. BBC News. February 17, 2006.
http://news.bbc.co.uk/go/pr/fr/-/1/hi/world/africa/4723008.stm (last accessed 10 October
2014).
California State Resources Control Board. 2011. A Guide for Private Domestic Well Owners.
http://www.waterboards.ca.gov/gama/docs/wellowner_guide.pdf (last accessed 14
August 2014).
Chalkias, C., M. Ferentinou, C. Polykretis. 2014. GIS Supported Landslide Susceptibility
Modeling at Regional Scale: An Expert-Based Fuzzy Weighting Method. International
Journal of Geo-Information 3(2): 523-539.
Danert, K. 2014. Drilling Deep in Sub-Saharan Africa. Water World 26(1): no page.

98

http://www.waterworld.com/articles/wwi/print/volume-26/issue-1/regional-spotlight-
middle-east-africa/drilling-deep-in-sub-saharan-africa.html (last accessed 27 August
2014).
Dufour, A., J. Bartram, R. Bos and V. Gannon. 2012. Animal Waste, Water Quality and
Human Health. Emerging Issues in Water and Infectious Disease Series, World Health
Organization.
Demesouka, O., A. Vavatsikos and K. Anagnostopoulos. 2014. GIS-based multicriteria
municipal solid waste landfill suitability analysis: A review of the methodologies
performed and criteria implemented. Waste Management & Research 32(4): 270-296.
Egeru, A., and M. Majaliwa. 2009. Landuse/Cover Change Trend in Soroti District Eastern
Uganda. Journal of Applied Sciences and Environmental Management 13(4): 79.
EPA (Environmental Protection Agency). No date. Getting Up to Speed: Ground Water
Contamination. http://www.epa.gov/region1/students/pdfs/gwc1.pdf (last accessed 26
July 2014).
Esri. 2014. How Fuzzy Membership works. http://resources.arcgis.com/en/help/main/10.1
/index.html#//009z000000rz000000 (last accessed 21 October 2014).
Faeth, P. and E. Weinthal. 2012. How Access to Clean Water Prevents Conflict. Solutions 3(1):
no page. http://www.thesolutionsjournal.com/node/1037 (last accessed 10 October
2014).
Farmer, S. No date. Ventilated Improved Pit Latrine. Schumacher Centre for Technology &
Development. https://www.engineeringforchange.org/static/content/Sanitation
/S00040/vip_latrines.pdf (last accessed 30 July 2014).

99

Hillie, M. and M. Hlophe. 2007. Nanotechnology and the challenge of clean water. Nature
Nanotechnology 2(11): 663-664.
Jakhrani, A., S. Samo, and I. Nizamani. 2009. Impact of Wastewater Effluents on Physico-
Chemical Properties of Groundwater. Sindh Univ. Res. Jour. Sci. Ser. 41(1): 75-82.
Kelly, M. 2013. The MDG Enterprise: Experiences and Thoughts from Zambia. IDS Bulletin
44(5-6): 34–41.
Kihoro J., N. Bosco and H. Murage. 2013. Suitability analysis for rice growing sites using a
multicriteria evaluation and GIS approach in great Mwea region, Kenya. SpringerPlus
2:265. http://www.springerplus.com/content/2/1/265 (last accessed 24 July 2014).
Lewis, L. No Date. Rural and Urban Water Issues in Africa. The Water Project.
http://thewaterproject.org/pdf/rural-and-urban-water-issues-africa.pdf (last accessed 24
July 2014).
MacDonald, A, J. Davies, and B. Dochartaigh. 2002. Simple methods for assessing groundwater
resources in low permeability areas of Africa. British Geological Survey, Department for
International Development. http://www-esd.worldbank.org/esd/ard/groundwater/
pdfreports/Simple_%20methods_African_lowperm_areas_Pt1.pdf (last accessed 23
October 2014).
Mitchell, A. 2012. The Esri Guide to GIS Analysis, Volume 3: Modeling Suitability, Movement,
and Interaction. Redlands, CA: Esri Press.
Muruli, J. 2009. Higher Local Government Statistical Abstract: Nakasongola District. Uganda
Bureau of Statistics. http://www.ubos.org/onlinefiles/uploads/ubos /2009_HLG_%20
Abstract_printed/Nakasongola-Final%208-02-10.pdf (last accessed 27 August 2014).

