ThesisPDF Available

Water demand and organization and payment schemes for solar pumping systems in emerging countries: a case study in Burkina Faso

Thesis

Water demand and organization and payment schemes for solar pumping systems in emerging countries: a case study in Burkina Faso

Abstract and Figures

Photovoltaic water pumping systems are a promising solution to improve health, education and social well-being through better water access in rural communities of emerging countries by reducing water collection duration and providing better quality water. Nevertheless, the technical solution must be accompanied by demand-driven management schemes to meet what people want and are willing to pay for, or systems end up disused. Current standardized water demand prediction methods and management practices yet do not fulfill these requirements. Localized demand evaluation method not only based on quantity of water per capita, but also on time and place of collection, time and money one is willing to spend, etc. would help to prevent system failure and enable truly demand-driven management schemes which can improve significantly water repartition. A new method to predict the average hourly utilization of a solar pump in Gogma, a rural remote village of Burkina Faso, was designed. A linear and a logistic regression-based prediction models fitted with data from a household survey, village mapping, boreholes account books and water quality tests reports were developed. The models took the village characteristics as an input and the solar pump predicted hourly load curve as an output. The household survey performed in Gogma is more exhaustive than any other found in the literature and its results are very promising yet difficult to compare as no equivalent study was found. The prediction accuracy was 77% for the logistic and 63% for the linear at the solar pump, while it was respectively 35% and 74% in the whole village. The logistic model had better results when only using distance and water quality as inputs, two easily measurable indicators, which gives it great potential. In both cases, discrepancies between the predicted and the factual load curves might be due to a lack of accuracy in the water collection time answers in the household survey. Then, three demand-driven management scenarios which use the latter prediction were proposed to improve time-saving benefits, health benefits and affordability. Their efficiency, feasibility, social acceptance and durability were assessed through structured interviews with a representative range of local key water development stakeholders. Results showed that financial incentives were not adapted to move the demand along the load curve as originally expected, as users may stop using improved sources if the water becomes too expensive. A system of restriction of uses at certain times was better accepted for this purpose. It would be very interesting to apply the same methodology to other rural communities to evaluate its generalizability.
Content may be subject to copyright.
Water demand and organization and
payment schemes for solar pumping systems
in emerging countries:
a case study in Burkina Faso
Author:
Vitali Caplain
Supervisor:
Dr. Judith Cherni
Co-supervisors:
Simon Meunier
Dr. Loïc Quéval
August 31, 2018
A thesis presented to Imperial College London in partial fulfillment of the
requirements for the degree of Master of Science in Sustainable Energy Futures
and for the Diploma of Imperial College.
Energy Futures Lab
Imperial College London
SW7 2AZ
Abstract
Photovoltaic water pumping systems are a promising solution to improve health,
education and social well-being through better water access in rural communities of emerging
countries by reducing water collection duration and providing better quality water.
Nevertheless, the technical solution must be accompanied by demand-driven management
schemes to meet what people want and are willing to pay for, or systems end up disused.
Current standardized water demand prediction methods and management practices yet do not
fulfill these requirements. Localized demand evaluation method not only based on quantity of
water per capita, but also on time and place of collection, time and money one is willing to
spend, etc. would help to prevent system failure and enable truly demand-driven management
schemes which can improve significantly water repartition.
A new method to predict the average hourly utilization of a solar pump in Gogma, a
rural remote village of Burkina Faso, was designed. A linear and a logistic regression-based
prediction models fitted with data from a household survey, village mapping, boreholes
account books and water quality tests reports were developed. The models took the village
characteristics as an input and the solar pump predicted hourly load curve as an output. The
household survey performed in Gogma is more exhaustive than any other found in the
literature and its results are very promising yet difficult to compare as no equivalent study was
found. The prediction accuracy was 77% for the logistic and 63% for the linear at the solar
pump, while it was respectively 35% and 74% in the whole village. The logistic model had
better results when only using distance and water quality as inputs, two easily measurable
indicators, which gives it great potential. In both cases, discrepancies between the predicted
and the factual load curves might be due to a lack of accuracy in the water collection time
answers in the household survey.
Then, three demand-driven management scenarios which use the latter prediction were
proposed to improve time-saving benefits, health benefits and affordability. Their efficiency,
feasibility, social acceptance and durability were assessed through structured interviews with a
representative range of local key water development stakeholders. Results showed that financial
incentives were not adapted to move the demand along the load curve as originally expected,
as users may stop using improved sources if the water becomes too expensive. A system of
restriction of uses at certain times was better accepted for this purpose.
It would be very interesting to apply the same methodology to other rural communities
to evaluate its generalizability.
iii
“The past can hurt. But the way I see it,
you can either run from it or learn from it.”
— Rafiki (1994)
Acknowledgements
I wish to thank various people for their contribution to this project:
Dr. Judith Cherni, my supervisor, for her time and guidance;
— Simon Meunier, my co-supervisor, for his enthusiastic encouragements, incredible
availability and his desire to make my thesis the best possible one;
Dr. Loïc Queval, my co-supervisor, for his patient guidance, very valuable critiques
and his constructive suggestions;
My special thanks are extended to Olivier Borne and the staff of the GeePs for their
help and availability;
Dr. Dale Manning for his selfless help on statistical correlation;
Gavin Eves for his support during the Risk Assessment process;
Arouna Darga for his involvement in the field trip organization;
Basile Darga who warmly welcomed me in his home for his trust and support;
— Séverin Darga for his willingness to give his time and to go the extra mile so
generously during my stay in Burkina Faso;
Ghislain, Achille, the Conseiller Zéba, Willy Sangha, Alasan Pierre Claver and all
the other wonderful people I met in Burkina Faso for their help;
My friends Louis, Martin and Philémon, a.k.a. La Colocscopie, for their support and
encouragement throughout the year;
And all the other people who made this thesis so enlightening and instructive.
v
Table of Contents
Abstract iii
List of Figures ix
List of Tables x
Introduction 1
1 Literature review 3
1.1 Challenges of water access ............................ 3
1.1.1 Challenges of water access are numerous ................ 3
1.1.2 There is evidence of improved water access positive impacts . . . . . 5
1.1.3 Solar water pumping technologies can tackle these challenges .... 6
1.2 Methods to quantify water demand ....................... 7
1.2.1 Definition of water demand ....................... 7
1.2.2 Precise estimate of water demand is essential ............. 8
1.2.3 Current estimation of water demand is unsatisfactory ......... 10
1.2.4 Water demand evaluation methods can be improved ......... 15
1.3 Organization and payment schemes for PVWPS ................ 17
1.3.1 The main current organization and payment scheme does not fulfill
its mission: community-based management .............. 17
1.3.2 Five alternatives organization and payment schemes to community-
based management still do not meet all requirements ......... 19
1.3.3 The private contract-based professional management scheme is a
promising solution ............................ 22
1.4 Conclusion ..................................... 24
2 Approach, Methodology and Data collection 25
2.1 Approach ..................................... 25
2.2 Introduction of the case study .......................... 26
2.3 Methodology to build the prediction model .................. 27
2.3.1 Building the prediction model ...................... 27
2.3.2 Data collection .............................. 30
2.3.3 Building the regression .......................... 31
2.3.4 Forecasting the load curve at the solar pump ............. 34
2.4 Methodology to propose and evaluate improvement scenarios to tackle the
three goals identified in section 1.3.3 ...................... 35
2.4.1 Elaborating scenarios that can tackle these three goals ........ 35
2.4.2 Evaluating the three scenarios ...................... 36
2.4.3 Data collection: structured interviews ................. 37
2.4.4 Interviews answers processing ...................... 37
vii
3 Prediction of water demand in Gogma 39
3.1 Building the regression in Gogma ........................ 39
3.1.1 Data cleaning and completion ...................... 39
3.1.2 Data selection ............................... 40
3.1.3 Data modeling .............................. 45
3.1.4 Assessment and validation of the regressions .............. 45
3.2 Forecasting the load curve of the solar pump in Gogma ............ 46
3.3 Discussion ..................................... 49
4 Evaluation of the three improvement scenarios in Gogma 51
4.1 The prediction model had trouble to give a valuable insight on the benefits
and implementability of the scenarios ...................... 51
4.2 Interviews results for the evaluation of the three scenarios .......... 52
4.2.1 The durability indicator was difficult to evaluate ........... 52
4.2.2 Results for scenario n°1......................... 52
4.2.3 Results for scenario n°2......................... 54
4.2.4 Results for scenario n°3......................... 55
4.3 Qualitative findings from the interviews .................... 57
4.3.1 The social facet of water is of very great importance ......... 57
4.3.2 Transition to private operation and maintenance: the case of Vergnet
Burkina .................................. 57
4.4 Discussion ..................................... 58
Conclusion 61
Bibliography A
Appendices H
A. Survey content (extract) .............................. H
B. Interview content .................................. M
C. Assumptions to process data from the surveys concerning quantity of water
fetched and time of water collection ....................... W
D. Predicted load curves and maps for the rainy season .............. Y
viii
List of Figures
1.1 Water consumption according to total travel time to collect water (summary
of results of water-use studies in East, West, and Southern Africa, Nicaragua,
India, Sri Lanka, and Bangladesh) ....................... 7
1.2 Example illustrating the need for a more precise evaluation of water demand 9
1.3 Typical average hourly water consumption for a family connected to the city
water distribution network for summer and winter seasons in South Africa . 11
2.1 Maps of Burkina Faso and the Centre-Est region ............... 26
2.2 Photos of the solar pump in Gogma ...................... 26
2.3 Structure of the prediction model ........................ 27
2.4 Detailed structure of the prediction model .................... 28
2.5 Data regression schematic explanation ..................... 32
3.1 Detailed structure of the prediction model (reminder) ............. 39
3.2 Scatter plots for continuous predictors ..................... 42
3.3 Box plots between continuous and categorical predictors ........... 43
3.4 Method of elimination of predictors by bivariate analysis ........... 44
3.5 Comparison between the factual load curve at the Gogma solar pump (mean
data from 14/01/2018, date the solar pump was installed, to 29/06/2018,
end of the dry season, from the data logger) and the two predicted load
curves during the dry season. .......................... 47
3.6 Comparison of the two predicted maps of Gogma linking households with
their drinking water source (linear and logistic regressions) with the factual
map from the surveys during the dry season .................. 48
4.1 Interviews quantitative results for scheme n°1 ranked from 1 (worst
performance) to 5 (best performance) ...................... 53
4.2 Interviews quantitative results for scheme n°2 ranked from 1 (worst
performance) to 5 (best performance) ...................... 54
4.3 Interviews quantitative results for scheme n°3 ranked from 1 (worst
performance) to 5 (best performance) ...................... 56
D.1 Comparison between the factual load curve at the Gogma solar pump (mean
data from 30/06/2018 (beginning of the rainy season) to 13/08/2018, from
the data logger) and the two predicted load curves during the rainy season. Y
D.2 Comparison of the two predicted maps of Gogma linking households with
their drinking water source (linear and logistic regressions) with the factual
map from the surveys during the rainy season ................. Z
ix
List of Tables
1.1 Advantages and drawbacks of photovoltaic water pumping systems . . . . . 6
1.2 Water use classification between basic water use and economic water use . . 8
1.3 Estimation of drinking water needs in the literature (per capita per day) . . 10
1.4 Comparative table of current water demand estimation methods ....... 14
1.5 List of factors that affect the choice of water source for a rural dweller . . . 16
1.6 Comparative table of alternatives schemes to community-based management 21
2.1 List of predictors chosen for this study for the elaboration of the prediction
model ....................................... 29
2.2 Three types of water sources in Gogma ..................... 30
2.3 Four possibilities for the prediction result at the new solar pump of Gogma 34
2.4 List of respondents for the interview ...................... 38
3.1 List of predictors for the elaboration of the prediction model (reminder) . . . 40
3.2 Univariate analysis for categorical predictors .................. 40
3.3 Pearson correlation coefficient between continuous predictors ......... 41
3.4 Two-way tables showing perceived water quality against factual water quality 41
3.5 Two-way tables for bivariate analysis of categorical predictors ........ 41
3.6 Assessment and validation indicators for the linear regression ........ 45
3.7 Assessment and validation indicators for the logistic regression ....... 45
3.8 List of predictors actually used in the two regressions ............. 46
x
Introduction
Photovoltaic water pumping systems are a promising solution to improve water
access in poor rural communities of developing countries which can lead to many other
enhancements such as social well-being, improved health and education. However, the
technical solution on its own will not save the day. It must come with “institutional
arrangements that are effective in providing people with the services that they want and
for which they are willing to pay. The solution will emerge from a demand-driven
approach” (The World Bank Water Demand Research Team 1993). Nonetheless,
quantifying water demand and implementing demand-driven organizational methods that
guarantee the sustainable use of the system over its lifetime remain challenging and
under-documented. A better comprehension of these two elements can allow to set up
solar water pumping projects which are better adapted to local communities.
