Escolar Documentos
Profissional Documentos
Cultura Documentos
A thesis presented by
Heather Carmichael
to
Harvard College
Cambridge, Massachusetts
March 2009
Acknowledgements
My research would not have been possible without the help of many
people in Accra as well. I have to thank our field assistants—Tete, Laye,
Nazar, Abu, and Kilu—for assisting me around Nima and Jamestown, and for
lively discussions about local and world politics. My second trip to Accra
would not have been successful without the help of Allison Hughes and Nana
Prempeh, who offered both advice and company on numerous occasions. I
would not have made it through the summer, or indeed, out of the airport, if
it were not for Raphael and Comfort Arku, who came to rescue me whenever
anything went wrong, even in the midst of their honeymoon. Finally, I
2
particularly have to thank Adam Abdul Fatah and his wife, Hawa, for their
generosity and hospitality, as well as their insight and inspiration, and for
lending me their baby daughter, Amirat, when I got lonely.
2
Abstract
human health and contribute substantially to the global burden of disease. However,
much of the research that has been done on environmental risks and interventions in the
developing world has focused on rural regions. Meanwhile, urban areas in poor countries
are experiencing rapid growth; while this expansion presents challenges to developing
providing services such as water supply, sanitation, and waste removal. Focusing on the
city of Accra, Ghana, my thesis moves beyond the study of risks in the household
environment and considers the neighborhood effects that occur in areas of concentrated
areas, my thesis also seeks to quantify the magnitude of environmental disparities within
particular the problems of littered solid waste, open drainage systems, and household
water quality in a quantitative analysis of original data, showing that these risks can be
researched and assessed in resource-poor settings at relatively little cost. My thesis also
serves as a body of empirical evidence that could be used to promote policies to improve
2
Table of Contents
Chapter 1 : Introduction...................................................................................................1
Chapter 2 : An Introduction to Accra and the Study Areas.........................................11
Chapter 3 : Datasets and Methods of Data Collection.................................................20
Chapter 4 : Water, Hygiene and Sanitation in the Greater Accra Metropolitan Area
Census and Other Household Surveys ..........................................................................30
Chapter 5 : The Spatial Distribution of Littered Solid Waste.....................................47
Chapter 6 : Liquid Waste and Roadside Drainage Gutters ........................................64
Chapter 7 : Household Drinking Water—Source versus Consumption.....................76
Chapter 8 : Discussion, Conclusions, and Suggestions for Future Policy...................94
Works Cited...................................................................................................................106
Appendix A....................................................................................................................107
Appendix B....................................................................................................................108
2
List of Figures and Tables
Figures
Figure 2-1: Beach area (top) and residential street (bottom) in Jamestown..............16
Figure 2-2: Rooftop view (top) and public waste collection site (bottom) in Nima.. .17
Figure 2-3: Residential area (top) and the Ring Road (bottom) in Asylum Down... .18
Figure 2-4: Developing area along the walking path (top) and gated houses in East
Legon (bottom).................................................................................................................19
Figure 3-5: Water collected from a public tap in Jamestown and stored outside a
house in a plastic barrel...................................................................................................24
Figure 3-6: Metal barrels used to store water in the courtyard of a household
compound in Nima...........................................................................................................24
Figure 3-7: An example of a highly polluted consumption sample collected from a
household in Jamestown. E. coli colonies appear blue or purple, while other
coliforms form pink colonies...........................................................................................25
Figure 3-8: Testing for E. coli and total coliforms using Coliscan® Easygel® system
—incubation at room temperature................................................................................25
Figure 3-9: Map of study areas and walking paths showing road type and surface,
as well their general location in the city (courtesy of Ari Friedman)..........................27
Figure 3-10: Partially covered cement gutter in East Legon with no trash and slow
flowing water (left) and clogged open gutter in Nima with stagnant water flow
(right). ..............................................................................................................................29
Figure 4-11: Population density in the Greater Accra Metropolitan Area (boxes show
the approximate locations of the study areas)...............................................................32
Figure 4-12: Persons per bedroom for households in the Greater Accra Metropolitan
Area, aggregated by enumeration area..........................................................................32
Figure 4-13: Water sources by EA in the Greater Accra Metropolitan Area.............34
Figure 4-14: Water sources in the four neighborhoods using data from the 2000
census (left) and the 2007 household surveys (right)....................................................35
Figure 4-15: Household methods of solid waste disposal include (clockwise from top
left) disposal at public collection sites, private collection of solid waste, illegal
dumping (for example, into open drainage gutters) and burning of household trash.
...........................................................................................................................................39
Figure 4-16: Household methods of solid waste disposal by EA in the Greater Accra
Metropolitan Area............................................................................................................40
Figure 4-17: Methods of solid waste disposal in the four neighborhoods, according to
the 2000 census (left) and the 2007 household surveys (right).....................................41
Figure 4-18: Liquid waste disposal methods by EA in the Greater Accra
Metropolitan Area............................................................................................................43
Figure 4-19: Liquid waste disposal methods in the four neighborhoods, according to
the 2000 census (left) and the 2007 household surveys (right).....................................45
Figure 4-20: Toilet facilities in the four neighborhoods, according to the 2000 census
(left) and the 2007 household surveys (right)................................................................46
2
Figure 5-21: "Sachet" water is a common commercial source of drinking water in
Accra (left). The plastic sachet then becomes a common source of littered waste
(right)................................................................................................................................49
Figure 5-22: The spatial distribution of plastic waste in the four study areas. Only
Asylum Down and Jamestown show significant spatial autocorrelation....................52
Figure 5-23: The spatial distribution of food waste in the four study areas. Only
Nima shows significant spatial autocorrelation............................................................53
Figure 5-24: The spatial distribution of paper waste in the four study areas. No
neighborhoods showed significant spatial autocorrelation..........................................54
Figure 5-25: Map depicting the littered waste data collection points in Asylum
Down, along with the population density raster. The red circles represent buffer
zones (50 meter radius) for which an average population density was calculated....57
Figure 5-26: Histograms of this test statistic for replicate datasets sampled from the
model. Lines in red show the test statistics for the actual dataset..............................62
Figure 6-27: Level of trash in open gutters summarized for the four neighborhoods.
...........................................................................................................................................66
Figure 6-28: Level of littered waste in gutters along the walking paths for the four
study areas........................................................................................................................67
Figure 6-29: Stagnant waters provide breeding grounds for mosquitoes and other
pests (left). Shallow gutter with high flood risk in Jamestown (right).......................68
Figure 6-30: Classification of water flow in gutters along the walking paths for the
four study areas................................................................................................................71
Figure 6-31: Measure of flood risk for gutters along the walking paths for the four
study areas........................................................................................................................72
Figure 6-32: Comparison of trash level to water flow for all measured gutters
(pooled across neighborhoods).......................................................................................74
Figure 7-33: Classes of coliform bacteria used as indication of biological
contamination in water (Source: Washington State Department of Health, 2007).. .82
Tables
Table 4-1: Mean population densities for EAs in each locality (standard deviations
shown in parentheses). ...................................................................................................31
Table 4-2: Mean number of persons per bedroom for households in each
neighborhood for the 10% household sample (standard deviations shown in
parentheses)......................................................................................................................31
Table 5-3: Mean level of different waste types by neighborhood (ranges shown in
parentheses)......................................................................................................................48
Table 5-4: Regression of log1plastic on neighborhood.................................................50
Table 5-5: Moran's I values for each neighborhood for different types of littered
waste. Data demonstrating high levels of spatial autocorrelation (unlikely to be
random) are marked by asterisks...................................................................................51
Table 5-6: Results for OLS regression of log1plastic on road capacity and surface..57
Table 5-7: OLS regressions of log1plastic on two measures of population density....58
2
Table 5-8: Results of the OLS regression of log1plastic on the percentage of the
population using a public dumpsite for solid waste collection.....................................59
Table 5-9: Summaries of posterior densities (mean and 95% confidence interval) for
important estimated parameters....................................................................................60
Table 6-10: Summary of important characteristics of gutter water in the four
neighborhoods..................................................................................................................70
Table 6-11: Average width and depth of gutters in the four neighborhoods (in
centimeters)......................................................................................................................70
Table 6-12: Flood risk of gutters with specified characteristics, namely, level of
littered waste (left) and water flow (right).....................................................................75
Table 7-13: Piped sources used by sampled households within each neighborhood. 78
Table 7-14: Of households that store drinking water prior to consumption,
percentages using various containers for water storage. ............................................79
Table 7-15: Location of the storage container within the home in each neighborhood
(as a percentage of households that store drinking water prior to consumption). ...79
Table 7-16: Percentage of households demonstrating particular hygiene behaviors in
the household prior to consumption from piped sources. ..........................................80
Table 7-17: Comparisons of E. coli contamination in piped samples, at the source
and at the point of consumption. ...................................................................................85
Table 7-18: Comparisons of total coliforms in piped samples, at the source and at
the point of consumption.................................................................................................85
Table 7-19: Difference in E. coli contamination between consumption samples and
source samples for all households with piped water sources, by source type.............88
Table 7-20: Difference in total coliform contamination between consumption
samples and source samples for all households with piped water sources, by source
type....................................................................................................................................89
7-21: Mean differences in E. coli and total coliform contamination between
consumption samples and source samples for all households with piped water
sources, by storage location. Numbers in parentheses include TNTC samples by
approximating TNTC counts at the minimum 200 cfus/5mL......................................90
7-22: Mean differences in E. coli and total coliform contamination between
consumption samples and source samples for all households with piped water
sources, by storage container type. Numbers in parentheses include TNTC samples
by approximating TNTC counts at the minimum 200 cfus/5mL................................92
2
Chapter 1: Introduction
In 2007 the global urban population exceeded the rural population for the first
time in history.1 Trends suggest that population growth in the next 30 years will occur
almost entirely in the urban areas of low- and middle-income countries (LMIC).2 The
rapid rates of growth in these urban areas are due both to high rates of rural-to-urban
migration and high urban birth rates. As cities struggle to cope with rapid growth and
development, infrastructure often lags far behind what is needed to manage pollution and
wastes.
One outcome of rapid urbanization in the developing world has been the
proliferation of city slums. Nearly one sixth of the global population lives in slums,
defined as areas where the population lacks security of tenure, adequate housing, safe
drinking water and sanitation.3 Almost four billion people are expected to live in slums
by 2030. In developing countries, the proportion of the urban population who live in
opportunity, and therefore attempted to accelerate rural development as the primary focus
indicated that they had initiated policies to slow urban growth by reducing rural-to-urban
1
2
3
4
5
2
migration.6 It is thought that by creating better opportunities for rural populations,
countries can discourage the influx of rural migrants into urban centers, and thus solve
the “problem” of urbanization.7 This perspective can foster a misconception of the urban
poor as transient populations that present a merely temporary problem, one that needs not
urban poverty, government officials also neglect to formalize patterns of settlement for
rural migrants as they flock to urban areas. The urban poor are thus a marginalized
population excluded from many aspects of social development and welfare including
understood concept, established by Michael Lipton (1976).10 Urban bias refers to the
tendency of city-dwellers, who live in centers of political, economic, and social activity,
to have greater access to policy-makers and government than their rural counterparts. The
corollary to this is the notion that urban poverty ought to be less of a concern when facing
poverty as a whole, and that the majority of the focus in development should be on the
However, ironically, the assumption of an urban bias may have resulted in greater
attention to means of improving the lives of the rural poor in developing countries, at the
6
7
8
9
10
2
expense of research focused on the living conditions of their urban brethren. Saumitra
If it is in fact the poorest migrants that gain from improved access to decision makers and
services by moving to the city, then urbanization, rather than being a strain on resources
that is to be discouraged, may in fact provide a means to better governance and poverty
alleviation. Unfortunately, due to an arguably strong “rural bias” in development data
and inquiry, little quantitative evidence is available either on the strategies of network
formation or on the actual political networks of the urban poor.11
Thus, it remains unclear whether the urban poor actually benefit from the urban bias in
development, and this lack of clarity is in part due to the imbalance of research and
Although Jha focuses on the lack of research that exists in the social and political
realm, this rural bias is also evident in the study of environmental risks and intervention
methods. While research exists on the associations of environmental risks with poverty
in the developing world, urban settings have generally received less attention than rural
hygiene in the developing world, conducted in 2005, considered all studies in the field
published prior to June, 2003. Of a total of 38 studies selected for review, only 11 were
areas (the locations of the other 6 studies were unstated). Of the 24 studies deemed to be
of “good” quality, only 3 were in urban areas and 3 in peri-urban areas, while 14 had
been undertaken in rural settings.12 While the quality and quantity of research on
environmental risks in urban areas is certainly improving13, there is still much progress to
be made.