100

Nobre, R., O. Rotunno Filho, W. Mansur, M. Nobre, and C. Cosenza. 2007. Groundwater
vulnerability and risk mapping using GIS, modeling and a fuzzy logic tool. Journal of
Contaminant Hydrology 94(7): 277-292.
Olsen, J., and L. Berry. 2004. Uganda - Extent and impact of Land degradation. Food and
Agriculture Organization of the United Nations.
http://www.fao.org/nr/lada/index.php?option=com_docman&task=doc_download&gid=4
81&Itemid=165&lang=en (last accessed 24 July 2014).
Ongley, E. 1996. Control of water pollution from agriculture. Rome: Food and Agriculture
Organization of the United Nations. http://www.fao.org/docrep/w2598e/w2598e00.htm
(last accessed 23 October 2014).
Qui, F., B. Chastain, Y. Zou, C. Zhang, and H. Sridharan. 2014. Modeling land
suitability/capability using fuzzy evaluation. GeoJournal 79: 167–182.
http://www.utdallas.edu/~ffqiu/published/2014QiuetalGeoJournal.pdf (last accessed 14
August 2014).
Raghuram, S. 2008. The MDGs in a World of Multiplying Inequalities and Differentiating
Complexities. Development 51 (2): 241-244.
Reuter H., A. Nelson, and A. Jarvis. 2007. An evaluation of void filling interpolation methods
for SRTM data. International Journal of Geographic Information Science 21(9): 983-
1008.
Rural Water Supply Network. 2010. Siting of Drilled Water Wells. A Guide for Project
Managers. http://www.sswm.info/sites/default/files/reference_attachments

101

/CARTER%20et%20al%202010%20Siting%20of%20Drilled%20Water%20Wells.pdf
(last accessed 26 July 2014).
Tsiko, R., and T. Haile. 2011. Integrating Geographical Information Systems, Fuzzy Logic and
Analytical Hierarchy Process in Modelling Optimum Sites for Locating Water
Reservoirs. A Case Study of the Debub District in Eritrea. Water 3(1): 254-290.
Üçisik, A., and P. Rushbrook. 1998. The Impacts of Cemeteries on the Environment and Public
Health. WHO Regional Office for Europe. http://whqlibdoc.who.int/euro/1998-
99/eur_icp_ehna_01_04_01(a).pdf (last accessed 12 August 2014).
Uganda Ministry of Health. 2014. Malaria Prevention in Uganda.
http://health.go.ug/mcp/mpre.html (last accessed 24 July 2014).
UNICEF (United Nations Children’s Fund). 1999. Towards Better Programming: A Water
Handbook. United Nations Children’s Fund Programme Division.
http://www.sswm.info/sites/default/files/reference_attachments/UNICEF%201999%20W
ater%20Handbook.pdf (last accessed 15 July 2014).
UNICEF and World Health Organization. 2012. Progress on Drinking Water and Sanitation.
2012 Update. WHO/UNICEF Joint Monitoring Programme for Water Supply and
Sanitation. http://whqlibdoc.who.int/publications/2012/9789280646320_eng_full_
text.pdf?ua=1 (last accessed 16 June 2014).
United Nations. 2000. United Nations Millennium Declaration. General Assembly Agenda item
60 (b) 00 55951. Resolution adopted by the General Assembly.
http://www.un.org/millennium/declaration/ares552e.pdf (last accessed 23 October 2014).