In Burkina Faso, only 53% of low income households have access to improved water
sources1, 42% of which only have access intermittently (PMA2020 2016,2017). This can
be explained by a lack of knowledge on water demand in these areas causing a mismatch
between the devices in place and the local needs or by unsustainable organization and
payment schemes which result too often in unusable water access systems (Zilles & Fedrizzi
1999). Developing a sustainable water access in this country is thus a huge challenge of
high emergency. This thesis ambition is to understand how water demand can be assessed
when building a new solar water access point and which organization and payment schemes
can answer this demand while improving its sustainability. It relies on a solar pump pilot
project in Burkina Faso led by the GeePs (Group of electrical engineering Paris), the CEP
(Centre for Environmental Policy) of Imperial College London, the SATIE (Systèmes et
Applications des Technologies de l’Information et de l’Energie) in Paris and Dargatech (a
Burkinabé start-up) to test its methodology.
This thesis aims first, at understanding community members’ individual
preferences concerning water demand to forecast their utilization of a new solar water
source before it is built. Second, at proposing and evaluating an organization and
payment scheme adapted to this forecast utilization that caters for water demand,
maximizes water use efficiency and ensures the long-term sustainability of water access.
1An improved water source is “one that, by nature of its construction or through active intervention,
is protected from outside contamination such as a protected dug well or a borehole. (World Health
Organization 2017)
1
To achieve these goals, a literature review (chapter 1) helps to understand the
challenges linked to water access in Sub-Saharan Africa, the methods currently used by
different water development stakeholders to evaluate water demand and the existing
organization and payment schemes for solar pumps. It identifies precisely the areas where
improvement is needed for a precise prediction of water demand and the goals that more
efficient organization and payment schemes should aim at.
The methodology is then explained (chapter 2). It cuts the problem into two.
First, it proposes a method to forecast the demand for water at a new water source.
Second, it proposes scenarios to adapt existing organization and payment schemes to the
demand and evaluates their feasibility. The methodology is then applied to Gogma, a
remote rural village of Burkina Faso. Chapter 3shows and analyzes the results of the
prediction of the demand for water in Gogma. Chapter 4explains the organization and
payment scheme feasibility evaluation. Each time, a discussion (sections 3.3 and 4.4) builds
on the benefits and challenges of this approach.
2
| Chapter 1. Literature review
The literature review is divided into four sections. Section 1.1 provides
knowledge of water access challenges in developing countries, shows the importance of
improved access for all and presents how photovoltaic pumps can help address these
challenges. Section 1.2 discusses the methods currently used by different water
development stakeholders to quantify water demand in rural communities when designing
and operating a water access point. It assesses their precision, assets and drawbacks
before concluding on the need to design a more efficient water demand estimation
method. Section 1.3 evaluates currently existing organization and payment schemes for
solar pumps, assesses their viability and advantages, before drawing the outline of an
improved demand-driven organization and payment scheme. Gaps remaining for a precise
evaluation of water demand and the elaboration of a more efficient organization scheme
are detailed in section 1.4.
Section 1.1. Challenges of water access
This section’s goal is to identify the main challenges of water access in Sub-Saharan
Africa, explain the benefits of solving them and justify how a solar water pumping system
combined with an adapted organization scheme can help tackling them.
1.1.1. Challenges of water access are numerous
Seven challenges of water access have been identified from the literature and are
detailed below.
Challenge n°1: Water access deficiency — In 2011, according the OECD
(2011), 884 million people cannot access a ‘safe water supply’, defined by the United
Nations (2000) as an “adequate amount of safe drinking water located within a convenient
distance from the user’s dwelling”. The meaning of ‘adequate amount’ will be discussed
in section 1.2.‘Safe drinking water’ is “water that does not contain biological or chemical
agents directly detrimental to health” (United Nations 2000, p.71). A ‘convenient distance’
is typically 1 kilometer in rural areas (Agence Française de Développement 2011). Among
others, 30% of under five years old child deaths in developing countries could be prevented
with better water access (OECD 2011, p.14) which makes it a major challenge.
3
Challenge n°2: Lack of political support — If negative impacts of poor
water access can easily be encompassed, positive impacts of good water access are often
under-estimated. Benefits are social, environmental, economic and health-related (see
section 1.1.2). This under-estimation results in a lack of political concern and thus, of
action (OECD 2011, p.14). Furthermore, national investment plans are scarce in
developing nations (OECD 2011, p.24), so that local actors and external support agencies
end up addressing this issue with no national coordination, hence in an uncontrolled and
suboptimal way.
Challenge n°3: Shortsightedness of new investments — In the last
decades, most development agencies have claimed doing sustainable programs while
leaving this responsibility to local level communities once the pump is built (Harvey &
Reed 2007). This proved inefficient as the OECD (2011) highlighted: half of manual hand
pumps are disused in Sub-Saharan Africa. Support from the capital investor is required
for technical, managerial and administrative training of the community (Practica
Foundation 2013). They ought to involve local actors as per their capacity and available
time to lay the foundations of an initial administrative organization and make sure
someone will be committed to the water source’s durability (Fedrizzi et al. 2009).
Challenge n°4: Too long time spent collecting water — In Sub-Saharan
Africa, 29% of the population (37% in rural areas) spend more than one hour per day to
fetch clean water. The mean value of a round trip is 33 minutes for rural dwellers (UNICEF
2016). Time-saving benefits are detailed in section 1.1.2.
Challenge n°5: Affordability of water — The United Nations (n.d.) considers
that spending more than 5% of one’s revenue in water makes it unaffordable. According
the IIED (2016), poor households can need from 13 to 112% of their income in some
regions like Tanzania where water is expensive. Well thought out payment schemes are
ergo fundamental for a universally affordable water access.
Challenge n°6: Water projects are supply-driven “The realization that
effective policy and planning must take into account what the rural clients want and are
prepared to pay for” is not recent: this quote is taken from The World Bank Water
Demand Research Team report on the demand for water in rural areas dated 1993.
Oversizing a water pump makes it unnecessarily expensive and might be a danger for
groundwater resources. Undersizing it leads some households in storing water in
individual tanks which depreciates its quality while others suffer from droughts.
Demand-driven operation programs are thus mandatory to tackle this challenge.
Challenge n°7: Lack of education on water quality — Some people are
unfamiliar with assets of safe water and good sanitation. Most consumers do not know if
they have access to ‘safe’ water or not (Cairncross & Valdmanis 2004, p.772). The paucity
of such understanding results in under-investment in water access facilities and inadequate
choice of a drinking water source in remote areas (OECD 2011, p.24).
All seven challenges could be tackled with smart usage planning and payment
schemes that take into account local peculiarities and water demand. The brought benefits
of doing so are detailed in section 1.1.2.
4
1.1.2. There is evidence of improved water access positive
impacts
Basic water service1was part of the seventh Millennium Development Goal:
“Halve, by 2015, the proportion of the population without sustainable access to safe
drinking water” (Millennium Development Goals 2016) which was fulfilled before the
deadline (United Nations 2016). The sixth Sustainable Development Goal, new target,
aims among others at “achiev[ing] universal and equitable access to safe and affordable
drinking water” and “implement [ing] integrated water resources management at all levels”
(United Nations Development Programme n.d.).
Why do these goals exist? The key role of water in development was highlighted
by Clarke (2015): three quarters of the statistical variance in the Human Development
Index (HDI) is explained by poor water access or poor sanitation, which makes it the
most influential driver. The HDI assesses countries’ development with three indicators:
health (life expectancy at birth), education (expected and mean years of schooling) and
standard of living (gross national income per capita). Improved access to water and efficient
planning schemes can help improve the three because of its health and time-saving benefits
that are now detailed.
Health Benefits Causality between poor quality as well as insufficient
quantity of water and diseases such as diarrhea, trachoma, malaria or dengue fever was
established by Howard & Bartram (2003) and White et al. (1972). Water access
development therefore directly impacts the HDI health indicator.