11
12
13
2
The environmental risk and the physical deprivation experienced by urban slum
residents can be staggering, with many health risks at levels substantially higher than
those considered safe or seen in other areas of the city. Inadequate access to safe water
influences the spread of disease via water-borne pathogens, and water scarcity also
the neighborhood and household environment, and in turn can contaminate water sources.
Inadequate waste removal exacerbates disease by providing breeding grounds for flies
and other pests, and littered waste also clogs drainage systems, which then act as
breeding grounds for mosquitoes that spread malaria and dengue. Poor structural quality
of housing leads to higher risk for physical injury and overcrowding facilitates the rapid
The environmental risks described above indicate a need for more detailed
empirical study in the urban developing world. This research needs to extend beyond the
study of the household environment, to the neighborhood effects that occur in areas of
concentrated poverty. Thus, one goal of this thesis is to examine closely both
14
2
The Underestimation of Urban Poverty: Definitions and Discrepancies
Studies on urban poverty can easily underestimate its pervasiveness and its
implications. This is in part because average urban incomes exceed average rural
incomes by a significant margin, which is all too often cited as a reason to divert focus
away from the urban poor. The UN Millennium Project Report on Improving the Lives
of Slum Dwellers states: “The conventional wisdom is that urban areas are better off than
rural areas and that urban slum dwellers live better lives than their rural counterparts.”15
This underestimation of poverty in the city is partially due to a lack of adjustment for cost
of living. Although the urban poor are on average wealthier than the rural poor, their
earnings often fail to cover even the most basic necessities when living in the city.
Definitions of lack of access to safe drinking water, sanitation, and shelter are also
inadequate and may misrepresent the experiences of the urban poor. While many urban
residents meet the definitions of having access to “improved provision” of water and
sanitation, these facilities may be shared by hundreds of people, greatly increasing the
hospitals or water sources does not ensure that poorer residents will have access to these
amenities.16
Finally, when considering the urban poor there is also a problem of aggregate
statistics, which obscure the huge disparities that often exist in the city. Middle and
upper class households are also concentrated in urban areas, and can obscure the
differences between rural and urban poverty. Under-five mortality rates in informal
settlements in Nairobi, for instance, are more than twice the city’s average, while these
15
16
2
informal settlements comprise nearly half the city’s population.17 It is unclear whether
the urban bias extends to the urban poor, who may be just as distanced as the rural
population from formal government and access to policymakers who live just next door.
Although the meaning and scope of the term environmental equity remains vague,
the United States Environmental Protection Agency defines it as “the equal protection
cities has been limited, although marked differences exist in the living conditions and
environmental differences and disparities in the urban setting of the developing world.
human health, accounting for close to one-fifth of the burden of disease in developing
factors in the developing world are inadequate water, sanitation, and hygiene (WSH),
indoor air pollution from solid fuels, urban ambient air pollution, and vector borne
health outcomes for populations in developing countries. It has been estimated that 4.79
million children under the age of five years died in sub-Saharan Africa in 2004, an
17
18
19
20
2
estimated annual child mortality of 39.1 per 1000 children.21 Among other things, the
two-thirds. However, one study has estimated that 50% of the gap towards meeting this
child mortality goal in sub-Saharan Africa could be reached if the three nutritional and
sanitation, and reduction in the use of solid fuels—were achieved for all households with
the environmental risk factor of indoor air pollution can be influenced, among other
things, by the fuel or stove types used by a household, the location of vendors for a given
fuel type, the amount of time that the stove is on, or the amount of time spent near the
stove.23 This can make causality, as well as the health effects of individual interventions,
difficult to establish.
21
22
23
2
Water, Sanitation and Hygiene (WSH)
Adequate water, sanitation and hygiene (WSH) have been identified as the key
environmental risk factors for diarrheal diseases, which contribute heavily to the burden
of disease in developing countries. Diarrheal diseases are estimated to cause 3.3 million
deaths globally each year and are responsible for more than one-quarter of childhood
pathogens such as viruses, bacteria, protozoa and parasitic worms that are transmitted via
fecal matter excreted by the human host. Thus, poor WSH contributes to the spread of
diarrhea, poor WSH can increase the risk of other diseases such as schistosomiasis,
trachoma, ascariasis, trichuriasis, and hookworm, and also contributes to the risk of
diseases such as typhoid, malaria, dengue, yellow fever, hepatitis A and hepatitis E.25
also plays a substantial role (transmitted via hands or food due to lack of water and
adequate sanitation).26 Thus, WSH can be defined broadly to encompass many factors
including ingestion of contaminated water, lack of water, poor hygiene practices in the
home and neighborhood, contact with unsafe water in agriculture and inadequate
management of water resources and water systems (such as drainage gutters).27 Chapter
24
25
26
27
2
water. Chapters 5 and 6 focus on neighborhood gutter systems and neighborhood solid
Risk can be influenced by factors at the household level or in the surrounding area
—household vicinity—in which case they are generally behavioral. Alternately, WSH
the infrastructural level, WSH can be affected by the availability and type of water
sources, the quantity of water available, and the sewage and waste disposal
and behavioral choices at the household level, such as the household source of water and
storage method, type of latrine, and the household or community techniques for waste
disposal. Additional technological factors include point of use decontamination for water
(such as boiling or filtration), and food and hand washing techniques (such as soap).
Finally, WSH can be affected by behavioral factors such as hygiene, methods of water
transportation and storage, and waste disposal methods.28 This hierarchy is summarized
below.
28
2
The interaction between urbanization, poverty and the environment requires
further inspection. I designed this study to assess some of the differences in the
hygiene and sanitation, while omitting the equally important problems of indoor and
The second chapter of this thesis describes some of the specifics of the Accra
metropolis, as well as the four neighborhoods chosen for the study. In the third chapter, I
describe the methods used in data collection for this thesis. The fourth chapter uses data
available on these areas from the Accra census as well as a much smaller but more recent
set of household surveys done in the four neighborhoods to explore demographic trends
in the four neighborhoods as well as drinking water sources, methods of solid and liquid
waste disposal, and sanitation facilities. The fifth, sixth, and seventh chapters present the
results of original data I collected in Accra during the summer of 2008. In the fifth
chapter, I focus on the spatial distribution of littered waste in the four neighborhoods.
The sixth chapter analyzes the distributions of open drainage gutters that line the streets
of Accra, intended for storm water runoff, but generally used for liquid waste disposal by
many city residents. The seventh chapter considers household drinking water quality,
hygiene and sanitation, including an analysis of samples that were tested for E. coli and
other coliforms. The final chapter will conclude and include suggestions for policy in the
future.
2
Chapter 2: An Introduction to Accra and the Study Areas
Accra is, in many ways, an ideal location for the study of urbanization in the
developing world and its associated health risks. Similarities in urban structure and
development, poverty, and past colonial history make Accra fairly representative of other
cities in coastal West Africa.29 Accra is characterized by a sprawling urban area and
independence, which has led to unplanned growth and the proliferation of slum
communities. UN Habitat estimates that 66% of the city’s population lives in urban
slums.30
generalization to other parts of sub-Saharan Africa, the city is unique due to the quantity
of data available on the environment, poverty, and health. Thus, Accra is somewhat of an
exception to the rural bias in research on environmental risk discussed in the previous
chapter. Several important studies and surveys should be mentioned here. The 2000
National Population and Housing Census collected information on water and sanitation in
the household environment at a high level of spatial detail. The Stockholm Environment
cities (Accra, Jakarta, and Sao Paulo) in 1991, including citywide surveys of 1000
households in Accra. Professor Jacob Songsore of the University of Ghana has published
29
30
2
important environmental health indicators in all residential areas. These studies assessed
water, sanitation, pests, sullage and drainage, food contamination, hygiene, solid waste,
housing problems, and indoor and outdoor air pollution at the neighborhood level.31
In the 2000 census, the Accra Metropolitan Area had a total population of
1,658,937. Based on an annual growth rate of 4.4% in the Greater Accra Region, the
expected population of the city in 2008 was approximately 2.3 million.32 This rapid
growth, which continues to the present day, is largely due to rural-urban migration, as the
total fertility rate in the metropolitan area, at 2.2 children per woman per lifetime, is the
lowest in the country.33 Indeed, the growth rate for this region far outpaces the national
average of 2.7% per year.34 Other estimates indicate that the floating population
[homeless and those without permanent residences] in the city could be as high as half a
million daily, with nearly half this population remaining in the city overnight.35 Forty-
eight percent of Accra’s population is below the poverty line and the city experiences
acute problems of sanitation, waste management and pollution coupled with this rapid
urban expansion.36 Data from the 2000 census indicated that 73.9% of households in
Accra did not have a toilet with a sewer connection, 56.3% had no piped water in the
house, and 47.8% had more than three people per room.37
31
32
33
34
35
36
37
2
The five most reported diseases at outpatient facilities in Accra indicate the
Songsore et. al. the health problems in the city are primarily “preventable and
and poverty.”38 Malaria is the most common, followed by upper respiratory tract
Within the city, large disparities exist in the health and environmental conditions
rates have been found between poor and wealthy neighborhoods.39 Household living
study by Songsore et. al. of 1000 households in Accra found that while only 10% of
households in the wealthiest quintile had to fetch water from outside their household
compound, nearly 80% of households in the poorest quintile did. While 71.7% of
households in the poorest quintile used pit latrines, 77.3% of households in the wealthiest
quintile had access to a flush toilet. Similarly, 68.8% of the poorest households shared a
toilet facility with more than ten other households, while 64.6% of the wealthiest
of varying socioeconomic status. Two of these communities, Jamestown and Nima, can
of Accra, while East Legon is an area of relative affluence on the outskirts of the city.
38
39
40
2
Jamestown and Usshertown
census. The study area includes both neighborhoods, but will henceforth be referred to
as Jamestown only. Jamestown is located along the coastline in the old colonial area of
Accra, and many of the older colonial buildings have been preserved and now house
multiple households. The primary ethnicity in this study area is Ga, and the population is
for the most part poor and uneducated. The study area includes the main road that passes
from central Accra along the coast to the Korle-Bu area, as well as a large area along the
beach with a sizable squatting settlement and large boat yard. It is bordered on the south
Nima
In Nima, the population according to the 2000 census was 69,044. Located in the
central part of the city, Nima has a large Muslim population of Hausa origins, and is
neighboring slum areas of Mamobi (population 49,812) and New Town (population
45,130). The study area enjoys a mix of residential and commercial use, with a large
market place located to the east of the main road. The Al-Hamdu Gutter, one of the main
open drainage gutters for the city, passes along the western border of the study area, and
the land has a substantial gradient going down towards this drain.
Asylum Down
2
Asylum Down, with a population of 9,363 according to the census, is located in
the central part of the city and has a mix of residential and commercial areas, as well as a
mix of ethnicities and socioeconomic statuses. Poorer populations live in areas along the
main gutter that are prone to severe flooding during periods of heavy rain. The study area
also includes the Central Ring Road, one of the larger highways in the city, which is
East Legon
The population of East Legon was 7,681 at the time of the 2000 census; however,
the area has grown substantially since then. East Legon is located near the University of
Ghana on the periphery of the city. The area is relatively affluent and largely residential,
developed, and many areas lack some of the infrastructure found in other parts of the city,
such as paved roads and sewage systems. Also, the community includes pockets of
2
Figure 2-1: Beach area (top) and residential street (bottom) in Jamestown.
2
Figure 2-2: Rooftop view (top) and public waste collection site (bottom) in Nima.
2
Figure 2-3: Residential area (top) and the Ring Road (bottom) in Asylum Down.