102

--------. 2012. The Millennium Development Goals Report 2012.
http://www.un.org/millenniumgoals/pdf/MDG%20Report%202012.pdf (last access 24
July 2014).
--------. 2013. The Millennium Development Goals Report 2013.
http://www.un.org/millenniumgoals/pdf/report-2013/mdg-report-2013-english.pdf (last
accessed 24 July 2014).
UNDP (United Nations Development Programme). 2004. Water Resources Management,
Environmental Degradation and Human Welfare: Lessons Learned from the Lake Kyoga
Catchment in Uganda. Policy Brief. UNDP-UNEP Poverty-Environment Initiative.
http://www.unpei.org/sites/default/files/e_library_documents/Uganda-Policy-brief-
WaterResourceManagement.pdf (last accessed 14 August 2014).
University of Illinois at Urbana – Champaign Water Resources Center. 1990. Planning Your
Well: Guidelines for Safe Dependable Drinking Water.
http://web.aces.uiuc.edu/vista/pdf_pubs/PLANWELL.PDF (last accessed 7 October
2014).
USGS (United States Geological Survey). 2008. Shuttle Radar Topography Mission.
http://srtm.usgs.gov/mission.php (last accessed 24 July 2014).
--------. 2014. Elevation Derivatives for National Applications (EDNA): Compound Topographic
Index (CTI). http://edna.usgs.gov/edna/datalayers/cti.asp (last accessed 24 July 2014).
Van der Wal, A., 2010. Understanding Groundwater & Wells in manual drilling: Instruction
handbook for manual drilling teams on hydro-geology for well drilling, well installation

103

and well development. Papendrecht, The Netherlands: PRACTICA Foundation.
http://www.unicef.org/wash/files/04.pdf (last accessed 23 October 2014).
WaterAid. 2008. Technology Notes. WaterAid. http://www.sswm.info/library/3546 (last accessed
4 July 2014).
WHO (World Health Organization). 2014. Water Sanitation Health. Water-related diseases:
information sheets. http://www.who.int/water_sanitation_health/diseases/diseasefact/en/
(last accessed 23 October 2014).
World Weather Online. 2014. Soroti Monthly Climate Average, Uganda.
http://www.worldweatheronline.com/Soroti-weather-averages/Soroti/UG.aspx (last
accessed 23 October 2014).
104

104

APPENDIX 1 PRELIMINARY MODEL EXISTING WELL SUITABILITY VALUES
This data in the table below is of six criteria because in the preliminary model, the
wetlands and stagnant water criteria were only treated as one criterion.
ID Water Point Name Wetlands Slope Roads Schools
Populated
Places
Contaminant
Flow
1 Kazwama 0.1 0.9 0.9 0.4 0.6 1.0
2 Kamiga 0.0 0.8 0.7 0.0 0.8 1.0
3 Iryema 0.1 0.8 1.0 0.4 0.4 1.0
4 Kabandi 0.3 0.9 0.0 0.0 0.0 1.0
5 Kitaleba 0.0 0.9 0.8 0.0 0.0 1.0
6 Kalongo TC 0.0 0.8 0.8 0.0 0.3 1.0
7 Kiwambya 0.0 0.9 0.0 0.0 0.0 1.0
8 Nakinyama 0.1 0.9 0.0 0.0 0.0 1.0
9 Kinamwanga 0.0 0.8 0.3 0.0 0.0 1.0
10 Nakataka 0.1 0.9 0.0 0.3 0.4 0.2
11 Bagambira 0.1 0.9 0.9 0.0 0.8 1.0
12 Lutengo 0.0 0.9 0.0 0.0 0.5 1.0
13 Wanzogi 0.2 0.8 1.0 0.0 0.0 1.0
14 Rukooge 0.0 1.0 0.0 0.0 0.0 1.0
15 Kaisolo 0.2 0.9 0.0 0.0 0.0 1.0
16 Kiranga 0.0 0.9 0.0 0.0 0.6 1.0
17 Kigejjo 0.1 0.9 0.0 0.0 0.0 1.0
18 Kalalu 0.2 0.9 0.0 0.0 0.0 1.0
19 Kamirampango 0.0 0.8 0.0 0.0 0.0 1.0
20 Nalubobya 0.1 0.7 0.0 0.0 0.0 1.0
21 Kiwoole 0.1 1.0 0.1 0.0 0.0 1.0
22 kabazi 0.0 0.8 0.0 0.0 0.0 1.0
23 Kyalusaka Primary
School
0.1 1.0 1.0 0.0 0.2 1.0
24 Kyalusaka 0.0 1.0 0.6 0.0 0.3 1.0
25 Irima 0.0 0.9 0.6 0.0 0.2 1.0
26 Kigazi 0.0 0.9 0.8 0.0 0.9 1.0
27 Bagambira
Kyakabona
0.1 0.9 0.6 0.0 0.2 1.0
28 Dagala 0.0 0.9 0.9 0.4 0.0 1.0
29 Madaali 0.0 0.9 0.4 0.6 0.0 1.0
30 Kapundo 0.1 0.9 0.2 0.2 0.0 1.0
31 Ninga 0.0 0.9 0.5 0.6 0.0 1.0
105