Time-Saving Benefits This is the main positive impact from the point of
view of water users provided with a new pump (Churchill 1987, p.21–22). If the Sustainable
Development Goal focuses more on water quality than quantity, achieving this goal requires
adding new improved water sources, thus statistically lowering water collection time. As
62% of people collecting water in Sub-Saharan Africa are women and 15% are children
(United Nations 2015), new available time can be used for child-care, healthier cooking
and other income-generating activities (women) or for education purposes (children). This
impacts the HDI education and standard of living indicators. Cairncross & Valdmanis
(2004) and Whittington et al. (1990) assigned a monetary value to the time saved by
looking at how much rural dwellers pay other people to fetch water for them or how much
they spend on closer but more expensive water compared to further cheaper sources in
rural Kenya. The result is about $0.38 per hour, “very close to the average imputed wage
rate for such households” (Cairncross & Valdmanis 2004, p.773). Tackling challenge n°4 of
section 1.1.1 (“Too long time is spent collecting water”) is thus of paramount importance.
Merging both benefits, the OECD (2011) considered meeting the Millennium Goal
target generated (thanks to time reduction and decline of water-linked diseases) $84 billion
per year while only costing $12 billion. Other unquantifiable benefits include “dignity, social
status, cleanliness and overall well-being” (OECD 2011, p.16). In a nutshell, the benefits
of improving water access are colossal. Section 1.1.3 explains how solar water pumping
technologies can play a role in meeting this target.
1Abasic water service is “the drinking water coming from an improved source provided the collection
time is not more than 30 minutes for a round trip.” (World Health Organization 2017)
5
1.1.3. Solar water pumping technologies can tackle these
challenges
Motorized technologies to pump water in remote off-grid areas include
photovoltaic, wind and diesel (Campana 2015). Photovoltaic water pumping system
(PVWPS) have been used for more than forty years now (Vick & Neal 2012) and the
rapidly decreasing cost of photovoltaic modules has boosted its deployment (Foster &
Cota 2014) so that they have become the most promising solution in developing countries
for improved water access. For example, Bossyns (2013) found out PVWPS are by far
the most cost-effective solution in rural Mozambique. Table 1.1 details their advantages
and drawbacks. They are renewable energy-fueled, environmentally friendly, very reliable
during a long lifetime, standalone and enable communities to get purer water by drilling
deeper, which make them very appropriate for remote rural areas (Barlow et al. 1993).
However initial capital costs are usually important, making it difficult for rural
communities to self-finance.
Table 1.1: Advantages and drawbacks of photovoltaic water pumping systems
Advantages Drawbacks
No need for fuel (Cam,Ode) High initial capital costs (Cen,Fir)
No ‘fuel’ price fluctuation (Cam,Ode) High risk of theft (Cen)
No local pollution (Cam) Repair requires proficient technicians (Bos)
No greenhouse gas emissions (Cam) Dependency on weather conditions (Cua)
High reliability (Gho)
Long lifespan (Cua,Ode)
High flexibility (Mei)
No noise (Gho)
Sources: Bossyns (2013), Campana (2015), Centraider (2013), Cuadros et al. (2004),
Firatoglu & Yesilata (2004), Ghoneim (2006), Meier (2011), Odeh et al. (2006)
PVWPS are thus a promising solution, but as underlined in the introduction,
“institutional arrangements” are mandatory to realize their full potential. If many studies
hitherto focused on technical modeling and improvements of electrical parts’ performance
(Campana 2015,Ormsbee & Lansey 1994), section 1.2 will show little has been done to
consider water use optimization.
In this section, water access challenges have been explained, the importance of
improved access has been shown and assets of PVWPS have been presented. It is at
present necessary to understand how better knowledge of water demand can help designing
organization and payment schemes that can efficiently tackle the challenges of section 1.1.1.
6
Section 1.2. Methods to quantify water demand
This section aims at showing the state of the art of water demand estimation in
Sub-Saharan Africa. It first defines what water demand is (section 1.2.1), then explains
why precise water demand evaluation is crucial when designing and operating a water
access point (1.2.2), shows how this estimation is currently done in practice and points
out assets and drawbacks of these methods (1.2.3) and finally starts building a method to
improve current water demand estimation (1.2.4).
1.2.1. Definition of water demand
One should not confuse water needs and water demand. Water needs is an
intrinsic value depending among others on gender, age and physical activity and is usually
unknown by people. Water demand (or water consumption) refers to the needs “that
can be satisfied when taking into account supply constraints” (Bijl et al. 2016). Batteson
et al. (1998) supported by Howard & Bartram (2003) showed the difference between water
needs and water demand by linking the amount of water collected with the total travel
time (figure 1.1). Water needs should not depend on the water availability, whereas water
demand does: people collect less if collection is hard and more if collection is easy. Water
demand is what is at stake in this thesis as it corresponds to the utilization of the water
sources.
Figure 1.1: Water consumption according to total travel time to collect water (summary
of results of water-use studies in East, West, and Southern Africa, Nicaragua, India, Sri
Lanka, and Bangladesh)
(Taken from Batteson et al. (1998), p.74)
Water use in rural Sub-Saharan Africa is grouped into two main categories: basic
and economic uses (see table 1.2). This literature review focuses on demand for basic
water use, excluding economic water use and uses in hospitals, health centers, schools and
religious establishments.
7
Table 1.2: Water use classification between basic water use and economic water use
Basic water use Economic water use
Drinking (M 2017) Animals (M 2017)
Cooking (M 2017) Crops (M 2017)
Bathing (M 2017) Industry (M 2017)
Laundering (M 2017) Small scale horticulture (H&B 2003)
Construction (H&BM 2003)
Sources: Meunier (M) 2017,Howard & Bartram (H&B) 2003
1.2.2. Precise estimate of water demand is essential
One of the rationales why accurate knowledge of water demand is crucial when
building a new water pump is that designing a demand-driven operation scheme requires
to know the demand.
Secondly, a precise estimate of water demand helps reducing economic costs when
building a water pump, and thus reduces the price of water for users. Indeed, complex
models to simulate components’ efficiency under different irradiations and situations are
used, resulting in a pumping system that optimally fits the input water demand. The latter,
however, is often estimated with huge uncertainties, so that the technical calculations
turn out unnecessarily precise and costly (Narvarte 2001). Moreover, systems are usually
deliberately oversized by a factor of 2 or more to mitigate the uncertainties of the estimation
of water demand (Odeh et al. 2006). In other words, along with a PV array model (giving
power output according to location, choice of modules, time of the year, etc.), an inverter-
pump model (giving the water extraction flow according to power input and efficiencies)
and a groundwater model (giving the capacity of underground water reserves according
to water extraction), a water demand model is mandatory to design a pumping system
(Campana 2015). As underlined by Odeh et al. (2006), “at the time where enormous
efforts are exerted to raise the system efficiency by few percentages to decrease water unit
cost, it is found possible to decrease water unit cost by high percentages if the demand
pattern is well estimated”.
The third reason is a bad estimation of water demand can cause project failure.
The example of the PlayPump, described by Stellar (2010) as a “merry-go-round type
device” which pumps water as children play, seems anecdotal but is representative. It
built enthusiasm but failed shortly after implementation because demand was too big and
children had to keep playing for too long for enough water to be pumped. The same system
would maybe have worked in areas where demand was lower.
Fourthly, understanding the variations of water demand all year round allows, if
the system is sized to fit the maximum demand, to forecast how much spare energy can
be used for other purposes such as lighting, cooling or even phone charging when demand
is low (Campana 2015).
8
A very simple example may be worth a thousand words to underpin the importance
of knowledge on why, when and how much rural dwellers go fetch water to design an
effective organization and payment scheme. Figure 1.2 shows three households (A, B and
C) which have access to two water sources: a PVWPS and an unimproved source. If
one estimated each household needed 20 liters per day, one would have sized the PVWPS
accordingly. However, household A is so close he could fetch more than 20 liters, using the
water for laundry as well. B and C are so far away they may pump less at the PVWPS
and use the unimproved source to complement, and the PVWPS would thus be oversized.
Perhaps B would only go to the PVWPS if it takes him strictly one hour, that is, if there
is no queue at the PVWPS. If there is some queue at the time when B is available for
water collection, he will prefer the unimproved source. At the same time, maybe C is very
short on money, so he can only go at the PVWPS if the tariff is low otherwise he will go
at the unimproved source. If both B and C do not use the PVWPS, the latter would be
only used by A. The revenues would not be high enough to maintain it, and it would end
up disused. Moreover, if the unimproved source is only usable during the rainy season,
then B and C are forced to go the PVWPS during the dry season, or to walk even longer
distances to find another, more affordable source.
A
B
C
PVWPS
unimproved
source
30 min
5 min
5 min
Figure 1.2: Example illustrating the need for a more precise evaluation of water demand
(PVWPS: Photovoltaic Water Pumping System)
Possible scenarios are infinite, but the point is made: accurate knowledge on
population, their habits, the time and money they are willing to spend in water collection,
their preferences according to distance and quality of water, on other existing water sources
and on the influence of seasons on water levels (among others) must be acquired as part
of the water demand evaluation for each community where a water pump is installed,
otherwise it will not be adapted to the local demand and uses. A good estimation of water
demand is therefore important to design and operate a water point. Section 1.2.3 develops
how this process is currently done in practice.
9
1.2.3. Current estimation of water demand is unsatisfactory
The description of current methods to estimate water demand will enable to show
their deficiencies and understand what should specifically be improved. The United Nations
(2000), the Agence Française de Développement (2011) and Batteson et al. (1998) all
consider that a person needs 20 liters of water per day for basic water needs. Bossyns
(2013) suggested 30 liters and Gleick (1996) 50 liters. These rash estimates are hardly
convincing as they do not take into consideration any particularity: is it a man? a child?
a woman? is she pregnant? are they working hard? in harsh climate conditions? what
are the community’s socio-economic characteristics? how far is the collection point? which
uses do these estimates cover?. Fortunately, some research papers tried to take into account
some of these specificities, mainly by separating demand according to uses:
Estimates by the IPCS (1994), the World Health Organization (Howard & Bartram
2003) and White, Bradley & White (1972) for drinking water needs according to age and
gender are given in table 1.3. For cooking, estimates go from 1.5 (Katui-Katua 2002)
to 10 liters (Gleick 1996) per capita per day. Damerell et al. (2011) suggested 3-6 liters
depending on the type of food. For hygiene, basic hygiene requirements are 2–6 liters
and toilets with manual flush require 3–5 liters (Damerell et al. 2011) per capita per day.
Bathing and laundering adds about 50 liters per capita per day (Howard & Bartram 2003).