2
Figure 2-4: Developing area along the walking path (top) and gated houses in East Legon (bottom).
2
Chapter 3: Datasets and Methods of Data Collection
This chapter summarizes briefly the data used for analysis in this thesis, including
three original datasets that I collected in Accra during the summer of 2008 with the help
Household Data
The 2000 Population and Housing Census was the fourth population census
undertaken in Ghana since independence and the most recent since 1984. The census was
arrangements made for the floating population. The census was enumerated at the level
of the individual, but data were also collected for households. A household is considered
by the GSS to be “a person or group of persons who live together, in the same house or
compound, share the same housekeeping arrangements and are catered for as one unit.”41
The smallest statistical unit of the census is the enumeration area (EA), a
geographic area with clear physical boundaries that is enumerated by one census official.
There were 26,710 enumeration areas in the country, and 1,724 in the Greater Accra
Metropolitan Area (GAMA). The average population of each EA was 750, making the
EA similar in size to U.S. census blocks.42 Another important geographic unit in the
census is the locality, which is defined by the Ghana Statistical Services as a “nucleated
and physically distinct population cluster in which the inhabitants live in neighboring
41
42
2
living quarters and which has a name or locally recognized status.”43 In the GAMA,
communities.
I have access to a 10% random sample of the census for the entire Greater Accra
Metropolitan Area in 2000, courtesy of Professor Allan Hill. This sample includes
information on liquid and solid waste disposal methods, toilet and bathing facilities, and
water sources for a random sample of 10% of households in the GAMA. A rough
estimate of population density can also be obtained by dividing the total population in
The primary focus of these surveys was to gather information related to household
use of solid fuels, household exposure to indoor air pollution, and respiratory disease.
However, some survey questions were directed more generally toward the household
environment, and pertain to water, hygiene and sanitation. As in the 2000 census,
respondents were asked about their water sources, toilet facilities, and methods of
disposal for solid and liquid waste. Although the sampled households represent a much
smaller dataset than is available through the census, with only 18-20 households per
neighborhood, the data are more recent, and may provide clues as to how the household
environment is changing in these areas of Accra. This survey was approved by the
Institutional Review Board (IRB) of the Harvard School of Public Health and the results
43
2
Household Water Quality and Water Storage (Original Dataset, 2008)
the air pollution studies described above. Two 12 mL samples were taken in each
household; one is denoted the consumption sample, and was taken one at the point of
consumption (as it would be consumed by a household member) and was taken from the
source (the tap or location from which the household fetches water). If the household’s
source was commercial (i.e. bottled or sachet water) only one sample was taken. The
household head was asked to identify his or her household’s primary source of drinking
water. Drinking water was defined as water that is used for direct consumption, and not
used for bathing or cooking. Specifically, the household head was asked to provide a
The household’s primary source of drinking water was categorized as one of the
following:
Additionally, the source was classified as being private (only for the household or
The household head was also asked to identify any alternative sources of drinking water
2
I then observed the following characteristics of the source and storage of the
consumption sample:
5mL of each sample was tested in field conditions for E. coli and total coliforms,
indicators of fecal contamination, using the Coliscan® Easygel® system.44 Results are
presented in colony-forming units per 5mL (cfus/5mL). These data are right-censored, as
E. coli and total coliform counts greater than 200 cfus/5mL were denoted too numerous
to count (TNTC). A more detailed description of the testing methods and protocol can be
found in Appendix A.
44
Micrology Laboratories, LLC (Goshen, IN).
2
Figure 3-5: Water collected from a public tap in Jamestown and stored outside a house in a plastic
barrel.
Figure 3-6: Metal barrels used to store water in the courtyard of a household compound in Nima.
2
Figure 3-7: An example of a highly polluted consumption sample collected from a household in
Jamestown. E. coli colonies appear blue or purple, while other coliforms form pink colonies.
Figure 3-8: Testing for E. coli and total coliforms using Coliscan® Easygel® system—incubation at
room temperature.
2
Neighborhood Data
“walking path” used in previous studies for urban air pollution research. Each path
road capacities and surfaces, which have been classified and recorded in previous studies
(Figure 3-5). Along the walking paths, I took a survey of the level of litter and trash,
Alternating points were sampled on each day. The following parameters were measured
within a radius of 5 meters, not including any trash or waste that was within a gutter, or
within a household/individual waste container (if the number of items in a given count
exceeded 100, the level of waste was estimated from a smaller area count):
2
Figure 3-9: Map of study areas and walking paths showing road type and surface, as well their
general location in the city (courtesy of Ari Friedman).
2
Although the data were not analyzed in this study, I also located and marked all public
toilets, bathing facilities, and public rubbish dumpsites or collection points within each
neighborhood area.
I also classified and recorded the location of all the gutters along the walking
paths. Each segment of gutter was classified according to its structure, size, and the
amount of water and trash in it. Each section of the walking path for each neighborhood
was described according to the following parameters, using a waypoint to mark its start:
1. ID: Recorded both with GPS waypoint and other attributes (below)
2. Location: Recorded using GPS waypoint
3. Altitude: Recorded using GPS waypoint
4. Time: Recorded GPS time
5. Approximate Location: A brief description of the waypoint location.
6. Type: Each segment of the walking path was classified as having no gutter
or as having a gutter
7. Side of road (L or R): When gutters are on both sides of the roadway used,
the right side was classified. If there was no gutter on the right side the left
side was classified.
8. Opposite: The opposite side of the path was classified as having no gutter
or as having a gutter.
Only gutters running parallel to the path were classified, unless the width of the
gutter crossed was greater than 1 meter. Isolated gutters less than 3 meters long were not
recorded. All gutters were also classified according to the following variables:
2
8. Trash: The quantity of solid waste in the gutter was classified as none, low
(<10 % estimated coverage), medium (<40 % estimated coverage), high (<
70 % coverage), or clogged (>70 % coverage)
9. Turbidity: Transparent or opaque.
10. Color
Figure 3-10: Partially covered cement gutter in East Legon with no trash and slow flowing water
(left) and clogged open gutter in Nima with stagnant water flow (right).
Other Research
I also visited important agencies within Accra to learn about sanitation and water
issues from a more qualitative perspective. I visited the National Water Research
Institute, as well as the Ghana Water Company (GWCL), which handles the distribution
of water throughout the city. I also met with an official at the Ministry of Local
as the Federation of Youth Clubs (FYC) in Nima, which has been actively campaigning
for the construction and improvement of the major gutter that runs through the
2
Chapter 4: Water, Hygiene and Sanitation in the Greater
Accra Metropolitan Area Census and Other Household
Surveys
This chapter uses household-level data from the 2000 National Census and from
may generate higher levels of littered waste, or may have more acute problems of sullage
and drainage. Crowding can also directly impact the spread of communicable disease by
The map of population density for the Greater Accra Metropolitan Area (Figure
4-1) highlights that the majority of the urban population is concentrated in urban slums
neighborhood using the mean population density of enumeration areas in that locality as
found in the 2000 census; Table 4-1 below summarizes these results. In particular, the
EA-level population densities in Nima and Jamestown are more than five times greater
than those of Asylum Down, and are almost fifteen times greater than the densities in
2
East Legon. While population densities in Jamestown and Nima are nearly identical, the
variation between EAs in Nima is much greater than in Jamestown, indicating that
household level. Differences in persons per bedroom (Figure 4-2) are less pronounced
across the Greater Accra Metropolitan Area than are differences in population density.
Similarly, the differences between the study neighborhoods are not as dramatic for this
measure (Table 4-2); while households in Nima and Jamestown, on average, have more
persons per bedroom than do households in Asylum Down or East Legon, the differences
are relatively small. Notably, while the population density in Asylum Down is twice that
of East Legon, households in Asylum Down are less crowded than those in East Legon.
This reflects the different patterns of land use in each community; while nearly all land in
Nima and Jamestown is residential and built on, portions of land in Asylum Down are
2
Figure 4-11: Population density in the Greater Accra Metropolitan Area (boxes show the
approximate locations of the study areas).
Figure 4-12: Persons per bedroom for households in the Greater Accra Metropolitan Area,
aggregated by enumeration area.
Accra experiences problems of acute water stress due to the rapid expansion of
the city with little complementary development in the piped water supply for developing
areas. Additionally, the piping distribution system in some parts of the city is more than
2
thirty years old and suffers from deteriorating quality.45 Almost without exception,
households that I visited in Accra described some inconsistency in their supply of water.
In wealthier areas, households generally have indoor piping and store water in large,
sealed “polytanks” for when there is water outage. In poorer areas, most households use
a communal source of water—either a public tap or a tap in the yard of the household
outside the home have, on average, higher diarrheal prevalence for children under the age
of six than do households with indoor piping. Similarly, households that must store water
or have interrupted water supply have significantly increased risk (4.3 and 3.1
approximate relative risk, respectively) for diarrheal diseases in children under the age of
six.46
Both the 2000 census and the 2007 household surveys conducted by Harvard
researchers show clearly differentiated patterns of water sources for households across
neighborhoods. The results from the 2000 census show that the majority of households
in Nima (n = 1534) and Jamestown (n = 917) use pipes outside their home as their main
supply of water. Households in Asylum Down (n = 439), on the other hand, generally
had an inside pipe, with a smaller minority using outside pipes. In East Legon (n = 271),
the results were mixed, with relatively similar numbers of households using outdoor and
indoor pipes. Only a small fraction of the population in each neighborhood used other
45
46
2
Figure 4-13: Water sources by EA in the Greater Accra Metropolitan Area.47
47
Note: Drinking water sources are evaluated at the household level. The following options are listed:
outside pipe, indoor pipe, tanker supply, borehole, spring/rain water, river/stream, dugout, other. For clarity,
enumeration area boundaries are not shown. The bottom frame (alternative methods) includes households
that use a source that is not pipe-borne.
2
The results of the 2007 household surveys were quite different; this can be
attributed to a number of factors. First, this survey represented a much smaller sample of
sources of drinking water in this survey, which is often distinguished in source from
water for other purposes such as cooking, bathing, and cleaning. Also, households were
given a broader selection of response options on this survey that included commercial
sources of water such as water sachets and bottled water. Households were allowed to
list multiple water sources on the later survey, but only one on the census. Households
may also have different conceptions of the definitions of different water sources; a tap
that is in the yard of the family compound but is shared by several households in the
importantly, sources of water could have changed significantly in the intervening years
Figure 4-14: Water sources in the four neighborhoods using data from the 2000 census (left) and the
2007 household surveys (right).
2
The 2007 survey gives insight into sources of drinking water in the four
primarily listed a public tap as their water source (57%), followed by water piped into the
yard (21%), and lastly water piped into the dwelling (14%). Households with two
sources (n = 7), almost without exception, used a combination of water piped into the
In Jamestown, public taps (43%) and water piped into the yard (43%) were the
most common sources of drinking water for households with one primary source (n =
14), with a minority using water piped into the dwelling (14%). Households with two
sources (n = 7) most frequently used both sachet water and water from a piped source,
In Asylum Down, the sources used were much more varied. Of those households
using one water source (n = 11), there was a more even distribution of sources: sachet
water (36%), piped into dwelling (27%), public tap (18%), bottled water (9%) and piped
into yard (9%). For households with two sources (n = 9), all households used a
combination of sachet water and a piped source. For most multi-source households, this
second source was either piped into the yard (66%) or piped into the dwelling (22%).
Households with only one source (n = 8) used either water piped into the dwelling (38%),
bottled water (25%), sachet water (25%), or water piped into the yard (13%).
Households with more than one source (n = 10) used, almost without exception, a
combination of sachet water and an alternative source. This alternative source was most
often water piped into the dwelling (50%) or bottled water (30%).
2
Importantly, it appears that in some neighborhoods, particularly in Asylum Down
and Jamestown, households may supplement an existing piped source with water from
commercial sources, primarily sachet water. This phenomenon was not reflected in the
census data, both because the census did not specify water from commercial sources on
their survey, and because households could not list multiple water sources. Thus, it is
impossible to determine from these data if the relatively common use of commercial
water usage.