105

ID Water Point Name Wetlands Slope Roads Schools
Populated
Places
Contaminant
Flow
32 Kisenyi 0.0 0.9 0.4 0.2 0.0 1.0
33 kakoola 0.0 1.0 0.5 0.2 0.0 1.0
34 Nakinyama 0.0 0.8 0.6 0.0 0.0 1.0
35 Kigali 0.0 1.0 0.8 0.2 0.9 1.0
36 Ruunyu 0.0 0.7 0.8 0.7 0.4 1.0
37 Kyakabombo 0.2 0.8 0.0 0.0 0.0 1.0
38 Kisweramainda 0.0 0.9 0.9 0.6 0.1 1.0
39 Kaleire 0.1 0.8 0.6 0.0 0.0 1.0
40 Kigejjo 0.0 0.9 0.0 0.0 0.0 1.0
41 Nalukonge 0.2 0.9 0.0 0.0 0.0 1.0
42 Kinamwanga 0.0 0.8 0.3 0.0 0.0 1.0
43 Namungolo 0.0 0.9 0.0 0.0 0.3 1.0
44 Nakatuba 0.0 0.9 0.0 0.0 0.0 1.0
45 Kasozi 0.4 0.8 0.4 0.2 0.6 1.0
46 Iryema 0.1 0.9 1.0 0.7 0.4 1.0
47 Kabaandi 0.1 1.0 0.0 0.0 0.0 1.0
48 Kanyonyi 0.1 0.9 0.0 0.6 0.0 1.0
49 Rukooge 0.0 0.8 0.0 0.0 0.0 1.0
50 Kisenyi TC 0.0 0.9 0.9 0.0 0.0 1.0
51 Kasozi 0.1 0.8 0.4 0.6 0.8 1.0
52 Kalungi 0.0 0.8 1.0 0.0 0.0 1.0
53 Kitaleeba /katuugo 0.1 0.9 0.9 0.4 0.0 1.0
54 Kanyonyi 0.1 0.8 0.4 0.6 0.0 0.0
55 Nakataka 0.0 0.9 0.0 0.3 0.2 1.0
56 Kamirampango PS 0.1 0.9 0.0 0.8 0.0 1.0
57 Kaisolo 0.1 0.8 0.6 0.0 0.0 1.0
58 Nakataka 0.1 0.9 0.0 0.7 0.9 0.0
59 Bulwandi 0.2 0.8 0.0 0.0 0.9 1.0
60 Kasambya SW 0.0 1.0 0.8 0.1 0.9 0.0
61 Lutengo 0.0 0.8 0.0 0.0 0.9 1.0
62 Bagaya primary
school
0.0 0.7 0.0 0.8 1.0 1.0
63 Bamungolode 0.0 0.9 0.7 0.5 0.9 1.0
64 Wanzogi primary
schoo
0.1 0.9 0.3 0.0 0.8 1.0
65 kiranga 0.1 0.9 0.8 0.0 0.0 1.0
66 kyakabombo 0.0 0.9 0.0 0.1 0.0 1.0
67 Ndaiga DWD 0.0 0.9 0.9 0.0 0.9 1.0
106

106

ID Water Point Name Wetlands Slope Roads Schools
Populated
Places
Contaminant
Flow
41739
68 Kiswerwa DWD
41743
0.2 0.7 0.8 0.0 0.9 1.0
69 Namungolo DWD
41742
0.2 0.8 0.0 0.2 0.9 1.0
70 Kalungi DWD
41740
0.1 0.8 0.7 0.0 1.0 1.0
71 DWD 41738 0.0 1.0 0.7 0.1 0.9 1.0
72 DWD 41771 0.1 0.8 0.1 0.0 0.0 1.0
73 DWD 41773 0.2 0.9 0.6 0.0 0.5 1.0
74 DWD 41770 0.1 0.9 0.4 0.0 0.1 1.0
75 DWD 41741 0.2 0.9 0.3 0.0 0.0 1.0
76 DWD 41772 0.0 0.8 0.0 0.0 0.0 1.0