Table 1.3: Estimation of drinking water needs in the literature (per capita per day)
Source IPCS (1994)Howard & Bartram (2003)White et al. (1972)
Adult male
2.0L normal conditions 2.9L normal conditions
3.4L at 32°C 4.5L high temperatures
3.7L moderate activity 4.5L manual labor 3.0L normal conditions
Adult female
1.4L normal conditions 2.2L normal conditions 4.5L moderate activity
4.5L in the sun / 25°C
6.0L hard word / 30°C
4.5L manual labor
4.5L high temperatures
1.9L pregnant / nursing 4.8L pregnant
5.5L lactation
10 years old
child 1.0L normal conditions 1.0L normal conditions
The variance between the different estimations is large which makes them too
imprecise to appraise water demand in a village where the multiplication by the number
of people will build up the error. Indeed, water demand depends too much on local
peculiarities, so numbers differ significantly depending on where estimations are
performed.
Furthermore, water demand is a fundamentally dynamic process for three main
reasons:
10
1. The relationship between water demand and travel time showed in figure 1.1 results
in what Fedrizzi et al. (2009) called “repressed demand”: water consumption changes
one way or another when access conditions change (new pump built closer, water
cost change, period of drought, etc.). Easier access results in an increase of demand
to meet new needs or upgrade comfort.
2. Long-term variations (typically years) have demographics and climate change causes.
Demographics. Fedrizzi et al. (2009) distinguished natural demographic
growth from migrations. Demographic growth is easily predictable and results,
at least in a first approximation, in an increase in water consumption
proportional to the increase of the population. Migrations are less predictable,
driven by the unequal improvement of basic services such as water pumps.
Indeed, people will tend to move from low to high level of service areas,
resulting in an increased use of the service.
Climate change. According to the OECD (2012), climate change will
exacerbate the effects of droughts on aquifer levels causing increasing water
scarcity. Moreover, demand is higher during water stress periods, especially for
irrigation (Solomon et al. 2007,Kaushal et al. 2017). Basic water needs will
also be affected: higher risk of contamination (Macdonald et al. 2009), average
increase of water consumption (Al-Najar & Jarboo 2015) and increase of the
frequency of high-demand events (Goodchild 2003).
3. A typical range of daily variations of water demand is shown in figure 1.3. It was
obtained using water meters reading, hence it should take into account water for
drinking, cooking, washing oneself and clothes, gardening, washing cars, etc. It shows
a higher consumption around 08:00 and 20:00. Campana et al. (2013) developed a
dynamic model to size a PV array according to water demand hourly values and
tested it for irrigation water demand. It could be applied to basic water needs as
well provided that such estimation is carried; however, no research paper estimating
an hourly distribution of basic water demand in rural areas has been found.
Figure 1.3: Typical average hourly water consumption for a family connected to the city
water distribution network for summer and winter seasons in South Africa
(Taken from Coetzee et al. (2017))
Overall, anticipating these changes of the demand in time is mandatory for an
efficient planning scheme, albeit difficult as users themselves usually do not know their
repressed demand before experiencing easier access (Fedrizzi et al. 2009), migrations are
hardly determinable and no hourly estimation of basic water demand has been carried yet.
11
Other examples of operational situations where water demand was estimated are
described and analyzed below. Their analysis will enable to show their deficiencies.
Warner & Abate (2009) in their instruction paper for the elaboration of small water
access projects, stated that the evaluation of water requirements must be undertaken by
the community itself through internal discussions (p.22), but further wrote projects must
all have the capacity to deliver at least 20 liters per day per person to match their primary
needs (p.26). Having a lower threshold ruins the authenticity of their recommendation
which cannot be demand-driven if there is a minimum quantity required by someone else
than users themselves.
Dessu et al. (2014) estimated demand in the Mara River Basin (Kenya and
Tanzania) at 45 liters per day per person by accounting for an adjustment of water
availability. Notwithstandingly, if it is true that demand depends on the accessible supply
(repressed demand), basing a demand estimation on the available supply is not an
approach that can be condoned.
Bijl et al. (2016) designed a macro-model to predict the evolution of municipal
water demand, taking into account population size, activity, GDP per capita, regional
deviations and all their long-term variations. Different scenarios were built to account for
different factors. The results were compared with data from years 1971-2011 and reproduce
“quite well” the reality for all regions of the world studied. Their effort to integrate the
specificity of every area in their model is very interesting and represents a first step towards
a more accurate estimation of water demand.
Most development associations or NGOs do not describe publicly how they
evaluate water demand in a community before building a new water source, and let
people understand that the need for new water sources is so pressing that any
contributions is welcomed which is a moot point: what good would it do in the long term
to meet international targets if local users are not willing to pay for it for example?
Four main sizing softwares often used by these actors are cataloged here to show
that they are not made to design systems according to hourly variations of demand.
Information about those softwares was verified on April 1, 2018.
1. The Grundfos sizing tool only asks for a constant m3day1water demand value
to find the matching pump. The client has to check if the solution can match water
demand expectations every day considering solar radiation and other dynamic factors.
2. The same applies to Lab - Sines France, a PVWPS sizing firm specialized in
turnkey rural electrification projects in Africa.
3. PVsyst is a software for the sizing and simulation of PV systems. In its PVWPS
sizing tool, water demand must be specified as a m3day1value which can only
vary on a monthly basis. This choice is explained by PVsyst 6 Help:“Specifying
[water] needs in terms of hourly values does not make sense, as most of the time
the pumping system includes a storage for at least one day of consumption.” This
argument, supported by Bakelli et al. (2011), can be found relevant for sizing, albeit
12
the risk of oversizing it sensibly, but cannot be precise enough to design an effective
usage planning scheme. Furthermore, the software only uses a yearly mean value
of water demand to size the system; monthly values are only used for post-sizing
simulation.
4. Homer is a micro-grid modeling software. Its input is the quantity of energy
necessary for the system to work (so conversion between water demand and energy
must be computed separately) on an hourly basis. It can also account for deferrable
loads, which can model the tank/battery storage capacity (Lambert et al. 2006).
These two characteristics make this model very adaptative and precise. It still
requires the work of understanding and measuring water demand to be done
beforehand. Moreover, the conversion from water demand to energy demand is not
trivial.
Advantages and drawbacks of all the demand estimation methods listed above are
summarized in the comparative table 1.4 on page 14. The conclusion is straightforward:
no estimation meets all necessary criteria.
As shown in table 1.4, the Homer software allows hourly variations of water
demand as an input, which is encouraging, but it is still unusable as no such estimate has
been found for basic water needs. At the most, daily quantities have been evaluated
according to use, gender, age and physical activity. Only two papers took into account
the decreasing water consumption with travel time. Likewise, only two account for
seasonal variations of demand during wet and dry seasons. Overall, each precision
indicator (represented by a column in table 1.4) is met by half or less of the methods.
None of the demand estimation presented above thus go into enough detail.
Section 1.2.4 gathers useful elements to achieve a more precise evaluation of water
demand.
13
Table 1.4: Comparative table of current water demand estimation methods
Study By use By age
or gender
By working
conditions
Daily
distribution
Yearly
distribution
Long-term
variations
Distance between
house and source
Agence Française de
Développement (2011)No No No No No No No
United Nations (2000) No No No No No No No
Batteson et al. (1998) No No No No No No No
Bossyns (2013) No No No No No No No
Warner & Abate (2009) No No No No No No No
The Water Project No No No No No No No
Dessu et al. (2014) No No No No Yes No No
Katui-Katua (2002) Yes N/A N/A No No No No
Gleick (1996) Yes N/A N/A No No No No
Damerell et al. (2011) Yes N/A N/A No No No No
Bijl et al. (2016) No No Yes No No Yes No
Fedrizzi et al. (2009) No No No No No Yes Yes
White et al. (1972) Yes No Yes No No No No
IPCS (1994) Yes Yes Yes No No No No
Howard & Bartram (2003) Yes Yes Yes No No No Yes
Campana et al. (2013) N/A N/A N/A Yes Yes No No
Softwares
Grundfos N/A N/A N/A No No No N/A
Lab - Sines France N/A N/A N/A No No No N/A
PVsyst N/A N/A N/A No Yes No N/A
Homer N/A N/A N/A Yes Yes Yes N/A
14
1.2.4. Water demand evaluation methods can be improved
It was shown in the previous section that current water demand estimation
methods did not go into enough details to tackle effectively the challenges raised in
section 1.1. Thorough data gathering is required to collect all necessary information for a
comprehensive water demand evaluation. One of the main element to comprehend, aside
from the quantity of water inhabitants need, is the criteria according to which they
decide to go to a certain source rather than another (as illustrated in the example of
figure 1.2). If these criteria were fully understood, one could predict in advance where
each person would fetch water. Then, before installing a new solar pump for example, it
could be forecast who in the community will use it, and who will not. This could be the
basis for a truly demand-driven water delivery system. Some factors according to which
rural dwellers decide to use (and pay for) an improved source were set forth by The
World Bank Water Demand Research Team (1993) and grouped into three categories.
1. The social and economic influences:
Better educated households are more informed of health and social benefits
of improved water access and are willing to pay more or to do more efforts to
access an improved source.
Gender is a massive influence, but its direction is dependent on the context.
Women have stronger views on the matter than men, but they do not all think
that improved water access is better.
Minor determinants include income level, households’ religion, professional
activity as well as family size and composition.
2. The characteristics of the improved sources and its alternatives:
The cost of existing sources and of the improved source has a real influence.
The more complicated and time-costly the unimproved access, the higher the
willingness to pay for an improved access.
The lower the perceived quality of the existing sources, the higher the will
for improved access.
People are willing to pay more if the improved water point’s reliability is high
as preoccupation of being able to gather water is anxiety-provoking.
3. The perceived role of the government seems to be determinant in some areas as well.
Boone et al. (2011) also studied characteristics that affect households’ choice of
water source and concludes the collection time, defined as the time spent at the pump
to queue and collect water, and the distance from the household to the water source
are negatively correlated to the likelihood of choosing a water source. Geere et al. (2010)
underlined the negative health effects and physical injury linked to water collection, thus
adding the arduousness of this task to the list of factors.
In a nutshell, the main factors that affect the choice of water source for a rural
dweller according to the literature are summed up in table 1.5.
15
Table 1.5: List of factors that affect the choice of water source for a rural dweller
Education level
Gender
Cost of water
Perceived water quality
Reliability of water point
Collection time
Distance
Arduousness of water collection
On the whole, this section helped to understand the need for a precise and thorough
water demand evaluation when sizing or managing a PVWPS. Water demand estimation
does not only mean determining the quantity people need, but also when, for which use,
where they fetch it, how much they are ready to pay for, etc. This section also helped to
fathom more precisely the method currently used by different stakeholders to do so and to
see why they are unsatisfactory. The findings of section 1.2.4 about factors that influence
people’s behavior on water access could be the basis of a new methodology to improve
water demand evaluation, as this is a mandatory step to adapt planning and payment to
the demand. Section 1.3 now assesses currently existing organization and payment schemes
for water access and see how they are linked to water demand.
16
Section 1.3. Organization and payment schemes for
PVWPS
The management, which is defined as the process of operation,
maintenance and replacement of equipment, is one of the weak points of
programs of introduction of new technologies in rural communities.
Zilles & Fedrizzi (1999)
As seen in section 1.1.3, standalone PVWPS have become a very promising
solution in developing countries for improved water access. This literature review will
therefore now focus on organization and payment schemes for PVWPS. This section first
describes the baseline management of such systems and discusses its limitations (section
1.3.1). It then evaluates five alternatives schemes (1.3.2) that have been tested to
conclude on requirements an ‘ideal’ organization and payment scheme should fulfill.
Relevance of including knowledge of water demand in this ‘ideal’ scheme is explained and
three means to achieve so are then proposed (1.3.3).
1.3.1. The main current organization and payment scheme
does not fulfill its mission: community-based management
The main concern of the previous decades was the installation of as many water
systems as possible to ensure the largest water access. However, as more countries saw their
coverage enhanced, attention has turned more and more to their quality and sustainability,
hence the necessity of organizing teams to manage pumps and collect money, two tasks
that could not be achieved by central governments or development agencies on their own
(Nampusuor & Mathisen 2000). Local management has thus become the most commonly
used organization method and will now be detailed.
Community-based management (CBM) was created as governments and
development organizations looked more in-depth at the utilization of water access
systems (Lockwood & Smits 2011). It has become the baseline solution in Sub-Saharan
Africa. CBM relates to all forms of participatory organization where a panel of users,
called the water point committee (WPC), is formed at local-level to take decisions and
improve functionality and sustainability. Functionality informs on the condition of a
water source at a given time, sustainability on its capacity to deliver water over a long
period (Carter & Ross 2015). Both refer together to the mechanical device and to the
local-level governance necessary to operate and maintain the system (Colin & Cotton
1999). A pump can be non-functional today, but if a sustainable management plan has
been implemented, it will be repaired shortly.
Whaley & Cleaver (2017) defined the intended role of WPC as the “management,
administration, operation, maintenance and repair of the water pump”. They wrote that
before construction, attention should be devoted to the panel composition, especially as
far as gender and social stratum are concerned. Indeed, a diversified composition can
17
make water access more equitable (Ryan & Sulemani 2013). The choice of technology and
design should be discussed with the whole neighborhood. Education and sensitization are
valuable to build a positive environment for future rules enforcement (see challenge n°7 of
section 1.1.1: “Lack of education on water quality”). Among others, the feeling of ownership
which drives people in participating actively in the operation and maintenance of the pump
and taking responsibilities is important (Naiga et al. 2015,Harvey & Reed 2004,Drouin
2005). First funds should be collected and a fee schedule should be proposed. After
construction, the WPC should be responsible for fixing water tariffs, money collection,
funds management, enforcement of rules, repair work, equity of use and prevention of theft
(Chowns 2014). The WPC should meet regularly to discuss issues and report their actions
and accounts in a book. Last but not least, they should care about users’ satisfaction on
quality, quantity, accessibility and reliability (Whaley & Cleaver 2017).
However, CBM suffers from harsh criticism. They are said to meet not regularly
enough with each other and with local leaders, to be unrepresentative of the users, and in
the worst case to no longer exist because of a lost interest (Practica Foundation 2013).
Smits (2012) even writes: “Community-based management is dead”. The five main
difficulties raised by the implementation of CBM for water access are detailed below.
1. Payment failure One of the conditions for CBM success is that users
pay for operation and maintenance. Linked with the challenge n°5 of section 1.1.1
(“Affordability of water”), Smits (2012) reports a widely spread failure to pay a regular
fee in Ghana and Uganda. Same thing in Kenya (Adams 2012). When extensive repairs
are needed, contributions do not cover costs or take too much time to collect. Tertiary
International (2012) blames the deficient money collection system, the poor management
of funds by committees and the lack of sanctions for non-payment. The two main reasons
are thus unwillingness to pay and WPC inefficient financial management.
2. Lack of motivation WPC are usually made up of volunteers who
act in their spare time for the well-being of the community. Lockwood & Smits (2011)
and Bossyns (2013) defended that people’s involvement is then very dependent on the
motivation and availability of the moment. One solution could be to organize WPC rather
as a public utility with corresponding rights and duties. In this case, Baumann (2006)
advocated the need for a more precise definition of people’s role and obligations.
3. Lack of external support CBM requires external support for tasks it
cannot manage on its own such as major reparation, money collection and access to spare
parts for maintenance (WaterAid 2011,Bossyns 2013). Challenge n°2 of section 1.1.1
(“Lack of political support”) is therefore not met. Harvey & Reed (2007) proved lack of
such support can cause organization schemes’ unsustainability by showing a positive
correlation in Sub-Saharan Africa between government playing a dynamic role in
supporting communities and operational sustainability.
4. Misfit with local habits For Whaley & Cleaver (2017), water planning
must be merged with other local governance. Among others, accounting and predetermined
money collection timetable are two elements that regularly cause the system to fail because
of the misfit between implemented schemes and usual local governance schemes (Cleaver
18
& Elson 1995). For example, concerning money collection timetable, DeGabriele (2002)
raises that users do not necessarily have available funds at any time of the year, like farmers
whose revenues depend on seasons, so the timetable must take this local peculiarity into
account.
5. Individual abuse Dikoto-Wachtmeister (2000) wrote WPC can be subject
to abuse from some panel members to advantage some users. To prevent such abuses
and guarantee transparency, Schouten & Moriarty (2003) and Stawicki (2012) heavily
recommended to change panel members or organize new elections frequently.
Therefore WPC fall short of their responsibilities at many levels: payment failure,
lack of motivation, lack of external support, misfit with local habits and individual abuse.
Or, put differently, CBM was useful to provide first-time access to improved water supply
and to switch attention from the mechanical device to the institutional arrangements, but
it reaches its limits (informality and formalism) as users’ expectation increase (Moriarty
et al. 2013). Section 1.3.2 shows how these difficulties can be overcome.
1.3.2. Five alternatives organization and payment schemes
to community-based management still do not meet all
requirements
New ideas have been tested to overcome problems raised by CBM. They are
described and analyzed according to the five difficulties detailed in section 1.3.1. Five
alternative schemes were tested: WPC clusters, private ownership, private operation and
maintenance, Water Kiosk Entrepreneurs and Community Management Plus.
1. WPC clusters WPC clusters were tried in Uganda (Nekesa & Kulanyi
2012,Practica Foundation 2013) where all WPC of a district joined in a common
organization to cooperate, share ideas and benefit from economies of scale to buy spare
parts. However, in practice, this did not change the structure of the WPC, so payment
failure, lack of motivation, lack of external support and individual abuse remained.
2. Private ownership More radically, a system of private ownership was
tested in Kenya to get rid of CBM completely (Adams 2012). A private individual installs
and owns a pump. They take responsibility for operation and maintenance and ask users
for money in exchange for water. Motivation thus comes from the revenues for the owner.
Repairs are quicker as the owner will not make money while the pump is disused (Harvey
& Reed 2004) and users have to pay otherwise they cannot access water. However, the
big investment cost is a main barrier. External support is not really fostered, except if
governments or development agencies consider financing this capital intensive solution to
lower risks for private owners. Lack of skills in fund management can still be critical, so
the owner must be trained on technology and management. Individual abuse can be worse
if the owner decides to only sell water to a part of the population. Price for users can
be higher if the owner must pay off their investment quickly, so users’ willingness to pay
would decrease. Finally there is a risk of misfit if the community does not get to decide
the type and location of the pump.
19
3. Private operation and maintenance Keeping community participation
while discarding community management is possible. Private operation and maintenance
was tested in Burkina Faso and Madagascar, where communities still own the pump, but
contract out operation and maintenance to a private company (Practica Foundation 2013).
The latter signs a contract with the local government to ensure a good quality work and
external support. They are accountable for regular inspection, repair work and money
collection. Lack of skills and abuses are hence not a problem anymore, but users’ willingness
to pay still is as sanctions are apparently still not easily implementable. Motivation is there
as long as actors are paid for doing a good job. Misfit problem is not addressed and the
cost of hiring a company will surely increase water price for users.
4. Water Kiosk Entrepreneurs Keeping CBM assets while discarding CBM
itself is possible if improved water supplies are turned into a non-profit business driven by
demand. In India and Nigeria, Aqua for all, a non-profit organization, created Water
Kiosk Entrepreneurs. A private company owns pumps and assigns employees to operate
them. Like WPC clusters, experience is shared. Funds are brought together and allow
an optimized financing of all pumps in an area (bringing down prices for users). Such
organizations are likely to be supported by governments. Payment failure and motivation
are addressed the same way as private ownership or private operation and maintenance.
Like private ownership, people may not understand the concept of making water supply a
profitable activity, thus a potential misfit leading to unsustainability. Employees must be
partial to tackle abuses.
5. Community Management Plus To organize a reliable government
support, Baumann (2006) encouraged Community Management Plus, where
responsibilities are officially split between communities (operation, minor repairs, 30% of
major repairs), local government (70% of major repairs, problem resolution) and the
central government (monitoring and supply chain). Therefore, the WPC and its defects
remain but money collection becomes easier with the support of local government. This
idea has not been implemented yet (Practica Foundation 2013, p.8).
In the analysis of CBM issues, five main problems were raised: payment failure
(PF), lack of motivation (LM), external support (ES), misfit with local habits (MF) and
individual abuse (IA). Table 1.6 p.21 sums up how the presented alternatives manage to
tackle these difficulties using the analysis done in this section.
Overall, table 1.6 shows that the most convincing propositions seem to be Water
Kiosk Entrepreneurs and private operation and maintenance. They both tackle payment
failure and motivation issues by transferring the responsibility of operation to private
companies which revenues depend on the well-being of the pump. In both cases, regular
maintenance is also guaranteed which ensures the system sustainability. Non payers can
be baned easily. Focusing on the red cells of table 1.6, it appears the less convincing ones
are CBM, WPC clusters and Community Management Plus as they solve nearly none of
the raised issues. Also, what strikes in this table is that no alternative scheme manages to
tackle the misfit between users’ expectations and needs and what is delivered to them as
a service.
20
Table 1.6: Comparative table of alternatives schemes to community-based management
PF: payment failure; LM: lack of motivation; ES: external support; MF: misfit with local habits; IA: individual abuse;
CBM: community-based management; WPC: water point committee; O&M: operation and maintenance; WKE: Water Kiosk Entrepreneurs; CM+: Community Management Plus
Tackling CBM difficulties
Solution Country Advantages Drawbacks Source PF LM ES MF IA
CBM widespread
Community involved so
solution better fits to
needs.
Payment failure.
Decrease of motivation with
time.
Government support missing.
Whaley &
Cleaver
(2017)
– – – –
WPC
cluster Uganda
Cooperation and
learning between many
WPC at district level.
Economies of scale.
CBM operation problems
(money collection, fund
management, motivation, lack
external support) remain.
Practica
Foundation
(2013),
Nekesa &
Kulanyi
(2012)
– – – –
Private
ownership Kenya
High functionality and
sustainability.
Rapid repair.
Clear roles distribution.
Efficient payment
scheme.
High capital cost difficult for
private individuals.
Lack of knowledge on pump
management.
Higher prices.
Adams
(2012)+ +
Private
O&M
Burkina Faso,
Madagascar
Good maintenance.
Government support.
No skills or motivation
problem.
No incentive for payment.
Higher costs with more
intermediates.
Need for government
regulation.
Practica
Foundation
(2013)
+ + + +
WKE India,
Nigeria
Optimized operation
district-wise.
No committee required.
Government support.
Low prices.
Potential misfit in making
water a business.
Aqua for
all (n.d.)+ + + +
CM+ none yet Easy money collection.
Government support.
Lack of motivation and
potential individual abuse
remain.
Baumann
(2006)+ – + –
21
1.3.3. The private contract-based professional management
scheme is a promising solution
As table 1.6 and section 1.3.2 showed, none of the presented alternatives tackle all
the difficulties of CBM. A list of organization and payment guidelines will now be dressed
in light of what has been learned so far to sum up what is the most efficient in tackling
CBM issues.
Operation and maintenance is contracted out to a private company, which can own
the water system or not.
The private company signs a contract with the local government to ensure external
support. Local governance procedures are respected (Practica Foundation 2013).
The national government regulates, controls and facilitates the work of the private
company. Tasks are clearly split between pump managers, the local government and
the national government (Adams 2012).
The government regularly checks the good quality of the private company service.
The pump managers, helped by the local government and, if necessary, by the
national government, sanctions individual abuse if necessary.
Users pay for the operation, maintenance and if possible for depreciation and future
replacement (Practica Foundation 2013).
Money collection is performed by the private company. It is adapted to local habits,
but fund raising must chronologically precede need for money otherwise delays are
too long for maintenance (Smits 2012).
Pump managers are officially trained employees. They earn revenue for their work
(Practica Foundation 2013). They are skilled fund managers and incentives exist for
them to do a good quality work. They have authority to sanction non-payers in the
respect of local habits (Tertiary International 2012). The community trusts them
and recognizes them as legitimate in their role (Whaley & Cleaver 2017). They do
not go along with favoritism (Dikoto-Wachtmeister 2000). They have the capacity
to enforce decisions (Whaley & Cleaver 2017).
Maintenance is guaranteed by the private company (Whaley & Cleaver 2017). Users
can easily claim a need for repair to the private company or contact a government
official to report any work not done by the private company.
Users’ awareness is heightened on the need to pay for safe water and the possibilities
it presents by the private company (Practica Foundation 2013).
The private company ensures enough water is available at any time (Mala-Jetmarova
et al. 2017).
A scheme that respects these guidelines improves the way the five issues raised in
section 1.3.1 are tackled compared to the other schemes described above. Its functionality
can be assessed via interviews (Adams 2012), the efficiency of government help by the
existence of supportive policies (Whaley & Cleaver 2017), the quality of money collection
and financial management by looking at accounting records and bank account (Adams
2012), and users’ satisfaction can be appraised with interviews or surveys. Let such a
scheme be called a private contract-based professional management scheme.
22
Aprivate contract-based professional management scheme would still fail to tackle
challenges n°4(“Too long time is spent collecting water”) and n°5(“Affordability of
water”) of section 1.1.1. Time-saving benefits and health benefits developed in section 1.1.2
would not really be improved either. This scheme could thus be ameliorated by taking into
account the estimation of water demand discussed in section 1.2 to adapt the organization
to local usage and improve the issues above-mentioned. The three goals to be aimed at
are summed up in the following list:
1. Decrease time spent queuing (to address challenge n°4and time-saving benefits);
2. Make water more affordable for the population by opening opportunities of cheap
clean water (challenge n°5);
3. Use clean water for drinking and cooking preferentially (health benefits).
In this section, CBM was described and evaluated. Its five main drawbacks were
depicted, leading to the conclusion that it is not a sustainable model anymore. Five other
alternatives were raised, which all have their advantages and drawbacks. Their detailed
study allowed the formulation of a list of guidelines that a sustainable organization and
payment program should respect. A scheme that would respect these guidelines, named
private contract-based professional management scheme, would still fail to tackle some issues
raised in section 1.1. Three goals were set from the literature to be aimed at in making
this scheme more adapted to water use and improve CBM where it failed.
23
Section 1.4. Conclusion
The challenge [...] is simple to state: devise institutional arrangements
that are effective in providing people with the services that they want and for
which they are willing to pay. The solution will emerge from a demand-driven
approach.
The World Bank Water Demand Research Team (1993)
In 1993, the global community acknowledged demand-driven processes for water
access development projects in rural communities. Twenty-five years later, it still appears to
be a radically new concept as no-one really considered a serious evaluation of demand when
building a new water access point. The latter would serve two purposes: optimal sizing
of the new water source (which is not this thesis subject) and personalized organization
around the new water source.
In section 1.1, seven water access challenges were identified. It was shown they
can all be tackled thanks to a PVWPS combined with an efficient organization program.
The positive impacts of improved access on health, education and standard of living have
been proven through the health and time-saving benefits it leads to. The assets and the
promising future of PVWPS in developing countries, and especially in Sub-Saharan Africa,
were presented.
In section 1.2, it was shown how crucial it is to make a precise estimation of water
demand when sizing or managing a PVWPS: it makes technical optimization calculations
worth their cost, prevents oversized or inadequate systems and allows operational planning
to take it into account. It was shown that knowing water demand does not only mean
knowing how much liters people need, but also when, for which use, where they fetch it,
how much time and money they are willing to spend in water collection, etc. Different
methodologies used by a variety of water development stakeholders to estimate water
demand were described, but they were all imprecise enough to be satisfactory. Factors
that influence the choice of water source rural households decide to go to where put in
the limelight to prepare an improved method for evaluating water demand in a rural
community.
In section 1.3, CBM, the baseline management method for PVWPS, was described
and its five main drawbacks were highlighted. To tackle them, five alternative schemes were
evaluated. None could tackle all the difficulties spelled out for CBM. They were all supply-
driven, overlooking demand. A list of guidelines was dressed to tackle most issues raised
in the literature concerning water access, and three additional goals were set to reach an
organization and payment scheme that truly tackles all the issues raised in the literature
concerning water access.
24
| Chapter 2. Approach,
Methodology and Data
collection
Section 2.1. Approach
The literature review showed the need to create an efficient method to evaluate
precisely and thoroughly water demand in a given community to build and manage
sustainably a PVWPS. Indeed, a good knowledge of demand allows to adapt the
organization to this demand, thus improving significantly water repartition and the users’
standard of living. The literature review also concluded with a promising solution for
water access organization, the private contract-based professional management scheme
(section 1.3.3). Three goals were set to incorporate demand into this scheme, thus
helping to tackle major challenges concerning the time and money spent collecting water
and to improve time-saving and health benefits.
The problem of making a truly demand-driven organization and payment scheme
at a new solar pump was thus split into two. First design a prediction model that takes
as an input the characteristics of the village before the solar pump is built and as an
output the predicted load curve of this new pumping system. The prediction model grasps
the preferences of users and anticipates their choices. Second, propose scenarios to tackle
the three goals identified in section 1.3.3 and incorporate knowledge of water demand.
Their benefits and feasibility are tested using the prediction model to try to predict the
community’s reaction to those scenarios and on-site structured interviews to ask local
water development stakeholders’ opinion on their implementability.
The methodology for the prediction model elaboration is detailed in section 2.3.
The methodology to design and test the scenarios that fulfill the three goals is described in
section 2.4. The methodology was tested in Gogma, a rural village of Burkina Faso where
a solar pump was recently built. Results for the prediction model application to Gogma are
shown in chapter 3and results for the elaboration and testing of the scenarios in chapter 4.
Before that, Gogma, the place of case study, is presented in section 2.2.
25
Section 2.2. Introduction of the case study
Burkina Faso is a Sub-Saharan country of West Africa (figure 2.1a) with a
population of about 20 million people and a Human Development Index of 0.402, the
fourth lowest in the world in 2015 (United Nations Development Programme 2017). The
Centre-Est is one of the thirteen regions of Burkina Faso shown in red in figure 2.1b.
Gogma is a rural remote off-grid village of Centre-Est. It hosts 1100 inhabitants living
in 125 households (in December 2017). People live without electricity nor domestic water
supply. The majority lives from agriculture and has revenues below 1 USD per day.
(a) Map of the world by
Alvaro1984 (Public domain)
from Wikimedia Commons
(b) Map of Burkina Faso by Profoss,
original work by Uwe Dedering (CC BY-SA
3.0) via Wikimedia Commons
Figure 2.1: Maps of Burkina Faso and the Centre-Est region
A pilot project is underway since the beginning of 2017 in Gogma by the GeePs
(Group of electrical engineering Paris), the CEP (Centre for Environmental Policy) of
Imperial College London, the SATIE (Systèmes et Applications des Technologies de
l’Information et de l’Energie) in Paris and Dargatech (Burkinabé start-up). A solar
pump was built in December 2017 by Dargatech with a 10 m3reservoir, 750 Wc of solar
panels and a 56 m borehole (figure 2.2). Around 250 beneficiaries use it since
January 14th, 2018. It was financed by the NGO ResPublica and the association Eau Fil
du Soleil. The project aims at studying, on a technical and socio-economic perspectives,
the impacts of the solar pump on the community.
(a) (b)
Figure 2.2: Photos of the solar pump in Gogma
26
Section 2.3. Methodology to build the prediction
model
2.3.1. Building the prediction model
The prediction model was designed to take as an input the characteristics of a
community before a new solar pump is built and give as an output the predicted load
curve for this new pumping system (figure 2.3). The load curve represents the quantity of
water fetched during an average day on an hourly basis.
characteristics of
the community prediction model predicted load curve
of the new pumping system
Figure 2.3: Structure of the prediction model
This prediction process was undertaken in two steps. First, predict which
households will change their habits of water collection to go to the solar pump. Second,
evaluate how much water they will take from the solar pump and at what time of the day.
1. To predict which households will change their water collection habits, factors
that affect the probability that a household goes to a given water source were identified
in table 1.5: education, gender, cost of water, perceived quality of water, reliability of the
water sources, collection time, distance and arduousness of water collection. These factors
are as of now called predictors and are a part of the characteristics of the community
required to design the prediction model. Their means of evaluation are now explained.
The level of education was evaluated by comparing, for each household, the
perceived water quality (obtained in a survey) of a water source with its factual
quality (obtained by reading water quality lab tests reports). Therefore, only
perceived and factual water quality were kept as predictors;
Gender was put apart as it appeared too ambitious to ask in an unbiased way who,
between women or men, decide where the household should fetch water;
Reliability could be understood as the perceived reliability of a water source by
a household (obtained in a survey), or by the factual reliability that reflects the
frequency at which the water source dries up at some point in the year (obtained in
a survey as well), so two predictors were created;
Distance was calculated using the Euclidean distance, easy to calculate from GPS
coordinates. It was proven accurate enough to evaluate the distance between all
households and water sources by Nygren et al. (2016) (p.1148);
Cost of water, called price from now on as it is the cost for users, came from
account books and on-field observations (to identify free of charge water points). It
corresponds to total annual price for each household;
Collection time was asked in a survey;
Perceived arduousness of water collection was asked in a survey.
27
Ultimately, eight predictors were chosen in this study, summarized in the
table 2.1 on next page. Factual water quality, price of water source, factual reliability and
distance are objective predictors, based on demonstrable facts. Perceived water quality,
perceived reliability, collection time and arduousness of water collection are subjective
predictors, based on people’s experience and feelings. Price of water source, collection
time and distance are continuous predictors. Factual water quality, perceived water
quality, perceived reliability and arduousness are categorical predictors (typically ranked
on a 1 to nscale). Factual reliability was considered as a continuous predictor, but could
have been seen as categorical as well.
Data was collected regarding those predictors using a household survey, village
mapping, on-field observations, and by accessing account books for existing water sources
and water quality lab tests reports (see section 2.3.2 on data collection). The household
survey also served the purpose of knowing to which water source each household went to at
the time of the survey (that is, before the solar pump was installed) which is also part of the
characteristics of the community. Data was then analyzed using a linear regression and a
logistic regression (see section 2.3.3). The regressions forecast the water source destination
of each household for water collection.
2. To evaluate how much water users will take from the new solar pump and at
what time of the day, it was assumed that people who will go to the new water source
will take the same amount of water and at the same time of day as they did at the source
they went to previously. The main advantage of this approach was to only require data
on current habits of water collection. Indeed, questions such as “if you had access to a
new water source, how much water would you take there?” may be biased. This data was
collected through a household survey.
Finally, the predicted load curve of the new water source was drawn knowing which
households go to the solar pump, how much water they take and at what time (figure 2.4).
characteristics of
the community prediction model
predicted
load curve
predictors
quantity of water
time of collection
regression which households
use the solar pump
water source choice
Figure 2.4: Detailed structure of the prediction model
Data collection is presented in section 2.3.2, the elaboration of the regression in
section 2.3.3 and the application on the solar pump in section 2.3.4.
28
Table 2.1: List of predictors chosen for this study for the elaboration of the prediction model
Predictors Depends on Continuous/
Categorical Value Objective/
Subjective
Evaluation
method
Factual water quality OHousehold
VWater source binary 0 = high quality
1 = poor quality Objective Water quality lab
tests reports
Perceived water quality
of the water point
by the household
VHousehold
VWater source binary 0 = high quality
1 = poor quality Subjective Survey
Price household pays to
access water source
VHousehold
VWater source continuous total yearly cost in
local currency Objective Account books
Perceived reliability
of the water point
by the household
VHousehold
VWater source binary 0 = reliable
1 = unreliable Subjective Survey
Factual reliability
of the water point
OHousehold
VWater source continuous between 0 (reliable)
and 1 (unreliable) Objective Survey
Collection time
at water point
OHousehold
VWater source continuous
median collection
time according to all
households that use
this water point in
minutes
Subjective Survey
Distance between
household and water point
VHousehold
VWater source continuous Euclidean distance
in meters Objective GPS
Perceived arduousness of
extraction at the water
point by the household
VHousehold
VWater source trinary
0 = not arduous
1 = arduous
2 = very arduous
Subjective Survey
29
2.3.2. Data collection
Data was needed to build the inputs of the prediction model. It was collected in
Gogma using five means of data collection:
Village mapping (quantitative) – Using a GPS, the coordinates of all
households and water sources were collected in October 2017 by the GeePs, the CEP of
Imperial College London and the SATIE to compute the value of the Euclidean distance
between each household and water source. Three types of water sources were identified as
shown in table 2.2. Water sources’ type was logged as well.
Table 2.2: Three types of water sources in Gogma
Shallows A pond, small lake or river where the
water is not very deep.
Wells
A hole sunk by hand in the ground,
about 5-10 meters deep, used to collect
water by throwing a bucket into it.
Boreholes
A drilled deep hole (usually around
40-100 meters deep) where a pump is
used to raise water.
Water quality lab tests reports (quantitative) The quality of water
in all Gogma water sources was tested by the Aïna Lab (Ouagadougou, Burkina Faso)
on November 14th , 2017 upon request of the the GeePs, the CEP and the SATIE. Their
reports were used to evaluate the factual water quality of each source: if total coliforms
exceeds 10 UFC per 100 mL according to the lab test, the water quality was considered to
be poor.
Account books (quantitative) – Official documents such as boreholes money
record books were accessed in June 2018 by the CEP to collect the price each household
paid to access boreholes.
On-field observations (quantitative) – On-field observations by the CEP in
June 2018 showed all shallows and wells were free of charge.
30
Household survey (quantitative) A survey was performed in October 2017
by Dargatech, the GeePs, the CEP and the SATIE at 90 of the 125 households of Gogma to
collect data on their habits, time use, water uses, water sources preferences and opinion on
the current situation. The survey duration was about 45 minutes. The choice of households
was random. The list of questions relevant to this study is available in appendix A. In
short, data collected in this survey includes:
the composition of the household (number of men, women, children and elderly);
at which source the water is fetched for the household;
the quantity of water fetched each day there;
the time of the day when the household fetches water there;
the months of the year when the water is fetched there;
for which use among the ones listed in table 1.2 p.8is the water consumed;
the perceived water quality of this water point;
the perceived reliability of this water point;
the collection time at this water point according to this household;
the perceived arduousness of collecting water at this water point.
The factual reliability was computed from the survey by looking, for each water source, at
the households who go to this source in the rainy season but not in the dry season which
is a sign it has dried up.
Solar pump data logger (quantitative) – A data logger was installed on
Gogma’s solar pump by the GeePs, the CEP and the SATIE to measure instantaneous
flow rates from the borehole to the reservoir and from the reservoir to the taps at least
every second. This was used to compute the factual load curve of the solar pump to
compare it to the prediction.
2.3.3. Building the regression
The data collected gave a value to each triplet (household; water source; predictor)
for which data was available. It was then processed. The goal of the data processing
is to estimate the value of all the predictors for all households and all water sources.
This is done in section 2.3.3.1. From there, the regression of the prediction model is built
(sections 2.3.3.2 and 2.3.3.3). Its goal is to evaluate the probability each household has
of going to each water source, to deduce the water source each household chooses for
water collection. The goodness of fit between the regression and reality is then assessed
(sections 2.3.3.4 and 2.3.3.5).
2.3.3.1. Data cleaning and completion
Data cleaning and completion consists in making data consistent, with no
mistaken and no missing value (Ray 2016). When data was missing, it was completed
according to the general tendency detected among the other households with similar
characteristics. Outliers were deleted from the dataset if caused by a mistake in data
31
entry or data processing, or if they were a merely impossible value (e.g. 10,000 liters
collected per day). The source of the error was systematically traced back for a full
understanding. Duplicate entries were erased. Answers were standardized between
households (for example, some answered they collected 200 liters per day for the whole
household when others claimed they collected 10 liters per day per person or 500 liters
every three days). At the end of this step, a single value existed for each triplet
(household; water source; predictor).
2.3.3.2. Data selection
Data selection confronted predictors to each other to find relationships and tested
correlation hypotheses. The methodology described by Ray (2016) was adopted. First,
univariate analysis explored predictors individually. If the predictor value was the same
for all (household; water source) pairs, the predictor was discarded. Second, bivariate
analysis looked at relationships between predictors. Between two continuous predictors
Xand Y, the Pearson correlation coefficient given by Correlation =Covariance(X,Y )
V ar(X)×V ar(Y)
was used. It ranges between 1(perfect negative correlation) and +1 (perfect positive
correlation). 0means no correlation. There is no general rule to interpret the value
of the Pearson coefficient, it all depends on the context (Cohen 1997). Scatter plots
were used to graphically spot correlation between continuous predictors. Between two
categorical predictors Xand Y, two-way tables were dressed where each row is a possible
value for X, each column a possible value for Yand each cell the number of occurrence
of both happening at the same time. Between categorical and continuous predictors, box
plots showing the repartition of the continuous value for every possibility given by the
categorical predictor helped to see statistical correlation. If two predictors were too
strongly correlated, one of them was kept and the other discarded.
2.3.3.3. Data regression
The regression aimed at finding the relation between the values of the predictors
for each pair (household; water source) and the probability each household has of going
to each source. The water source each household chooses for water collection could then
be deduced by looking, for a given household, at the water source which had the highest
probability. The reasoning is shown on figure 2.5.
predictors regression
probability each
household goes to
each water source
destination of
all households
Figure 2.5: Data regression schematic explanation
32
The first idea was to use a linear regression. Considering a household hand a
water source w, the linear regression supposes the probability ρ(h, w)that hgoes to wis
a linear relation of the predictors:
ρ(h, w) = α0+
P
p=1
αp×predictorp(h, w)
where Pis the number of predictors and the αiare coefficients to be determined by the
regression. The aim was thus to find the P+ 1 values of the αicoefficients. To create the
output of the regression, ρ(h, w)was set to 1 if hgoes to wand 0 otherwise. Once the
regression was first launched with the Pcriteria, a look at the p-values gave information on
the statistical significance of each predictor. Conventionally, p-values above 0.05 suggest
the predictor has no or small influence in the regression (Craparo 2007). Corresponding
predictors were thus discarded and the regression was carried out another time.
If the linear regression was a simple way to model this problem, fixing ρ(h, w) = 0
if hdid not go to wis incorrect. Indeed, it puts all water sources where hdoes not go
on equal terms, which is not the case: nothing is known concerning the preferences of
households between all the water sources they do not go to. The only thing known is that,
for a given household, ρ(h, w)is higher for the water source where hgoes than for all the
other water sources. Therefore, a logistic regression seemed more appropriate. Indeed,
it is designed to correlate predictors with two possible outcomes, often symbolically labeled
0 and 1. Here, 0 would correspond to ‘this household does not use this water source’ and 1
‘this household uses this water source’. For a household hand a water source w, the logistic
regression supposes that the relation between the probability ρ(h, w)that the household h
goes to water source wand the Ppredictors is log-linear:
log ρ(h, w)
1ρ(h, w)=β0+
P
p=1
βp×predictorp(h, w)
where Pis the number of predictors and the βiare coefficients to be determined by the
regression. Like the linear regression, p-values were looked at to check the statistical
significance of all predictors.
Both linear and logistic regression were assessed.
2.3.3.4. Assessment of the regressions
To assess the quality of the linear regression, three indicators were used. The
sum of squares due to error (SSE)“measures the total deviation of the response values
from the fit (output) to the response values” (MathWorks 2018). The closer to 0, the
better the prediction. The coefficient of determination (R2) is a “measure of success of
predicting the dependent variable (the output) from the independent variables (the
predictors)(Nagelkerke 1991). It ranges from 0 (no fit) to 1 (perfect fit between the
prediction and the data). The root mean squared error (RMSE) is the square root of the
average of squared errors. The smaller the better the fit to data.
33
For the logistic regression, four indicators were calculated to evaluate the
goodness of fit. The deviance shows the quadratic difference between the fit and the
data (Hosmer et al. 2013). The smaller the deviance, the better the fit. The McFadden
R2is an equivalent of the coefficient of determination for logistic regression
recommended by Allison (2014). Last, the Akaike information criterion (AIC) and the
Bayesian information criterion (BIC) are measures of goodness of fit taking respectively
the number of predictors and the number of observations into consideration (Akaike 1974,
Wit et al. 2012). The smaller the AIC and BIC, the better the fit.
2.3.3.5. Validation of the regressions
For this step, each time a regression was completed, the destination of all
households were computed: the water source wwhere the household hdecides to go is
the one that maximizes ρ(h, w)for all w.
The households were then split randomly into two groups: one to do the regression
(70% of the households) and the other to validate it (30% of the households). During this
stage, for each household, either the regression predicted the good destination and it was
a success; or it did not predict the good destination and it was a failure. Dividing the
number of success by the total number of households gave the prediction success rate.
Linear and logistic regressions were launched fifty times with different validation sets to
observe the mean of the prediction success rate.
2.3.4. Forecasting the load curve at the solar pump
Both regressions were then used to predict the utilization of the solar pump in
Gogma. Gogma’s solar pump was added to the inputs, which means all corresponding
predictor values had to be evaluated as best as possible. The regressions evaluated the
probabilities of all households going to all water sources, this time with a new possibility:
the solar pump. Then, the water source destination for all households was deduced. For
each household, four cases described in table 2.3 could apply as far as the prediction was
concerned. The “solar” prediction success rate was calculated using
number of success
number of success + number of failure.
The new water source load curve could then be forecast and plotted (see figure 2.4).
It was then compared with reality, thanks to the data logger installed on Gogma’s solar
pump.
Table 2.3: Four possibilities for the prediction result at the new solar pump of Gogma
prediction reality
household goes to solar pump hh goes to solar pump Success
household goes to solar pump hh does not go to solar pump Failure
household does not go to solar pump hh goes to solar pump Failure
household does not go to solar pump hh does not go to solar pump Not relevant
34
Section 2.4. Methodology to propose and evaluate
improvement scenarios to tackle the three goals
identified in section 1.3.3
Section 1.3.3 identified three goals to improve the private contract-based professional
management scheme that already tackled most issues faced by the usual organizations
of motorized pumps. They are reminded below. These goals aimed at ameliorating it
by taking into account knowledge of water demand, thus improving time-saving benefits,
health benefits and affordability. Section 2.4.1 elaborates the scenarios that tackle these
three goals, then section 2.4.3 explains the method to evaluate them. Section 2.4.3 describes
the data collection process and section 2.4.4 how the collected data was analyzed.
1. Decrease time spent queuing;
2. Make water more affordable for the population by opening opportunities of cheap
clean water;
3. Use clean water for drinking and cooking preferentially.
2.4.1. Elaborating scenarios that can tackle these three goals
To decrease the time spent queuing, a solution is to flatten the load curve by
implementing differentiated pricing for peak time and off-peak time as is done in Australia
(PowerWater n.d.); no test of differentiation of price in Sub-Saharan Africa was found in
the literature. Given the presupposed effectiveness of economic incentives (the “repressed
demand” of Fedrizzi et al. (2009) explained in section 1.2.3), making water more expensive
at peak time and less expensive at off-peak time could translate part of the water collection
to off-peak hours thus flattening the load curve. Moreover, this action could enhance water
affordability by allowing users to go when water is cheaper. In the same vein, to assure
cleaner water is used preferentially for drinking and cooking, one could increase the price of
water at the boreholes when the water level is critically low to spread the remaining water
between as many households as possible. Or, if the economic incentive is unacceptable
to communities, it is possible to restrict the use of water to drinking and cooking only at
the boreholes in case of critically low water level or very high level of use. These three
scenarios can be summed up as follows:
1. Make water more expensive for those who come at peak time and less expensive
otherwise.
2. Make water more expensive if the borehole is nearly dry.
3. Restrict the use of water to drinking and cooking at peak time and when the borehole
is nearly dry.
They all require a precise knowledge of demand, which is what makes them truly
interesting: the literature review showed that no operation and payment scheme that
adapts to the local demand was tested. From now on, the three scenarios will be referred
to by their number in the list above.
35
2.4.2. Evaluating the three scenarios
The goal was then to evaluate these scenarios’ benefits and implementability.
Two means were used to do so. Firstly, the prediction model helped to get an insight on
the benefits and implementability of those scenarios thanks to its understanding of the
community’s preferences concerning water access. Secondly, ten structured interviews
were performed with a representative range of local key water development stakeholders
in the Centre-Est region of Burkina Faso to evaluate the efficiency,feasibility,social
acceptance and durability of the scenarios. The scenarios’ efficiency were evaluated as
their capacity to meet the three goals reminded above. The social acceptance refers to
people’s “assent to the reality of a situation, recognizing a process or condition without
attempting to change it or protest it” (Wikipedia 2018). The aim was not to compare the
three scenarios, but to evaluate them one by one. Detail on the evaluation of each
scenario is now given.
There are two reasons why scenario n°1 was proposed: flatten the water load curve
to decrease the time spent queuing, thus allowing women and children to concentrate on
other activities, and allow people who could not afford clean water before to use it during
off-peak time as it is less expensive. The efficiency of this scheme was evaluated in two
ways: it assessed whether the price variation affects the time at which people go and fetch
water or not, and whether it allows people who could not afford clean water before to fetch
clean water during off-peak time as it is less expensive. The feasibility of this scheme
hinges on the possibility to monitor the time at which users fetch water and to verify they
are allowed to do so. Finally, social acceptance refers to users’ readiness to accept the
scheme and durability to the length of the period this scheme could stay in place.
Schemes n°2 and 3 were suggested to better organize the repartition of water
during the day or during crises such as droughts, where the water level can be critically
low. Its aim is to guarantee that high quality water is used preferentially for drinking and
cooking.
Scenario n°2 is a financial incentive, its efficiency also depends on whether the
price variation affects the quantity of water that people take at those times. Feasibility
for scheme n°2 refers to the possibility of dynamically changing the price according to the
level of the water reserve. Social acceptance and durability were evaluated as well.
Scheme n°3 is not a financial incentive but a restriction. Its efficiency is
appraised in two ways: does the restriction affect the time people spend queuing? and
does the restriction allow more people to access high quality water for drinking and
cooking? Feasibility estimates whether it is possible to restrict the uses of water at
given times. Social acceptance and durability were evaluated as well.
36
2.4.3. Data collection: structured interviews
2.4.3.1. Interview format
Ten structured interviews with a representative range of local key water
development stakeholders were performed in the Centre-Est region of Burkina Faso in
June 2018 by the CEP to evaluate the benefits and feasibility of the three scenarios.
They contained closed and open questions. The full list of questions is available in
appendix B. The interview is cut into three parts. Answers analyzed in this thesis come
from part 3. The latter contains questions 43 to 76 which ask to rank the three scenarios’
efficiency, feasibility, social acceptance and durability. The ranking was always performed
on a 1 to 5 scale, considered the most appropriate in Burkina Faso, and supplemented by
open questions to explain the reasons for the marks given. The third scenario was
presented as a restriction on animals use at the pump, in order to make it clearer by
giving an example of water use restriction. Questions 77 to 79 are open questions which
let the interviewee talk freely about what has been said.
2.4.3.2. Choice of respondents
Some respondents were contacted beforehand, but most were found on the spot,
thanks to local contacts and networking. The only requirement was to have a wide variety
of respondents to get an insight from all perspectives: managers, funders, installers of
water pumps as well as policy-makers. Interviewees had to be educated enough to answer
questions about hypothetical possibilities objectively. Table 2.4 lists the ten interviewees
and their professional situation. Answers were anonymized so that respondents could
answer questions more freely.
2.4.4. Interviews answers processing
All quantitative answers were put together. Questions which asked for a 1 to 5
score were analyzed via simple frequency approach. Qualitative answers which justified
this score were processed thematically to extract reasons raised by interviewees as to their
scoring. Similarities and differences between answers were studied. Answers from open
questions were grouped by themes and were referenced or quoted.
Chapters 3and 4will now show the results of the application of this methodology
to Gogma.
37
Table 2.4: List of respondents for the interview
Name Professionnal
situation
Date and
time of
interview
Place of
interview
1Yvan Darga Manager of Dargatech
in Burkina Faso
12/06/2018,
16h00 Ouagadougou
2Koudougou Amadou
Founder of Gogalis,
design office in water
and environment
17/06/2018,
15h00 Tenkodogo
3Joanny Bambara
Director of water and
sanitation at Garango
town hall
18/06/2018,
9h30 Garango
4Ferdinand Kadoure
Director of water and
sanitation at Dakupa
and Director of a hand
pumps installation
company
19/06/2018,
11h00 Garango
5Conseiga
Passoukemanegba
Head of department at
ONEA (National
Office for Water and
Sanitation)
20/06/2018,
8h30 Garango
6Gouemkassoum Manager at Bousouma
di