Household solid waste disposal methods in Accra were found to be both diverse
dumpsites (61.3 % of all households), having trash collected (20.4 %), burning trash (6.6
%), burying trash (3.0 %), and dumping trash elsewhere (5.6 %). Mapping these
location (see Figure 4-6). The more central and densely populated slum areas, including
Nima and Jamestown, generally relied on public containers to dispose of solid waste.
Areas in the central part of the municipality, including Asylum Down, were serviced
primarily by private waste collection. Areas on the periphery of the urban metropolis,
such as East Legon, regardless of socioeconomic status, tended to use alternative methods
2
2
Figure 4-15: Household methods of solid waste disposal include (clockwise from top left) disposal at
public collection sites, private collection of solid waste, illegal dumping (for example, into open
drainage gutters) and burning of household trash.
The results of the 2007 surveys show similarly differentiated patterns of solid
waste disposal. The majority of households surveyed in Nima and Jamestown used
public dumpsites, while all households in Asylum Down reported that their waste was
collected. East Legon demonstrated a variety of waste disposal methods in both surveys,
with the majority of households using waste collection, but with significant minorities
2
Figure 4-16: Household methods of solid waste disposal by EA in the Greater Accra Metropolitan
Area.48
48
Note: Solid waste disposal methods are evaluated at the household level. The following options are
listed: public dumpsite, collected, burned by household, dumped elsewhere, buried by household, or other.
For clarity, enumeration area boundaries are not shown. The bottom frame (alternative methods) includes
households that bury or burn trash, or dump their trash elsewhere.
2
Figure 4-17: Methods of solid waste disposal in the four neighborhoods, according to the 2000 census
(left) and the 2007 household surveys (right).
have increased substantially in East Legon during the intervening years between the two
neighborhood. Once again the survey shows that, particularly in East Legon, multiple
methods of waste disposal were used by some households; in particular, while waste
collection was used by many households, the same households also used alternative
Household methods of liquid waste that were assessed in the 2000 census in
Accra included disposal via sewage system (12.4% of households), waste thrown on
street (16.3%), waste thrown in gutter (51.9%), waste thrown in the household compound
(16.3%), and other (0.7%). Thus, a remarkably small number of houses were serviced by
2
a sewage system as of 2000, and the majority of households instead relied on open gutters
that run alongside most city streets. These gutter systems, intended to manage rainwater
runoff during the rainy season, feed into larger “drains” across the city that eventually
The toilet facilities used by households also reflect this lack of sewage systems in
most of the city. Only 22.1% of houses reported using a flush toilet located in their
house, while 9.1% of households said they used a flush toilet located in another house.
used traditional pit toilets (6.1%), Kumasi Ventilated Improved Pits (11.5%), a
project is the case of Salvador, Brazil, a city similar in size to Accra. In 1997, an
intervention was planned for the city which aimed to increase the sewerage coverage
from 26% to 80% of households. The effects of the intervention were assessed using two
longitudinal studies of diarrheal morbidity in children under three years of age, one prior
to the intervention and one after its completion. The researchers found that diarrheal
prevalence was reduced by 21% (95% CI 18-25%), and that after adjustment for baseline
sewerage coverage and confounding variables, an overall reduction of 22% (19-26%) was
found. In areas where initial prevalence was highest, declines of more than 40% (39-
49
2
Figure 4-18: Liquid waste disposal methods by EA in the Greater Accra Metropolitan Area.50
50
Note: Liquid waste disposal methods are evaluated at the household level. The following options are
listed: sewerage system, dumped in gutter, dumped in compound, dumped in street, or other. For clarity,
enumeration area boundaries are not shown. The bottom frame (alternative methods) includes households
that dump liquid waste in the street or in their compound.
2
Analysis of the census data indicates that construction of sewage systems in Accra
has been limited to wealthier areas of the central city such as Cantonments, Airport
Residential, and, to an extent, Asylum Down. Within the city, and particularly in poorer
areas, most households disposed of waste into open drainage gutters. Alternative
methods, such as dumping in the compound or street, were used in the peripheral areas of
the city which lack infrastructural developments such as the construction of gutters or a
sewage system.
Both the census data and the 2007 survey reflect the prominence of dumping in
gutters as a primary method of liquid waste disposal in Accra, particularly in poorer areas
such and Nima and Jamestown. The fact that none of the households in Asylum Down
sampled in the 2007 survey indicated that they used a sewer for liquid waste disposal is
likely due to the small sample size in this survey as well as the heterogeneous nature of
sewer coverage in Asylum Down. The EAs in Asylum Down where households were
sampled in the 2007 survey did not include EAs with high levels of sewer coverage,
although parts of the area did, as the census indicated, have relatively high numbers of
households connected to sewer systems. Similarly, the apparent increase in the number
of households with sewer systems in East Legon and decrease in households dumping in
their compound between the 2000 census and the 2007 survey could reflect either the
infrastructural development of this area in these years or the fact that these households
represent only a small sample from a few EAs within the locality.
2
Figure 4-19: Liquid waste disposal methods in the four neighborhoods, according to the 2000 census
(left) and the 2007 household surveys (right).
neighborhood (Figure 4-10); the vast majority of households in Nima and Jamestown
relied on public facilities and more households had access to flush facilities in Asylum
Down and East Legon. In these wealthier neighborhoods, the surveys conducted in 2007
show increases in the percentage of households with access to a flush toilet, particularly
in East Legon, although this may be due to limited sampling. Notably, it does not appear
that any improvements have been made in either Jamestown or Nima, where all but a
2
Figure 4-20: Toilet facilities in the four neighborhoods, according to the 2000 census (left) and the
2007 household surveys (right).
2
Chapter 5: The Spatial Distribution of Littered Solid Waste
developing cities. Waste is generated in smaller quantities in the developing world than
in the developed world, and within developing cities poorer residents produce less
waste.51 Despite lower waste generation, there is often more littered waste in developing
cities, especially in poorer neighborhoods, because the infrastructure for waste removal is
limited.52 Poor households may have to resort to less effective alternatives, such as
other pests, clogging drainage systems, and creating cesspools which then act as breeding
grounds for mosquitoes that spread malaria and dengue.53 This study examines some of
the spatial patterns of littered waste in Accra, using original data collected in July and
hierarchical model are used to estimate effects at the enumeration area (EA) and
neighborhood level.
Plastic waste was by far the most common form of waste found in this study,
followed by paper and food waste items. By observation, the most common form of
littered plastic waste was the plastic water sachet, a common commercial source of
drinking water throughout the city (Figure 5-1). The density of plastic waste was higher
51
52
53
2
in the slum areas of Nima and Jamestown, and higher in East Legon than in Asylum
Down. Littered food waste was most common in Nima, which has a large open market,
and was more common in both slum communities than in either East Legon or Asylum
Down. Similarly, paper waste was less common in the wealthier communities. Metal
waste and wood/plant waste were more evenly distributed between the four
Table 5-3: Mean level of different waste types by neighborhood (ranges shown in parentheses).
Because of the ubiquity of littered plastic waste, it is the primary form of waste
millions of years, implying that it will remain in the environment perpetually unless
otherwise removed; while many other forms of littered waste, such as food and paper
products, also pose health risks to a community, these products will eventually
biodegrade. The use of plastic products on a large scale is also relatively new
phenomenon within the city of Accra, at least in part due to rising use of items such as
2
Figure 5-21: "Sachet" water is a common commercial source of drinking water in Accra (left). The
plastic sachet then becomes a common source of littered waste (right).
Thus, in order to do regression analysis, the variable was first log-transformed. Because
a value of zero is possible for plastic waste, the variable was transformed as follows:
This transformed variable was then regressed against a dummy variable for
neighborhood, yielding the results presented in Table 5-2. Because this is a log-linear
to the reference category (Asylum Down). Thus, both slum areas of Nima and
Jamestown have, on average, 78% and 88% more littered plastic waste than Asylum
Down, and these differences are significant at the 1% level. East Legon, on the other
hand, has 28% more littered plastic waste on average, but this difference was not
54
A lognormal distribution refers to the probability distribution of a random variable that is the logarithm of
a normally distributed random variable.
55
Similarly, a log-linear regression is the regression of a log-transformed dependant variable on an
untransformed (linear) independent variable.
2
Table 5-4: Regression of log1plastic on neighborhood.
There was also variance in the amount of plastic waste found at different points
plastic waste within each neighborhood, the littered waste data was mapped using the
GIS software ArcMap.56 The data were then analyzed for spatial clustering, to see if any
clustering, or autocorrelation—was calculated for three types of waste (plastic, food, and
paper) in each neighborhood. Values close to zero indicate that the data are most likely
random. Positive spatial autocorrelation values indicate that data are clustered in patterns
unlikely to be due to random chance; that is, if a particular collection point has high
levels of waste, points nearby are also likely to have high levels of waste, and visa-versa.
Negative spatial autocorrelation values indicate that neighboring data points are different
in patterns unlikely to be due to random chance, or that a point with high levels of waste
For plastic waste, there was significant clustering in Jamestown and Asylum
Down. Possible explanations for spatial clustering in Jamestown include high levels of
plastic waste found in the southernmost part of the neighborhood, which is along a beach
56
2
with an illegal dumpsite. In Asylum Down, very low levels of waste are found in the
commercial area along the Central Ring Road, where cleanup of littered waste occurs on
a daily basis. Conversely, high levels are found in the poorer central part of the area,
along the main drainage gutter. This could explain the spatial autocorrelation found in
Asylum Down. Finally, in Nima, spatial autocorrelation was detected for littered food
waste. From the map, it appears that most food waste was found either in the market area
along the main road or alongside the gutter that borders the western edge of the study
type. However, aside from the aforementioned exceptions, no waste types demonstrated
2
Figure 5-22: The spatial distribution of plastic waste in the four study areas. Only Asylum Down and
Jamestown show significant spatial autocorrelation.
2
Figure 5-23: The spatial distribution of food waste in the four study areas. Only Nima shows
significant spatial autocorrelation.
2
Figure 5-24: The spatial distribution of paper waste in the four study areas. No neighborhoods
showed significant spatial autocorrelation.
2
GIS Analysis
It is possible that both inter- and intra- neighborhood variation in plastic waste can
road surface and capacity, methods used by households for solid waste disposal or
population density. The littered waste measurements were conducted along walking
paths constructed to cover each neighborhood spatially and also to traverse a variety of
road capacities (i.e. alley, secondary and primary roads) and road surfaces (paved or dirt).
While less likely to explain differences between neighborhoods, road capacity could
used for household solid waste disposal, which vary both within and between
neighborhoods (see Chapter 4), represent another mechanism that could explain variation
in littered waste, as some methods may lead to more effective collection of waste from
the neighborhoods. Another possible explanation for the differences in littered waste
within and between neighborhoods is population density. The population density in the
slum areas of Nima and Jamestown is much higher than the density in East Legon, while
In order to analyze the spatial relationships between the littered waste data and
other demographic and environmental variables such as population density, road type,
and solid waste disposal methods, I created an overlay of relevant datasets in ArcMap.57
57
I accomplished this by joining a shapefile of all EAs in the Greater Accra Metropolitan Area (provided
courtesy of Allan Hill, Harvard School of Public Health) with a table of relevant census data, including
solid waste disposal and population density summarized for each EA. I then added a shapefile of the road
classifications along the walking path and the original trash dataset to the same data frame (provided
courtesy of Majid Ezzati, Harvard School of Public Health).
2
I used a spatial join58 to associate each trash data collection point with the relevant
EA-level data. This allowed me to associate each point of littered waste data with the
population density and the proportion of the population using a public dumpsite for solid
waste disposal for the EA in which it was located. Additionally, I used a spatial join to
associate each littered waste data point with the road type and surface at that location.
The table created from these spatial joins contains information on geographic location,
quantities of littered waste recorded, road type and surface, population density, and solid
waste disposal methods. This table was exported for further analysis using the R
As the littered waste data points were taken along roadsides, the data points are
often found at the intersection of multiple EAs with different population densities,
begging the question of which EA information is most appropriate to use. Thus, I created
calculating an area-weighted average of the population density for EAs within a certain
distance of the measure point;60 this method should address these discontinuities. The
“smoothed” population densities were then joined with the littered waste dataset.