107

107

APPENDIX 2 REFINED MODEL EXISTING WELL SUITABILITY VALUES

ID
Water Point
Name
Contaminant
Flow
Wetlands Depressions Slope Schools Roads
Populated
Places
1 Kazwama 1.0 1.0 1.0 1.0 0.7 1.0 0.6
2 Kamiga 1.0 1.0 1.0 0.9 0.3 0.9 0.7
3 Iryema 1.0 1.0 1.0 0.8 0.7 1.0 0.4
4 Kabandi 1.0 1.0 1.0 0.9 0.0 0.0 0.6
5 Kitaleba 1.0 0.0 1.0 0.8 0.2 0.9 0.0
6 Kalongo TC 1.0 0.0 1.0 0.9 0.4 0.9 0.3
7 Kiwambya 1.0 0.0 1.0 0.8 0.0 0.2 0.5
8 Nakinyama 1.0 1.0 1.0 0.7 0.0 0.1 0.8
9 Kinamwanga 1.0 1.0 0.4 1.0 0.2 0.7 0.0
10 Nakataka 0.4 1.0 1.0 0.8 0.7 0.4 0.4
11 Bagambira 1.0 1.0 1.0 0.8 0.0 0.9 0.7
12 Lutengo 1.0 1.0 0.9 0.9 0.0 0.4 0.6
13 Wanzogi 1.0 1.0 1.0 0.8 0.3 1.0 0.0
14 Rukooge 1.0 1.0 0.9 0.9 0.0 0.2 0.7
15 Kaisolo 1.0 1.0 1.0 0.9 0.2 0.2 0.0
16 Kiranga 1.0 1.0 1.0 0.9 0.0 0.1 0.6
17 Kigejjo 1.0 1.0 1.0 0.9 0.1 0.2 0.4
18 Kalalu 1.0 1.0 1.0 0.8 0.5 0.0 0.0
19 Kamirampango 1.0 1.0 0.5 0.9 0.3 0.2 0.6
20 Nalubobya 1.0 1.0 1.0 0.8 0.0 0.0 0.3
21 Kiwoole 1.0 1.0 1.0 0.8 0.0 0.6 0.2
22 kabazi 1.0 1.0 0.7 0.9 0.1 0.1 0.0
23 Kyalusaka
Primary School
1.0 1.0 1.0 0.8 0.0 1.0 0.2
24 Kyalusaka 1.0 1.0 1.0 0.9 0.0 0.8 0.3
25 Irima 1.0 1.0 0.7 0.9 0.0 0.8 0.8
26 Kigazi 1.0 1.0 0.7 0.9 0.0 0.9 0.9
27 Bagambira
Kyakabona
1.0 1.0 1.0 0.8 0.0 0.8 0.1
28 Dagala 1.0 1.0 0.5 0.9 0.7 0.9 0.0
29 Madaali 1.0 1.0 0.9 0.9 0.8 0.7 0.0
30 Kapundo 1.0 1.0 1.0 0.9 0.6 0.6 0.0
31 Ninga 1.0 1.0 0.6 0.9 0.8 0.7 0.8
32 Kisenyi 1.0 1.0 0.5 0.9 0.6 0.7 0.6
33 kakoola 1.0 1.0 1.0 0.9 0.6 0.8 0.0
34 Nakinyama 1.0 1.0 0.4 1.0 0.5 0.8 0.6
108

108

ID
Water Point
Name
Contaminant
Flow
Wetlands Depressions Slope Schools Roads
Populated
Places
35 Kigali 1.0 1.0 1.0 0.9 0.6 0.9 0.9
36 Ruunyu 1.0 1.0 1.0 0.9 0.8 0.9 0.4
37 Kyakabombo 1.0 1.0 1.0 0.9 0.5 0.2 0.8
38 Kisweramainda 1.0 0.0 1.0 0.9 0.8 0.9 0.0
39 Kaleire 1.0 1.0 1.0 0.8 0.5 0.8 0.6
40 Kigejjo 1.0 1.0 0.9 0.9 0.3 0.1 0.4
41 Nalukonge 1.0 1.0 1.0 0.9 0.0 0.0 0.0
42 Kinamwanga 1.0 1.0 1.0 0.7 0.2 0.6 0.0
43 Namungolo 1.0 1.0 1.0 0.9 0.4 0.1 0.3
44 Nakatuba 1.0 1.0 1.0 0.9 0.3 0.4 0.0
45 Kasozi 1.0 1.0 1.0 0.8 0.6 0.7 0.6
46 Iryema 1.0 1.0 1.0 0.8 0.8 1.0 0.3
47 Kabaandi 1.0 1.0 1.0 0.8 0.3 0.2 0.6
48 Kanyonyi 1.0 1.0 1.0 0.8 0.8 0.5 0.0
49 Rukooge 1.0 1.0 0.4 0.9 0.0 0.4 0.0
50 Kisenyi TC 1.0 0.0 1.0 0.9 0.3 1.0 0.7
51 Kasozi 1.0 1.0 1.0 0.9 0.8 0.7 0.8
52 Kalungi 1.0 1.0 0.9 0.8 0.5 1.0 0.7
53 Kitaleeba
/katuugo
1.0 1.0 1.0 0.7 0.7 0.9 0.0
54 Kanyonyi 0.0 1.0 1.0 0.9 0.8 0.7 0.3
55 Nakataka 1.0 1.0 1.0 0.9 0.6 0.3 0.2
56 Kamirampango
PS
1.0 0.0 1.0 0.8 0.9 0.2 0.7
57 Kaisolo 1.0 1.0 1.0 0.8 0.5 0.8 0.6
58 Nakataka 0.0 1.0 1.0 0.9 0.9 0.2 1.0
59 Bulwandi 1.0 1.0 1.0 0.8 0.0 0.0 1.0
60 Kasambya SW 0.0 1.0 1.0 0.8 0.5 0.9 1.0
61 Lutengo 1.0 1.0 1.0 0.7 0.0 0.4 1.0
62 Bagaya primary
school
1.0 1.0 0.8 0.7 0.9 0.3 1.0
63 Bamungolode 1.0 1.0 1.0 0.9 0.7 0.8 1.0
64 Wanzogi primary
schoo
1.0 1.0 1.0 0.8 0.2 0.7 0.8
65 kiranga 1.0 1.0 1.0 0.9 0.0 0.9 0.0
66 kyakabombo 1.0 1.0 1.0 0.9 0.5 0.4 0.6
67 Ndaiga DWD
41739
1.0 0.0 1.0 0.9 0.0 0.9 0.9
68 Kiswerwa DWD
41743
1.0 1.0 1.0 0.9 0.2 0.9 0.9
109

109

ID
Water Point
Name
Contaminant
Flow
Wetlands Depressions Slope Schools Roads
Populated
Places
69 Namungolo
DWD 41742
1.0 1.0 1.0 0.8 0.6 0.2 1.0
70 Kalungi DWD
41740
1.0 1.0 1.0 0.8 0.1 0.9 0.9
71 DWD 41738 1.0 1.0 1.0 0.8 0.6 0.8 1.0
72 DWD 41771 1.0 1.0 1.0 0.8 0.5 0.6 0.4
73 DWD 41773 1.0 1.0 1.0 0.7 0.5 0.8 0.4
74 DWD 41770 1.0 1.0 1.0 0.9 0.5 0.7 0.1
75 DWD 41741 1.0 1.0 1.0 0.7 0.4 0.7 0.0
76 DWD 41772 1.0 1.0 0.9 0.7 0.0 0.4 0.3

110

110

APPENDIX 3: OVERALL SUITABILITY OF EXISTING WELLS BY REFINED
MODEL

Membership
Range
Membership Index Overall Suitability
.75-100 Most Suitable 0
.50 - 749 Suitable 10
.25 - .499 Not Suggested 11
0 - .249 Not Suitable 55

Membership
Range
Membership Index Overall Suitability
.75-100 Most Suitable 0%
.50 - 749 Suitable 13%
.25 - .499 Not Suggested 14%
0 - .249 Not Suitable 73%

Abstract (if available)