58
A spatial join creates a table in which fields from one layer's attribute table are appended to another
layer's attribute table based on the relative locations of the features in the two layers. In this case, the data
attributed to the EA in which the trash data collection point is located is appended, as well as the road
classification for the nearest segment of the walking path.
59
60
In order to accomplish this calculation, the census shape file was converted to a raster, with population
density as the variable. A 50-meter buffer zone was defined around each littered waste data point. In these
buffers, zonal statistics were computed to find the mean population density within the buffer (see Figure
5-5).
2
Figure 5-25: Map depicting the littered waste data collection points in Asylum Down, along with the
population density raster. The red circles represent buffer zones (50 meter radius) for which an
average population density was calculated.
explain variations in littered waste within neighborhoods. For road capacity and surface,
only the coefficient for Divided Multi-Lane roads was significant at the 5% level.
However, the only road with this classification was the Central Ring Road in Asylum
Down, which had demonstrably low quantities of trash, as explained above. Thus, the
Table 5-6: Results for OLS regression of log1plastic on road capacity and surface.
2
While the coefficient on population density is very small in this regression, it can
be roughly interpreted as a 23% increase in the amount of waste between the 25% and the
75% percentiles (roughly 1 person per hectare and 430 persons per hectare, respectively)
for the dataset. While the effect of population density is statistically significant,
substantial unexplained variation in plastic littered waste remains (R2 = 0.03). The
smoothed population density created little change in the results for this regression. The
coefficient can be roughly interpreted as a 32% increase in the amount of waste between
the 25% quartile (roughly 3 persons per hectare) and the 75% quartile (roughly 463
Results for the regression on the percent households using a public dumpsite as
their method of solid waste disposal showed that the amount of waste increases when a
higher percentage of the population uses public dumpsites for waste collection. The
coefficient can be roughly interpreted as a 57% increase in the amount of waste between
the 25% quartile (13% of the population using public dumpsite) and the 75% quartile
(92% using a public dumpsite). The R2 for this regression was also low (0.06). When
population density was included in this regression (results not shown), the coefficient for
public dump decreased slightly, but remained significant at the 5 % level, while the
2
coefficient on population density was not significant. This makes sense, as these
Table 5-8: Results of the OLS regression of log1plastic on the percentage of the population using a
public dumpsite for solid waste collection.
There was significant variance in the amount of plastic waste found at different
occur at different levels of analysis—road surface and capacity are unique to a specific
data collection point, while the collection points are clustered within different EAs. Each
EA, in turn, has a population density and percentage using public waste collection, which
could also explain the level of plastic waste. Finally, the data collection points are
Thus, I chose to model the data using a hierarchical (multi-level) model. This
allows for gains in efficiency (because we are pooling across a larger dataset), but less
concern for bias if the variation between clusters (in this case EAs) is large. Bayesian
methods are ideal for this analysis, because they allow for the estimation of a relatively
large set of parameters with a relatively small dataset, without any computational
The model assumes that the dependent variable is distributed normal with linear
mean:
2
Thus, there are two independent categorical variables here, road surface and road
population density and the percentage of households using public waste collection. I
assumed vague normal priors on the slopes and gamma priors on precisions. Parameters
in this model were estimated using the JAGS program in conjunction with the R
statistical program.61
Table 5-9: Summaries of posterior densities (mean and 95% confidence interval) for important
estimated parameters.
61
This program uses a Gibbs sampler approach to create Markov Chain Monte Carlo (MCMC) simulations
that approximate the posterior densities. The chains were run for 50,000 iterations and trace plots were
used to diagnose convergence (all chains converged).
2
With the exception of public collection of waste, all estimated parameters have
confidence bounds that include zero. The mean of the posterior distribution for public
coefficients can be roughly interpreted as percent change. Thus, the coefficient can be
roughly interpreted as a 40 % increase in the amount of waste between the 25% quartile
(13% of the population using public dumpsite) and the 75% quartile (92% using a public
dumpsite). All of the coefficients at the lowest level of the hierarchy, which includes road
surface and capacity, have much smaller means and confidence intervals that contain
characterizes the differences between the four neighborhoods. This involves estimating
replicate datasets using the posteriors of the estimated parameters. If there are additional
differences at the neighborhood level that are not explained by the model estimated, we
would expect that the neighborhood means for the dataset would differ substantially from
the replicate datasets. Thus, the test statistics used were the mean of the independent
variable for each of the four neighborhoods. As can be seen below, the sample means do
not differ substantially from the distributions of the test statistics for the replicate
datasets. Thus, we can conclude that there is not a need for an additional level of
2
Figure 5-26: Histograms of this test statistic for replicate datasets sampled from the model. Lines in
red show the test statistics for the actual dataset.
Summary
It appears that patterns of littered plastic waste are not easily explained at a high
level of spatial resolution. While significant differences existed between the four study
areas, with much higher levels of waste in the two slum areas, patterns within each
households using public collection for solid waste management was predictive of higher
levels of littered plastic waste. This variable seems to adequately model differences
2
between the four neighborhoods, without the need for an additional “neighborhood
effect”.
explain the spatial patterns of littered waste. Road capacity and surface were also unable
neighborhood features such as market places, drainage gutters, and major roadways can
explain clustering and variation within neighborhoods, but this hypothesis remains to be
tested.
2
Chapter 6: Liquid Waste and Roadside Drainage Gutters
This chapter includes a spatial analysis of an original dataset that considers the
Each unit of data in this survey represents the beginning of a segment of the path
and the associated information pertaining to the gutter characteristics along that segment.
The walking paths were recorded using a GPS with sub-meter accuracy; the data were
then processed and smoothed to remove error associated with poor satellite signal when
the path was measured. The gutter data, on the other hand, were collected using a GPS
with an error of up to a few meters; therefore, while the locations recorded fall quite close
to the path, there is not an exact spatial match between the gutter data and the pre-
recorded path. The goal of the data processing procedure was to associate each gutter
data point with the closest point along the path, and then to calculate the length of the
path segment between gutter data points.62 These path segments were then used to
create maps displaying the various gutter characteristics along the path (Figures 6-2, 6-4,
and 6-5) and calculate relevant statistics for each neighborhood. The measurement error
62
Because the walking paths have points of overlap and intersection, it was also important to ensure that
each point of data collection was associated with the correct segment of the path in the correct order. To
accomplish this, each walking path was divided into two to three “analytic segments”, designed so that
each segment did not overlap itself or have any crossings. The gutter data were divided into these analytic
segments as well, so that the closest point on the path by distance could be associated with the appropriate
point on the walking path. ArcMap was then used to spatially join the datasets by associating the closest
point along the path with the gutter data. Thus, the gutter data points were effectively “snapped” to the
closest location along the path.
2
in the spatial data for the gutter measurement points—as well as the process used to
associate this data with the pre-recorded path—should not have produced any significant
Statistics in this chapter, unless otherwise noted, are expressed as the total length
total walking path length with gutters. (For example, the statement that 62.3% of gutters
in Jamestown had stagnant water means that, of the approximately 4.1 km of walking
path in Jamestown that has gutters, about 2.6 km of those gutters had stagnant water.) It
is important to note that gutter sampling was neither random nor complete, as the walking
path represents only a small subset of all roads and alleys within each neighborhood.
However, the walking paths are representative of all areas within the neighborhood; each
path covers a variety of road types within the neighborhood and traverses the entire study
area. Although both Aslyum Down and Nima contain portions of large, city-wide
drainage gutters, the analysis presented was robust to the exclusion of these larger gutters
from the sample, indicating that these gutters did not have a large influence on the results
The amount of littered waste found in the drainage gutters was estimated as a
percent coverage of the gutter, and classified in five categories: none, low (less than 10%
estimated coverage), medium (10-40% coverage), high (40-70% coverage), and clogged
(greater than 70% coverage). Overall, most gutters along the paths had low (38.5%) or
2
medium (28.2%) trash coverage. Fewer gutters were clogged (12.9%) or demonstrated
high levels of waste (11.8%). Only a small portion of gutters had no trash (7.5%).
of littered waste in open drainage gutters and the socioeconomic status of that
neighborhood. The data show that more gutters which are clogged or have high levels of
waste are found in the slum communities of Nima (42.6%) and Jamestown (36.3%). The
clogged gutters in Nima were concentrated along the main road through the marketplace,
which is heavily littered with trash and comprises a large portion of the walking path.
Overall, East Legon had the lowest levels of waste in open drainage gutters, with the
majority (65.5%) of gutters in this neighborhood having low levels of trash or no trash at
all. Asylum Down also had a majority of gutters with low or no trash (54.0%), but a
slightly higher proportion of gutters had medium levels of trash in this neighborhood.
Figure 6-27: Level of trash in open gutters summarized for the four neighborhoods.
2
Figure 6-28: Level of littered waste in gutters along the walking paths for the four study areas.
2
Water Quality and Flood Risk
Stagnant waters can act as a proxy measure for mosquito and other vector
breeding sites, which in turn increase the risk of diseases like malaria. It is not unusual to
find visible larvae in these stagnant gutters (see Figure 6-3). Thus, malaria is another risk
associated with water, sanitation and hygiene, although through pathways independent
Figure 6-29: Stagnant waters provide breeding grounds for mosquitoes and other pests (left).
Shallow gutter with high flood risk in Jamestown (right).
stagnant waters, with roughly equal portions having slow flow (19.4%), fast flow (17.4%)
and no water (20.1%). The proportion of gutters with stagnant waters differed across the
four neighborhoods, and was notably high in Jamestown (62.3%). Aslyum Down had the
lowest proportion of gutters with stagnant water flow (35.0%), and East Legon and Nima
had intermediate levels (39.6% and 42.6% respectively). One explanation for why Nima
had relatively few gutters with stagnant waters is due to the steep gradient on the western
2
side of the neighborhood; while gutters on the flat, eastern portion of the neighborhood
near the market place tended to have stagnant waters, the gutters on the steep hill near the
was classified as either transparent or opaque. While not a direct indicator of a health
risk, turbidity provides some indication of the use of open gutters for functions other than
their intended purpose of rainwater drainage from road surfaces. Gutters with opaque
waters in Accra are most frequently filled with grey, greenish, or white waters. Thus the
sullied water used for washing and cleaning in the gutter system. Indeed, during
measurement I frequently saw women disposing of wash water directly into gutters, or
witnessed opaque waters flowing out of household pipes directly into the gutter. In this
study, the percentage of gutters with opaque waters was considerably higher in both
Jamestown and Nima (80.9% and 87.6%, respectively) than in the wealthier communities
Another measure considered was what I have deemed the “flood risk” for a
particular gutter. This was measured by taking the depth of water in the gutter and
normalizing by the depth of the gutter itself. This measure was intended to approximate
the risk that water in the gutter will overflow the bounds of the constructed gutter.
Because many gutters contain fecal waste due to defecation, as well as other sewage and
waste products, flooding can spread this contamination in the neighborhood and lead to
2
Most gutters had very low flood risk according to this measurement; however,
some gutters, particularly those found in Jamestown, were shallow and less able to
capacitate their usual water flow. The “flood risk” for a neighborhood was measured by
taking the mean of flood risks for all gutters along the walking path, weighted by the
length of each gutter. According to this simple metric, the flood risk in Jamestown is
highest (0.146), making it more than twice the flood risk calculated in Asylum Down
(0.062) and three times that in East Legon (0.048). The flood risk calculated for Nima is
also high, and is nearly twice that found in East Legon (0.084).
Table 6-10: Summary of important characteristics of gutter water in the four neighborhoods.
These differences in flood risk are perhaps due to the shallow construction of
gutters in Jamestown and Nima in comparison to the gutters of Asylum Down and East
Legon. If all gutters with a width greater than one meter are excluded from calculations
(to exclude the very large drainage gutters in East Legon and Aslyum Down), it is clear
that gutters constructed in Nima and Jamestown are, on average, much smaller than those
found in Aslyum Down and East Legon. Average depths and widths are summarized in
Table 6-2. Additionally, it is possible that residents of Nima and Jamestown utilize the
drainage gutters more frequently for the disposal of household liquid wastes, resulting in
higher water levels, which would exacerbate the flood risk in these communities.
Table 6-11: Average width and depth of gutters in the four neighborhoods (in centimeters).
2
Figure 6-30: Classification of water flow in gutters along the walking paths for the four study areas.
2
Figure 6-31: Measure of flood risk for gutters along the walking paths for the four study areas.
2
Multiple Characteristics and Comparisons
The unit of analysis thus far has been the neighborhood; however, it is useful to
further, could lead to helpful policy suggestions. For example, if the accumulation of
trash in gutters is associated with stagnant water flow, then the removal of trash might act
as a means of improving water flow in street side gutters. There could be causality in the
other direction as well—fast water flow may remove trash that would otherwise
accumulate.
neighborhoods, instead of just the four considered in this study, and then considering
However, with such a limited sampling, it makes more sense to pool all the gutter data,
and consider some relationships between variables for this pooled dataset.
littered waste and water flow in the gutter; as previously discussed, stagnant water could
be caused by high levels of solid waste that clog gutters and prevent water flow. Gutters
with no water flow were excluded from this portion of the analysis. For gutters with no
trash, the proportion of fast flowing gutters is slightly higher than the proportion with
slow or stagnant flow. For low levels of waste, on the other hand, differences in water
flow were small, with a slightly higher proportion of slow and stagnant gutters. Gutters
classified as having medium trash levels had a slightly higher portion of stagnant gutters.
Notably, however, clogged gutters and gutters with high levels of trash were much more
2
likely to have stagnant waters than another water type; the proportion of clogged gutters
with stagnant waters exceeded the proportion of clogged gutters with fast flow more than
ten-fold. Thus it appears that there is a correlation between higher levels of waste and
Figure 6-32: Comparison of trash level to water flow for all measured gutters (pooled across
neighborhoods).
Higher levels of waste were associated with higher flood risk in gutters, measured
as the depth of water in the gutter divided by the total depth of the gutter. The flood risks
in gutters with high and clogged levels of waste (0.128 and 0.122, respectively) were
nearly three times flood risks for gutters with both no waste and low levels of waste
(0.045 and 0.037, respectively). Thus, there appears to be a correlation of higher trash
levels with increased flood risk. Also, increased flood risk was associated with stagnant
2
waters; the risk in gutters with stagnant waters was more than twice that of gutters with
Table 6-12: Flood risk of gutters with specified characteristics, namely, level of littered waste (left)
and water flow (right).
Summary
This analysis has demonstrated differences between the four neighborhoods in the
level of trash, the flow rate, the turbidity of water and the flood risk of drainage gutters.
In general, the poorer communities of Nima and Jamestown had a larger proportion of
gutters that were clogged or had high levels of waste than did their wealthier
counterparts. Jamestown had the highest proportion of gutters with stagnant waters and
the greatest flood risk in gutters of the four communities, while Nima had the highest
percentage of gutters with opaque waters, indicating the use of gutters for the disposal of
waste waters.
When the data are pooled together for the four communities, it is evident that
multiple characteristics are correlated; for example, gutters with higher levels of trash
also had higher flood risk and were more likely to have stagnant or slow-flowing waters.
Stagnant gutters also had, on average, higher flood risk than gutters with slow or fast
flow. While these correlations do not establish causality, they do suggest that improving
the better.
2
Chapter 7: Household Drinking Water—Source versus
Consumption
This chapter focuses on biological contamination of drinking water, that is, the
presence of organisms such as bacteria, protozoa, viruses and worms in water that is used
for direct human consumption. This biological contamination can be separated into two
categories. First, contamination can occur before the household has access to it—as
piped water travels from a treatment facility to the neighborhood or home, or when
alternatively, improvements in water quality, can occur after water is obtained by the
household from the source, as water is transported and stored prior to consumption.
neighborhood was not sufficiently random. The small sample size also leads to
considerable uncertainty surrounding the results. However, the analysis does indicate
Legon. Fewer samples were taken in East Legon because the residents of the households
2
were absent or away during the measurement period. The sampled households obtained
drinking water from a variety of sources, including bottled water, sachet water, and water
included households using either bottled or sachet water as a primary source of drinking
water. East Legon had the highest proportion of households using commercial sources
(45%), followed by Asylum Down (36%) and Nima (36%), with Jamestown having the
lowest use (17%). The most common form of commercial water available in Accra is the
bottled water that is sold by vendors throughout the city, usually costing only US$0.04-
residents told me that they did not trust the quality of sachet water because some
companies will fill sachets indiscriminately from piped sources without any tests for
drinking water, two households, both in East Legon, used bottled water, another
commercial source.
source of water as a piped source. These sources included taps inside the home, pipes in
the yard that are shared by a number of households within the compound, and public taps
used by many households in the neighborhood. Although these samples were not
sampled households do make up a rough estimate of the types of piped sources used
within each neighborhood. Notably, the majority of residents in Jamestown and Nima
2
use public standpipes as their primary water source, while households in Asylum Down
were more likely to use water that was piped into the yard or compound. Most
In Accra, existing water shortages and lack of infrastructure make water storage a
necessity for most households. Shortages in water supply do not originate from a lack of
water resources or from drought—water demand for 2020 is estimated at only 12% of
total surface waters—but rather from deficits in coverage.63 When collecting samples for
this study, households were often visited multiple times in order to obtain a source
sample, as the pipes would only have flowing water two or three times per week. Storage
time for household samples ranged between 0-7 days, with an average storage time of 1.6
days for piped sources. While not every household visited was storing water at the time
when samples were taken, every household described some means of storing water when
observed both the type of storage container and location of the storage container in each
household where water from a piped source was stored. Types of containers observed
included sealed bottles, polytanks (large, sealed, storage tanks), plastic barrels and metal
63
2
barrels. In all four neighborhoods, plastic barrels were the most common storage
containers, although Nima had a high percentage of households using metal barrels as
well. Use of bottles and polytanks was more common in the wealthier neighborhoods of
There was also evidence that households of different socioeconomic status stored
drinking water in different parts of the home or compound. Once again, in the wealthier
households often separated smaller quantities of water used for drinking from water used
for bathing and other purposes). In the poorer communities, none of the sampled
indoors (not refrigerated) versus outdoors does not appear to be as affected by changes in
socioeconomic status.
2
water prior to use, although these practices were generally uncommon in all four
neighborhoods. Washing hands prior to fetching water from the storage location was also
particularly in households where water was stored in a barrel or bucket and retrieved with
a dipping cup. However, only a very small minority of households utilized any of these
Total coliforms and E. coli are frequently used as indicators to compare biological
contamination of water samples. Coliforms are bacteria found in soils, surface waters,
and, importantly, in the guts and feces of endothermic species, including humans. While
coliforms are generally harmless, they are an indicator of the presence of enteric
coliforms can be used to more easily measure the potential risk for contracting water-
borne disease.64 Coliforms are also generally more resistant to disinfectants and other
water treatments such as boiling, and thus the absence of coliforms from water indicates
64
65
2
Total coliform count measures all coliform bacteria present in the sample,
potential bacterial contamination from soils, plants or fecal matter. Water that tests
positive for total coliforms alone is thus not necessarily contaminated specifically by
fecal sources; however, total coliforms are a standard test for water safety, as pathogens
from all of the above-mentioned sources are considered unsafe for consumption. In the
United States, standards for drinking water are based on total coliform counts, and
require a measure of zero total coliforms per 100 mL water in more than 95% of
samples.67
found in the human gut—and the subset Escherichia coli, are more specific to fecal
contamination from contact with human or animal wastes. The WHO standard for safe
drinking water calls for no detection of E. coli in a 100mL sample.68 As noted above,
most coliforms are harmless if ingested by humans. The 0157:H7 strain of E. coli is a
notable exception and has been the subject of recent media coverage. Cases of 0157:H7
infection are usually due to the consumption of undercooked meat, and are rarely
66
67
68
69
2
Figure 7-33: Classes of coliform bacteria used as indication of biological contamination in water
(Source: Washington State Department of Health, 2007).70
In this study, both E. coli and total coliforms are used to compare the quality and
potential health risk of drinking water samples. E. coli is used specifically to compare
the likelihood of fecal contamination. It is important to note that the absolute quantities of
these indicators do not necessarily indicate absolute differences in health risks, but are
contamination. Also, as mentioned in Chapter 3, the testing equipment used does not
yield reliable results when coliforms in the sample are in excess of 200 cfus. Thus,
coliform concentrations greater than 200 cfus/5mL were considered too numerous to
count (TNTC) and have been approximated in this analysis. A complete table of sampled
70
2
Contamination in Commercial Sources
individuals typically drink water directly from the container used for sale, and thus the
source sample and the consumption sample are the same. Only two households used
bottled water as a primary source of drinking water, and both of these samples contained
no total coliforms or E. coli. While the precision on these tests is not enough to
determine if this water passes WHO drinking standards, it is possible to conclude that this
water is less contaminated than most other samples, and is unlikely to suffer from the
large problems of contamination that occur during storage within the home, as discussed
considerably, was low. Importantly, no samples of sachet water detected any E. coli
sources of drinking water. Total coliforms detected in the 5mL samples ranged from 0 to
154 cfus, with a mean level of 12.7 cfus/5mL. Out of the 14 samples of sachet water, 6
had no total coliforms detected, and all samples but two had less than 10 cfus/5mL.
There was only one sample, from a household in Jamestown, with relatively high levels
of total coliform contamination (154 cfus/5mL). Thus, bottled and sachet water, while
they may not pass WHO standards for drinking water, are sources with relatively little
Previous studies in Accra have focused on the quality of piped water at the source,
considering only contamination that occurs in pipes between the water treatment facility
2
and community taps due to leaks or the presence of biofilms within the piping
city conducted by the Public Utilities Regulatory Commission in 2006 indicated that
quality varies substantially, with total coliform counts ranging between 0 and 220
whole, fecal contamination, while in excess of the WHO standards for drinking water,
was relatively low in comparison to the levels of contamination found in my study both at
The levels of contamination found in piped source samples in my study were also
relatively low (see Table 7-5). Of these samples, 33 source samples detected no E. coli
had contamination greater than 1 cfu/5mL. The mean number of E. coli detected for all
piped source samples was 1.88 cfus/5mL. It should be noted, however, that these counts
were much higher than those found in the previous Public Utilities Regulatory
coli contamination (four of these sampled households with no contamination boiled their
households had greater than 10 cfus/5mL, with one sample being denoted too numerous
to count (TNTC). The mean level of contamination for the piped consumption samples
was 6.21 cfus/5mL, but this estimate excluded the TNTC sample and is therefore an
underestimate.
71
2
Total 43 43
No contamination (0 cfus/5mL) 33 25
Low contamination (1 cfu/5mL) 8 3
Intermediate contamination (2-10 cfus/5mL) 1 8
High contamination (>10 cfus/5mL) 1 6
TNTC (>200 cfus/5mL) 0 1
Mean level of contamination (cfus/5mL) 1.88 6.21
Table 7-17: Comparisons of E. coli contamination in piped samples, at the source and at the point of
consumption.
coliform count, also greatly exceeded the contamination in found in source samples (see
Table 7-6). While the vast majority of source samples had no more than 10 cfus/5mL
(74.4%), and a large portion of consumption samples had total coliforms exceeding 200
cfus/5mL (39.5%).
The mean number of total coliforms in consumption samples was more than five
times greater than that found in the original source samples, but even this is a strict
underestimate of the difference in contamination. This mean does not include censored
samples denoted TNTC, and is therefore much lower than the true mean. If the total
coliform count of these TNTC samples is set to 200 cfus/5mL, which would be the
minimum possible number of coliforms present in the sample, the mean of total coliforms
2
in the consumption samples is 108.6 coliforms (nearly fifteen times greater than that
practices that occur after water leaves the source pipe play an in important role in the
quality of the drinking water consumed by households in Accra. When water is obtained
outside the home at a public standpipe, that water must be transported back to the
household compound and stored prior to use; contamination can occur if water is carried
in unclean or unsealed containers. Once the water reaches the household, it must be
stored until it is used, and if not properly sealed, may be contaminated by the surrounding
That the bacteriological quality of water differs between source and point-of-use
is not a new finding72; one study of 93 households in Lima, Peru indicated that
transportation and initial storage played the largest role in fecal contamination of water.
While all water sources used by the households tested negative for E. coli, more than
30% of households had positive tests for E. coli at the point of consumption in this
study.73
72
73
2
In my study, I use the difference between contamination in the consumption
sample and contamination in the source sample to determine any patterns in water quality
contamination) that can be attributed to household storage and hygiene practices. One
mentioned previously, transportation and storage prior to consumption play a large role in
determining the overall quality of water at the point of use. In particular, one hypothesis
is that the further the source is from the home, the greater the likelihood of contamination
prior to consumption.
2
Overall, seven households showed improvements in E. coli contamination at the
point of consumption relative to the source, that is to say, E. coli in the consumption
sample were less than E. coli found in the source sample. On average, households with
cfus/5mL, and only 31.3% of samples showed any deterioration in water quality. Water
piped indoors, on average, showed some further fecal contamination occurred, with a
mean increase in E. coli counts of 3.88 cfus/5mL. However, only 25% of sampled
households in this category had any increases in contamination, and 75% of households
showed no change or improvements in quality. Thus, it appears that for both water that is
piped indoors and water that is piped into the yard, further fecal contamination during
storage and transport is less of a concern. Households that bought water from a public
standpipe, however, showed, on average, an increase of 7.5 cfus/5mL, with more than
half of samples showing increases in E. coli counts between source and consumption.
Once again, this is an underestimate due to one sample being TNTC; the mean increase is
Only five samples showed any decrease in the number of total coliforms between
the source and the consumption sample. Most samples (72.1%) showed a large or TNTC
2
increase in total coliforms (see Table 7-8). The total coliform increase in this study is
consistent with the hypothesis that contamination is related to distance from the source.
The overwhelming majority of samples from public standpipes (89.5%) showed increases
in excess of 10 cfus/5mL for total coliforms between source and consumption, and more
than half showed increases that were TNTC (in excess of 200 cfus/5mL). These
differences are augmented when the TNTC samples are incorporated into the average. At
minimum, water from a source in the yard had approximately twice the contamination
found in water piped into the house, and water from a standpipe contained, on average,
more than three times the number of total coliforms found in indoor piped samples.
attributable to the specific storage and hygiene practices of the household. One example
households that stored water, drinking water was stored in a refrigerator. Only one of
these households demonstrated any increase in E. coli between the source and the
consumption sample, and even this change was small (2 cfus/5mL). Similarly, only one
2
sample had any increases in total coliform counts, although the change was large
(TNTC). Outdoor storage and indoor storage were less distinguishable in quality. While
both indoor and outdoor storage had roughly the same proportion of households that
56.7%, respectively), outdoor storage had a slightly higher mean difference in E. coli
counts.
water had substantially lower deterioration of water quality from source to consumption
when compared with samples that were stored indoors or outdoors. Only one household
with refrigerated storage showed any increase in total coliforms, although this change
was large (TNTC). When this sample is taken into account (estimated at 200 cfus/5mL),
coli counts, the differences between indoor and outdoor storage were small, with an
cfus/5mL for outdoor storage. The vast majority of samples in both categories (88.2%
and 86.7%, respectively) had increases in total coliforms between source and
2
Similarly, the contamination that occurs between source and consumption could
vary by the type of storage container. While the large majority of samples use plastic
barrels for water storage, six households, concentrated in Asylum Down and East Legon,
use sealed plastic storage tanks—known as polytanks—or plastic bottles to store water.
These sealed containers were associated with much lower levels of deterioration during
container had increases in E. coli counts between the source and consumption sample.
On the other hand, households using metal barrels and plastic barrels had large increases
in E. coli counts, with 80% of all households using metal barrels experiencing
deterioration in quality, and 58% of households using plastic barrels having deterioration
in quality. The mean differences between consumption and source samples are shown in
Table 7-10.
The results for changes in total coliforms between source and consumption
samples also demonstrate that contamination varies by storage type. Once again, no
households that used plastic bottles for water storage had increases in total coliform
counts, and only one sample from a household that used a polytank had an increase in
contamination (12 cfus/5mL). Once again, households that used metal or plastic barrels
for storage had much larger increases in total coliform contamination, with all households
with metal barrels showing increases in excess of 10 cfus/5mL, and the majority of
households with storage in plastic barrels showing increases in total coliforms that were
TNTC.
2
Metal barrel 5 4.6 121.3 (156.0)
Plastic barrel 26 6.8 (14.3) 38.8 (135.7
7-22: Mean differences in E. coli and total coliform contamination between consumption samples and
source samples for all households with piped water sources, by storage container type. Numbers in
parentheses include TNTC samples by approximating TNTC counts at the minimum 200 cfus/5mL.
between source and consumption; however, very few households sampled used these
methods. It is notable, however, that of households that boiled or filtered water prior to
use, none had any increase in E. coli or total coliform concentration between source and
in E. coli of -16.5 cfus/5mL, and a mean decrease in total coliforms of -54.0 cfus/5mL.
prior to consumption. Hand washing had no notable effect on reducing E. coli or total
Summary
consider biological contamination that exists in the household water source. However,
the results of this study indicate that in Accra it is equally, if not more important, to also
consider the effects of transportation, storage and hygiene behavior that occur after water
is obtained from the source. In particular, different water sources were associated with
water such as water bottles and sachet water contained no E. coli and only low levels of
total coliforms. Households with piped water sources also had low contamination in the
initial source samples, but the samples that were taken at the time of consumption had
2
unexpectedly high levels of both E. coli and total coliform contamination. In particular,
consumption samples from households that obtained water from public taps had very
high levels of contamination. It is important to note that commercial samples and indoor
taps were most frequent in the wealthier communities, while public standpipes were used
Infrequent and intermittent water supply in the city has led to a variety of storage
behaviors; households in this study most frequently used plastic or metal barrels for
storing drinking water, although a small subset of households in East Legon and Asylum
Down used sealed water storage tanks (polytanks) or sealed bottles for storage. Both E.
coli and total coliform counts indicate that households that used these sealed containers
had, on average, much less change in contamination between the source and point of
consumption. Similarly, households that used refrigerated storage had low levels of
measures such as boiling or filtration had an effect on the quality of water consumed,
hand washing did not have a noticeable effect. Importantly, the storage and hygiene
methods that were associated with lower levels of household contamination we found
most frequently, if not exclusively, in the wealthier communities. Thus, despite the
problems of sampling discussed earlier, there is evidence for the existence of substantial
2
Chapter 8: Discussion, Conclusions, and Suggestions for
Future Policy
impacts on human health, particularly in the developing world. The World Health
Organization estimates that 12.2% of all deaths in Ghana can be attributed to unsafe
water, sanitation and hygiene.74 Thus, improving access to adequate WSH can greatly
reduce the burden of common diseases such as diarrhea and malaria. The focus of
research in this field has generally demonstrated a “rural bias” and has been located
primarily in rural areas and at the household level. However, with rapid urbanization
risks among subpopulations in the urban community, and to work to rectify the disparities
in access to services.
While the ability to improve environmental conditions for the urban poor is
without concrete evidence on the intervention need. This thesis has shown that it is
the developing world. In particular, these risks can be researched and assessed in
resource-poor settings at relatively little cost and without the need for an on-site research
74
2
evidence that can be used to promote policies to improve the situation in the urban slums
primarily on the environmental risk factor of water, sanitation, and hygiene. In Chapter
4, I presented some important contrasts in household water supply, liquid and solid waste
management, and sanitation facilities, using data from the 2000 National Census and
household surveys conducted in 2007. This analysis showed that the slum communities
of Jamestown and Nima had markedly different profiles for these different household
attributes than did the wealthier communities of Asylum Down and East Legon.
waste because it is the predominant form of littered waste found in my research, and a
unlike organic waste products, littered plastic waste will remain perpetually unless
collected and removed. This longer life cycle makes plastic waste a particularly good
The level of plastic waste litter found differed significantly between the four
neighborhoods, with the slum areas of Jamestown and Nima having higher levels. The
percentage of households using a public dumpsite was a good predictor to explain intra-
2
causally; that is, to establish that public dumpsters are a less effective means of removing
waste from these neighborhoods. If this is the case, it might be possible to improve the
characteristics of open drainage systems (Chapter 6). The level of littered waste in the
gutters, classified by percent coverage, was higher in Nima and Jamestown, with the
lowest levels of waste in East Legon. Nima and Jamestown also had a higher percentage
of gutters that were characterized by stagnant water flow, which is an indicator of malaria
risk. A much higher percentage of gutters in these slum communities had turbid waters,
which indicates disposal of household waste water used for cooking and cleaning into the
gutter system. Finally, these communities, particularly Jamestown, had higher levels of
“flood risk”—a measurement of the capacity of the gutter system to deal with existing
While the gutter data have primarily served to explore differences between the
study areas, it is also possible to pool the data and consider patterns and correlations
between the characteristics observed. In particular, there was a correlation between the
level of littered waste in gutters and water flow; gutters with high levels of waste and
clogged gutters were most often characterized by stagnant water flow. Likewise, gutters
with higher levels of waste were also characterized by higher flood risk. Finally, gutters
with stagnant waters were associated with higher flood risk when compared to gutters
with slow or fast flowing waters. This suggests that an effort to improve one
characteristic of gutters in the poorer communities (such as reducing high trash levels)
2
could lead to improvements in other areas (for example, lowering flood risk or decreasing
stagnant waters).
on microbial contamination that occurred between the source—the tap—and the point of
the source pipes, and while contamination in the source pipes is thus relatively low, most
piped water in Accra does not meet WHO standards for drinking water.
Use of commercial sources such as water bottles and sachets appears to be one
means of reducing the risk of fecal contamination in drinking water, as none of the
commercial sources tested positive for E. coli contamination. However, this strategy
comes with its own environmental concerns relating back to solid waste management;
plastic sachets are one of the most common littered items in Accra. Similarly, households
using public taps had, on average, higher levels of E. coli and total coliform
contamination at the point of consumption, which would indicate that access to taps
closer to the household improves the quality of water consumed. Households that used
sealed storage containers such as plastic water bottles or polytanks, particularly when the
storage container was located in a refrigerator during storage, demonstrated lower levels
of contamination during storage and thus these storage practices appear to be more
effective. Finally, there is evidence that boiling water is an effective means of reducing
coliform contamination prior to consumption. Notably, the storage and hygiene practices
2
that led to lower levels of household contamination were most common in the wealthier
communities of East Legon and Asylum Down. There is no evidence that the hygiene
practices found in the poor communities of Nima and Jamestown, such as hand washing,
comparative environmental risks and burden of disease in the different localities of the
Institute (SEI) and the University of Ghana in the 1990s. While these studies were much
more comprehensive, taking into consideration all neighborhoods within the city, my
methodology has attempted to make these comparisons more quantitative and detailed.
and rating the risk they posed on a point-scale for each neighborhood. For example,
and “frequent water supply interruptions within the community” were among the 81
proxy indicators assessed in studies by Songsore et. al. Each proxy indicator was
considered by a panel of experts and given a maximum point value based on their
assessment of its relationship to disease burden (more important risks were assigned
higher point values). Each study community was then scored for each proxy indicator,
and relative environmental risk was calculated from the total score of all proxy indicator
2
values.75 According to this scale, Nima was determined to be the most environmentally
deprived residential area of Accra, and was calculated to have approximately 4.75 times
the level of environmental risk as the least deprived quintile of residential areas.76 On
second most deprived quintile, Asylum Down in the third, and East Legon in the fourth.77
to the study of environmental risks. In general, the magnitude of the disparities between
drainage gutters, etc) are smaller than those approximated in the methods of previous
studies; none of the environmental indicators that I examined demonstrated four- or five-
previously, they still clearly articulate the need for action to rectify disparities within the
urban context, and specify the targets of such actions. Furthermore, my attempt to
quantify environmental inequities has led to the observation of differences that cannot be
My research was limited by both time and resource constraints in the data
collection process. It was not possible to extend the analysis beyond the four study areas
of Jamestown, Nima, Aslyum Down, and East Legon. While these neighborhoods do
75
76
77
2
vary in socioeconomic status, and thus allow for some conclusions about environmental
inequity in Accra, the conclusions of this study would be much more robust with a larger
the patterns observed are socioeconomic, particularly when the neighborhoods also differ
the city, and types of land use). Given that I collected this data with only a small team of
field assistants over a period of less than eight weeks, I limited myself to the four
In the case of the neighborhood-level data on littered waste and drainage gutter
quality, it would also be helpful to have data collection methods that incorporated
temporal variability. It is possible that the levels of waste and the characteristics of
gutters change over time—either short term (by day) or long term (seasonally). The data
that I have collected are thus limited to a snapshot of conditions in the neighborhood.
Also, data were not collected simultaneously in each neighborhood, but rather over a
period of several weeks, and thus comparisons rest on the assumption that conditions did
not change dramatically in each neighborhood during the measurement period. However,
measures were taken to limit the effect of temporal variation on the data; for example,
measurements were not taken on days with unusual weather, such as days following
rainstorms, which occur only occasionally in the summer. Similarly, all data collection
was undertaken in the early morning, to limit the effects of any diurnal variation that
might exist. Also, the littered waste data collection was structured so that it was possible
to check for variations between measurement days; indeed, there were no significant
2
Another concern was the uncertainty due to changes in the study neighborhoods
—be they demographic, environmental, or spatial—during the time between the National
Census (2000) and my own data collection (2007-2008). While both the census data and
previous environmental research in Accra indicate that conditions in the slum community
of Nima are on par, or even worse, than those in Jamestown, both the 2007 household
surveys and my own experience on the ground indicate that environmental conditions in
organizations, such as the Federation of Youth Clubs, that have partnered with lawyers,
community members, and political leaders to actively campaign for improved conditions
have occurred in East Legon over the past decade, which could make the data of the 2000
In the case of the analysis of household water quality, an obvious difficulty was
limited testing equipment and the lack of access to a laboratory in Accra. Because of this,
field testing kits were used as a substitute. These techniques are highly imprecise, and as
used in this study, could not detect the presence of coliforms below a level of 20
cfus/5mL, and were thus denoted too numerous to count (TNTC), it would have been
For analysis, it would have also been helpful to have a more comprehensive set of
household water sources, methods of solid and liquid waste disposal, and sanitation.
2
However, other spatial information, such as land use and density of commercial versus
residential use, might have been helpful in considering patterns of littered waste. For
example, one qualitative observation I made while conducting the survey was that areas
in front of store shops tended to have lower levels of waste, perhaps due to the fact that
these areas are actively maintained by to attract customers. Similarly, areas where there
was no clear private ownership of the land—such as market places, back alleys, and
no information was available on land use, it was not possible to analyze these
the use of satellite data and remote sensing, which could be used to infer land use in the
study areas.
in much of the data. Although efforts were made to measure and classify characteristics
as accurately as possible, there was often room for error in classifications. Many
quantities, such as the level of littered waste in gutters or in the surrounding five-meter
radius, were estimated visually. However, efforts were made in the design of the research
protocol to minimize this measurement error by specifying both methods for performing
One reaction to the numerous problems of urbanization and poverty has been to
focus on rural areas in an attempt to rectify the urban bias in development; accelerating
economic development in rural areas, it is hypothesized, will raise living standards and
2
discourage rural-to-urban migration, thus “solving” the growing problems of the urban
opportunity for developing countries. While denser populations can mean concentrated
wastes and environmental risks, it is possible, with good governance, to provide services
in providing clean water, disposing solid and liquid waste, and constructing adequate
storm drainage systems.79 In turn, cities often provide a source of remittances to rural
areas, and thus simultaneously improve rural conditions.80 Studies have also shown that
positively correlated with the level of urbanization, although the causality of this
areas impedes both advocacy and programmatic progress towards this ideal. Residents in
urban slum communities experience the reality of environmental risks every day of their
lives. Many slum residents in Nima and Jamestown expressed awareness of the
detrimental effects of poor water quality and neighborhood sanitation on their health. As
has been stated by human rights advocate Chidi Anselm Odinkalu: “People are acutely
aware of the injustices inflicted upon them…What they need is a movement that channels
these frustrations into articulate demands that evoke responses from the political
process.”82 While any inhabitant of Nima or Jamestown could probably describe some of
the problems of water, sanitation and hygiene in his or her community, empirical
78
79
80
81
82
2
evidence can serve as a catalyst for action and a means of targeting limited resources to
possible that the results could place political pressure on governing bodies to put
presented in this study can act as a foundation for community organizers to elicit
government funding, or for local NGOs to press for resources from international donors.
This type of community-based approach to water and sanitation issues has been shown to
This study has also emphasized the fact that environmental inequities in the
sanitation and water quality does not require complicated technological solutions, such as
those currently under consideration to counter the effects of global climate change or to
manage the byproducts of nuclear facilities. Indeed, the estimated costs to improve the
global situation of WSH are minimal. Among the UN Millennium Development Goals is
the target to reduce by half the number of people without access to safe drinking water
and basic sanitation by the year 2015. The World Bank estimates that it would take an
additional 5 to 21 billion dollars annually to reach these goals84; even at the high end of
these estimates, this is less than 0.1% of the GDP in the United States alone.85 Improving
water, sanitation, and hygiene in developing cities, such as Accra, is both an achievable
83
84
85
2
2
Works Cited
2
Appendix A
Coliform Measurements – Testing Protocol
Samples in the study were tested for total coliforms and E. coli using the
Coliscan® Easygel® system (Micrology Laboratories #25001). This testing method uses
chromogenic substrates that react with the enzymes galactosidase and glucuronidase,
which are produced in the lactose fermentation of coliforms and E. coli, respectively.
The pre-treated petri dishes contain these chromogenic substrates, as well as inhibitors
that limit the growth of other bacteria in the sample. The dishes are incubated at room
temperature for 36-48 hours, at which point colonies that have grown on the substrate can
1.2. Collection
1. Record the Household ID and location of the sample using GPS.
2. Collect each sample in a pre-labeled and sterile collection tube (Micrology
Laboratories #TST15) using nitrile gloves.
3. Keep all samples in a cooler on ice until they are processed.
1.3. Preparation and Incubation
1. Test all samples for coliforms within 12 hours of collection.
2. Pre-label all petri dishes before testing samples according to Household ID.
3. Transfer 5.0 mL of each water sample to a bottle of Coliscan® Easygel® medium
using a sterile pipette.
4. Swirl sample and medium to mix thoroughly before pouring into pre-treated, pre-
labeled petri dish (do not shake).
5. Place all petri dishes on a level surface in a warm part of the room for 48 hours.
1.4. Inspection
1. Take digital photograph of petri dish on a white paper, including sample
identification in the photograph.
2. Count all purple/blue colonies (disregard light blue, green, or white colonies),
report as E. coli per 5.0 mL. Do not count colonies smaller than a pinhead.
3. Count all pink and purple colonies, report as coliforms per 5.0 mL. Do not count
colonies smaller than a pinhead.
1.5. Disposal
1. Place all materials in a large pan, cover with water, and boil for 45 minutes,
discard in normal trash.
86
Cite?
2
Appendix B
Compiled Table of Household Samples
Consumption Consumption Source E. Source Source Type Neighborhood Storage Type Storage Storage Hands Boiled Filtered
E. coli total coli total Location Time Washed
coliforms coliforms (days)
0 0 0 0 commercial EL
0 0 0 0 commercial EL
0 0 0 0 commercial EL
0 0 0 0 commercial EL
0 0 0 0 commercial EL
0 13 1 1 indoor EL polytank outdoor NA NA FALSE FALSE
0 0 0 0 indoor EL bottle fridge 0 NA TRUE FALSE
0 0 0 17 indoor EL bottle fridge 0 NA TRUE TRUE
1 TNTC 0 4 in yard EL plastic barrel indoor 7 FALSE FALSE FALSE
2 TNTC 0 0 in yard EL plastic barrel fridge 7 NA FALSE FALSE
32 TNTC 7 48 public tap EL plastic barrel outdoor NA FALSE FALSE FALSE
0 11 0 11 commercial AD
0 0 0 0 commercial AD
0 1 0 1 commercial AD
0 0 0 0 commercial AD
0 0 0 0 indoor AD bottle fridge 2 NA FALSE FALSE
0 5 1 9 indoor AD not stored not stored NA NA FALSE FALSE
0 0 0 0 indoor AD polytank indoor 0 NA TRUE FALSE
32 TNTC 0 16 indoor AD plastic barrel indoor 1 FALSE FALSE FALSE
26 TNTC 1 2 in yard AD plastic barrel indoor 1 TRUE FALSE FALSE
0 3 1 4 in yard AD plastic barrel fridge 1 NA FALSE FALSE
Consumption Consumption Source E. Source Source Type Neighborhood Storage Type Storage Storage Hands Boiled Filtered
E. coli total coli total Location Time Washed
coliforms coliforms (days)
0 1 66 TNTC in yard AD bottle fridge 3 NA TRUE FALSE
0 153 1 83 in yard AD plastic barrel outdoor 0 0 FALSE FALSE
0 0 1 12 in yard AD plastic barrel indoor 3 0 FALSE FALSE
1 TNTC 0 1 in yard AD plastic barrel indoor 3 0 FALSE FALSE
0 102 0 38 public tap AD plastic barrel indoor 0 0 FALSE FALSE
0 5 0 5 commercial NM
0 3 0 3 in yard NM not stored not stored NA NA NA NA
0 84 0 1 in yard NM metal barrel indoor 1 TRUE FALSE FALSE
0 17 0 17 in yard NM not stored not stored NA NA FALSE FALSE
0 0 0 0 in yard NM not stored not stored NA NA NA NA
15 0 1 public tap NM metal barrel outdoor 0 TRUE FALSE FALSE
0 0 3 public tap NM plastic barrel indoor 3 FALSE FALSE FALSE
0 0 0 public tap NM plastic barrel indoor 0 FALSE FALSE FALSE
4 0 0 public tap NM plastic barrel outdoor 1 FALSE FALSE FALSE
0 55 0 0 public tap NM plastic barrel outdoor 2 NA FALSE FALSE
5 133 0 0 public tap NM metal barrel outdoor 0 FALSE FALSE FALSE
2 0 0 public tap NM NA indoor NA NA FALSE FALSE
0 154 0 154 commercial JT
0 0 0 0 commercial JT
0 6 0 6 commercial JT
1 164 0 16 indoor JT metal barrel outdoor 0 TRUE FALSE FALSE
2 TNTC 0 0 in yard JT metal barrel indoor 3 NA FALSE FALSE
0 61 0 0 in yard JT not stored not stored NA NA NA NA
0 52 0 2 in yard JT plastic barrel indoor 0 FALSE FALSE FALSE
0 45 0 1 in yard JT plastic barrel outdoor 3 FALSE FALSE FALSE
Consumption Consumption Source E. Source total Source Type Neighborhood Storage Type Storage Storage Hands Boiled Filtered
E. coli total coliforms coli coliforms Location Time Washed
(days)
0 3 1 3 public tap JT plastic barrel outdoor 3 FALSE FALSE FALSE
0 26 0 0 public tap JT plastic barrel indoor 2 FALSE FALSE FALSE
57 TNTC 0 16 public tap JT plastic barrel indoor 2 FALSE FALSE FALSE
8 TNTC 1 3 public tap JT plastic barrel outdoor 3 FALSE FALSE FALSE
TNTC TNTC 0 6 public tap JT plastic barrel outdoor 1 FALSE FALSE FALSE
0 TNTC 0 6 public tap JT plastic barrel outdoor 3 FALSE FALSE FALSE
2
14 84 0 1 public tap JT plastic barrel indoor 0 FALSE FALSE FALSE
4 88 0 1 public tap JT plastic barrel indoor 0 FALSE FALSE FALSE
3 0 0 public tap JT plastic barrel outdoor 0 TRUE FALSE FALSE
0 0 0 0 public tap JT plastic barrel outdoor 3 FALSE FALSE FALSE