Abstract In 2000 the United Nations (UN) created the Millennium Development Goals (MDGs) to focus on addressing major issues like poverty, education, children’s health, sustainable environment, disease prevention, and economic development. One of the targets (7C) of the MDGs is to halve the portion of the population that does not currently have sustainable access to safe drinking water and basic sanitation by 2015. As a region, sub-Saharan Africa is not on track to meet the goal. In fact, the region has the lowest clean drinking water coverage of any region in the world. This project develops a general framework to improve water resource planning in sub-Saharan Africa. The project defines criteria, data and methods to improve planning for clean drinking water wells. The result is a general framework for 1) finding locations where contamination of water wells is least likely to occur, and 2) ensuring the benefits of clean water support overall community health and education. This is all with the aim to increase efficient water resource planning to support the MDG to increase safe drinking water coverage. The general framework is implemented as a model which is the functional component of the framework. The general framework was refined through the implementation of the model in a model fitting study in rural Uganda. The result of the implementation is a suitability map identifying locations where (1) risks to drinking water are minimized and (2) benefits to people living in the study area are maximized. The success of the model was evaluated by assessing the locations of existing wells against what the model identified as suitable well locations. The framework and model fitting process can be used as a tool by governments and non-government organizations (NGOs) to improve current water site suitability workflows.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Selection of bridge location over the Merrimack River in southern New Hampshire: a comparison of site suitability assessments

PDF

Predicting the presence of historic and prehistoric campsites in Virginia’s Chesapeake Bay counties

PDF

Modeling nitrate contamination of groundwater in Mountain Home, Idaho using the DRASTIC method

PDF

Redefining urban food systems to identify optimal rooftop community garden locations: a site suitability analysis in Seattle, Washington

PDF

Bringing GIS to a small community water system

PDF

Water Rights Permit System (WRPS): a GIS‐based tool for the Umpqua Drainage Basin

PDF

Using GIS to identify potential dynamic marine protected areas: a case study using shortfin mako shark tagging data in New Zealand

PDF

Using Maxent to model the distribution of prehistoric agricultural features in a portion of the Hōkūli‘a subdivision in Kona, Hawai‘i

PDF

Scenario-based site suitability analysis and framework for biodiversity conservation: agricultural zone, Galapagos Archipelago, Ecuador

PDF

Relocation bay: identifying a suitable site for the Tampa Bay Rays

PDF

The role of GIS in asset management: integration at the Otay Water District

PDF

Using pattern oriented modeling to design and validate spatial models: a case study in agent-based modeling

PDF

Estimating populations at risk in data-poor environments: a geographically disaggregated analysis of Boko Haram terrorism 2009-2014

PDF

Concentrated solar thermal facilities: A GIS approach for land planning

PDF

Site location suitability analysis for a smart grid network

PDF

Preparing for immigration reform: a spatial analysis of unauthorized immigrants

PDF

Analysis of future land use conflict with volcanic hazard zones: Mount Rainier, Washington

PDF

A Maxent-based model for identifying local-scale tree species richness patch boundaries in the Lake Tahoe Basin of California and Nevada

PDF

Concentrated solar thermal facilities: A GIS approach for land planning [map]

PDF

Site suitability analysis for implementing tidal energy technology in southern California

Asset Metadata

Creator Hyneman, Jared Charles (author)

Core Title Developing and implementing a GIS-based framework to identify optimal locations for clean water wells in sub-Saharan Africa

School College of Letters, Arts and Sciences

Degree Master of Science

Degree Program Geographic Information Science and Technology

Publication Date 10/29/2014

Defense Date 10/02/2014

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag clean water,fuzzy overlay,GIS,OAI-PMH Harvest,site suitability

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Kemp, Karen K. (committee chair), Lee, Su Jin (committee member), Warshawsky, Daniel N. (committee member)

Creator Email hyneman@usc.edu,jared.hyneman@gmail.com

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c3-509652

Unique identifier UC11287947

Identifier etd-HynemanJar-3038.pdf (filename),usctheses-c3-509652 (legacy record id)

Legacy Identifier etd-HynemanJar-3038.pdf

Dmrecord 509652

Document Type Thesis

Rights Hyneman, Jared Charles

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the a...

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA