Citation: Timotewos, M.T.;
Barjenbruch, M.; Behailu, B.M. The
Assessment of Climate Variables and
Geographical Distribution on
Residential Drinking Water Demand
in Ethiopia. Water 2022,14, 1722.
https://doi.org/10.3390/w14111722
Academic Editor: Marco Franchini
Received: 12 April 2022
Accepted: 24 May 2022
Published: 27 May 2022
Publisher’s Note: MDPI stays neutral
with regard to jurisdictional claims in
published maps and institutional affil-
iations.
Copyright: © 2022 by the authors.
Licensee MDPI, Basel, Switzerland.
This article is an open access article
distributed under the terms and
conditions of the Creative Commons
Attribution (CC BY) license (https://
creativecommons.org/licenses/by/
4.0/).
water
Article
The Assessment of Climate Variables and Geographical
Distribution on Residential Drinking Water Demand
in Ethiopia
Mosisa Teferi Timotewos 1,*, Matthias Barjenbruch 1and Beshah M. Behailu 2
1Department of Urban Water Management, Technical University of Berlin, Gustav-Meyer-Allee 25,
2Department of Water Supply and Environmental Engineering, Arba Minch University,
*Correspondence: [email protected] or [email protected]; Tel.: +49-15214328066
Abstract:
Water managers have increasingly shown that demand management solutions are more im-
portant than searching for alternative sources to resolve the challenges and shortages of water supply
services. This study identifies the impact of climate variables on residential water demand in three
geographically and spatially dispersed towns (Arba Minch, Ziway, and Debre Birhan) in Ethiopia.
Monthly mean temperature, relative humidity, and precipitation are analyzed using multivariate
regression models to identify and evaluate the impacts of the parameters on water consumption.
Principal component analysis (PCA) is also used to determine the dominant independent variable
affecting the rate of water consumption. Mean temperature is shown to be the dominant variable
causing the changes in water consumption in Arba Minch. The water consumption at Debre Birhan is
slightly affected by relative humidity. Analyzed climate variables do not affect the water consumption
changes at Ziway. The main findings of this paper show that geographical distribution and other
determinants are more important determinants of residential water demand. It is concluded that
the analyzed climate variables are not the dominant determinants which impact drinking water
consumption at the study sites. Thus, it is recommended to include relevant information about the
climate variables alongside other determinants in order to enhance the water management system in
evaluating and auditing water usage.
Keywords:
water demand; climate variables; multivariate regression analysis; principal compo-
nent analysis
1. Introduction
Water used as a source for municipal water supply is worldwide likely experiencing
stress from diminished water resources and greater demand [
1
]. Due to the demographic,
socio-economic, and hydrological factors, the gap between the availability and demand of
water resources is reaching a critical situation, even though the access to safe, affordable,
and reliable drinking water is a fundamental human right [
2
]. Bio Intelligence Service [
3
]
states that the existing changes to the Earth’s climate will have far-reaching effects on the
future availability of freshwater resources. This is an important challenge, because water
availability is one of the major limiting factors of the economic growth for a town [
4
]. Thus,
the implementation of successful water resource management plans necessitates in-depth
knowledge of the hydrological parameters and their spatial distribution characteristics.
Thus far the research into water management has had a limited scope of study.
Most researchers focus on residential water demand forecasting in order to manage
the currently available water sources in use and plan for future needs. They classify water
demand into two major categories: base water demand and seasonal water demand [
5
].
Water 2022,14, 1722. https://doi.org/10.3390/w14111722 https://www.mdpi.com/journal/water
Water 2022,14, 1722 2 of 20
Accordingly, researchers classify household water use as base water consumption and out-
door water consumption as seasonal water consumption. Water used indoors is generally
analyzed independent of the effects of climate variables, whereas seasonal use is influenced
by weather conditions such as temperature, evaporation, and rainfall [
6
,
7
]. However,
Gato et al.
[
7
] caution that the conclusion that indoor water use is not impacted by climate
variables overlooks certain climates where base water use may still be weather-dependent.
According to Schleich and Hillenbrand [
8
], drinking water consumption factors are clas-
sified into categories such as economic, social, and environmental, and climate variables’
impacts are grouped under the environmental category.
M Alshaikhli et al. [
9
] predicted factors affecting water consumption in Qatar by
using a multi-regression model. Correlation analysis is also used to avoid multicollinearity
between different independent variables. Population density and temperature is used
as an independent variable to identify their impact on water consumption of Qatar. The
result shows that the climate factor of temperature has a significant influence on the urban
water demand.
Shibao Lu et al. [
10
] uses a daily water consumption per capita of 38 cities in China
(2004–2014) dispersed spatially and temporally distributed to identify the major factor
which increases the residential water consumption by using multiple regression models.
From the nine independent variables that they used (annual average temperature, precipi-
tation, amount of source water available, importance of the city, population density, per
capita GDP, domestic water price, per capita housing area, and rate of using water-saving
appliances), two of them (water price and the usage rate of water-saving appliances) were
recognized as dominant influencing factors. Due to the lower price of the water, some
towns do not have any awareness of water saving habits or water management methods.
However, those parts of the country whose water prices are higher save water and use
water management methods. Because of the humid-hot climate of southern cities, the
residential water consumption is increasing tremendously.
Many related studies in Northern and Southern US and European cities show that
water consumption is positively related with temperature and evapotranspiration and
negatively related to precipitation [
11
,
12
]. Balling and Gobar [
13
] found significant correla-
tion between climate variables and residential water consumption for Phoenix, Arizona
between 1980 and 2004. Accordingly, they found that temperature and drought increase
residential water consumption, while precipitation reduces it. Hoyos and Artabe [
11
] found
that regions in Spain with lower average temperature and higher precipitation levels have
lower water consumption, while regions with higher average temperatures and lower
precipitation levels have higher water consumption. It may seem that the geographical
distribution of the cities having different climate conditions may lead to different water
consumption patterns, but all of the studies discussed above were performed in developed
cities having well-organized water infrastructures and water supply management services.
Thus, for the case of tropical sub-Saharan countries such as Ethiopia, which has spatially
distributed weather conditions with no proper water supply infrastructures and manage-
ment, the weather influences on water demands needed to be studied too. Some studies
have focused on the estimation of price and income elasticities of water demand [
14
,
15
],
while others have estimated the short-term and long-term demand management by suggest-
ing that water demand is changing because of economic and weather conditions [
16
–
18
].
However, the study of the effects of climate variables on residential water consumption
remains limited. The lack of literature regarding the influence of climate variables on
residential water demand in developing countries such as Ethiopia is one challenge for
such a study. Therefore, this paper fills the knowledge gap on water management issues in
developing areas, both in terms of policy implication and sustainable water supply services
at water-scarce cities.
The objectives of this study are (i) to evaluate the impact of hydrological factors on
domestic water demand by using multiple regression models; (ii) to identify whether
spatial variability and interannual climate difference within the country have a relation to
Water 2022,14, 1722 3 of 20
residential water consumption; and (iii) to identify the dominant variable/s which cause/s
water consumption changes.
Overview of Ethiopia
Ethiopia is characterized by high and ragged mountains, flat-topped plateaus, deep
gorges, river valleys, and plains. There are 12 major river basins, with a capacity of
annual surface water and ground water resources potential of about 122 billion m
3
and
36 billion m3
, respectively [
19
]. In spite of this immense resource potential, a considerable
proportion of the country faces uneven water distribution and discrepancy of accessibility
in terms of time and space [
20
]. Ethiopia has made enormous progress in extending access
to safe water the last two decades (1990–2020), but sustaining systems and services remains
a huge challenge. Non-governmental organizations (NGO) are also active in serving the
remote areas of Ethiopia by providing clean water supplies. NGOs and other external
agencies support about 56% of the required funds for water supply, while the government
provides only 44% [
21
,
22
]. Still, almost half of the Ethiopian population has no access to
clean water supplies nearby. In addition, weather predictions made by WHO/DFID [
23
]
indicate (Vision 2030) that average temperature and rainfall are likely to increase in East
African countries (including the Ethiopian highlands), and, consequently, there might be
increased frequency of extreme weather events which would pose negative impacts on
residential water supplies. As a result, uneven temporal and spatial distributions of water
resources and climate variability within Ethiopia is causing a stress on a residential water
supply service. Ethiopia is currently facing a problem of unreliable and insufficient water
demand across the country because of socioeconomic, demographic, and hydrological
changes. Residents of urban and rural towns of Ethiopia are receiving 15–20 L of water
per capita and day on average [
24
]. The dearth of access to clean water in the face of
tremendous potential resources adds a level of urgency to the Ethiopian case.
2. Study Area
This study was conducted in three urban towns of Ethiopia, each located at different
altitudes: Arba Minch, Ziway (1600 m), and Debre Birhan (2800 m). Traditionally there are
three climate classifications in Ethiopia. The “Kola”, which is characterized by a semi-desert
climate, is below 1500 m above sea level. The next class is called a “Weina Dega”, ranging
from 1500 to 2400 m above sea level and similar to a tropical climate. The highest class is
the “Dega”, which includes areas higher than 2400 m and commonly features a temperate
climate. Each of the case study sites is at different elevation point and characterized under
the different climatic zones stated above. Arba Minch (1300 m) is classified as a Kola; Ziway
(1600 m) is in a Weina Dega area; and Debra Birhan (2800) is a Dega site, one of the few
towns of its size at such an elevation.
Arba Minch is a city found 500 km south from the capital city Addis Abeba, located
at geographical coordinates of 6
◦
2
0
0” N 37
◦
33
0
0” E. Even though the city is categorized
as one of the hotter climate areas, it is surrounded by two big natural lakes (Abaya and
Chamo) and considered as one of the areas with a good groundwater potential zone in the
country. The total area of the city was estimated at 1095 hectares, with a total population
of 200,373. The town is located at an altitude of 1300 m above sea level with an average
annual temperature of 29 ◦C and average annual rainfall of 900 mm.
Ziway is located 163 km southeast of Addis Abeba. The town has a latitude and
longitude of 7
◦
56
0
N 38
◦
43
0
E with an elevation of 1643 m above sea level. Ziway’s annual
rainfall is 700–800 mm, and the mean annual temperature is 20
◦
C. The town has a total
population of 72,350. Ziway Lake (Hora-Dambal) has a major role in the economic activity
of the town and surrounding county, as it supplies water for the traditional irrigation
system of the community as well as the huge rose farms surrounding the lake.
Debre Birhan is one of the coldest towns in Ethiopia, located 120 km northeast of
Addis Abeba. The town has a latitude and longitude 9
◦
41
0
N 39
◦
31
0
E, with an elevation of
2840 m above sea level and a mean annual temperature of 15
◦
C. Having a mean annual
Water 2022,14, 1722 4 of 20
rainfall of 1219.2 mm makes the groundwater potential of the area a potential water supply
source for all residents of the town. The town is home to about 180,000 residents. Details of
the study area’s locations are presented in Figure 1.
Water 2022, 14, x FOR PEER REVIEW 4 of 20
of 2840 m above sea level and a mean annual temperature of 15 °C. Having a mean annual
rainfall of 1219.2 mm makes the groundwater potential of the area a potential water sup-
ply source for all residents of the town. The town is home to about 180,000 residents. De-
tails of the study area’s locations are presented in Figure 1.
Figure 1. Study area locations in Ethiopian map.
3. Data and Methodology
3.1. Climate Data
Climate data for all the study sites were collected from the National Meteorological
Agency Office of Ethiopia. Daily and monthly data of meteorological variables of the past
30 years, from 1990 to 2020, were used in order to see the trend analysis of the selected
climate variables data of the past years. Nine years of climate data (2012–2020) were used
to assess the impacts of the variables on water demand. Three climate variables were se-
lected to assess the relationship with drinking water consumption within the three case
studies: rainfall, mean temperature, and relative humidity. These parameters are among
the most known factors which illuminate the possible relationship between
Figure 1. Study area locations in Ethiopian map.
3. Data and Methodology
3.1. Climate Data
Climate data for all the study sites were collected from the National Meteorological
Agency Office of Ethiopia. Daily and monthly data of meteorological variables of the past
30 years, from 1990 to 2020, were used in order to see the trend analysis of the selected
climate variables data of the past years. Nine years of climate data (2012–2020) were used
to assess the impacts of the variables on water demand. Three climate variables were
selected to assess the relationship with drinking water consumption within the three case
studies: rainfall, mean temperature, and relative humidity. These parameters are among
the most known factors which illuminate the possible relationship between environmental
resource degradation. Table 1summarizes all the climate data for the study areas. Most of
the research papers produced in this area use similar variables such as temperature and
Water 2022,14, 1722 5 of 20
rainfall [
25
]. In a similar model, Sarah Praskievicz and Heejun Chang [
26
] used relative
humidity, sunshine hour, cloud cover, maximum and minimum temperature, precipitation,
and wind speed to correlate their significance with residential water consumption. Renwick
and Green [
27
] used maximum temperature and monthly cumulative rainfall in their
research model to estimate the water demand management policies in California. With
slight alterations, a focus on rain, temperature, and humidity has generally been shown
to be a key element to assessing water demand. The terminology “water demand” and
“water consumption” were used interchangeably in this study with the same meaning.
Table 1. Average monthly climate data of year 2012–2020.
Months
Arba Minch Debre Birhan Ziway
Max
Temp.
(◦C)
Precipitation
(mm)
R.
Humidity
Max
Temp.
(◦C)
Precipitation
(mm)
R.
Humidity
Max
Temp.
(◦C)
Precipitation
(mm)
R.
Humidity
January 31.34 12.66 0.66 16.75 3.07 0.54 20.41 3.54 0.40
February 33.23 22.01 0.63 18.51 13.57 0.54 21.76 26.43 0.35
March 32.89 66.00 0.70 19.95 44.09 0.57 22.96 38.63 0.37
April 30.14 176.34 0.85 19.81 52.10 0.65 23.29 59.13 0.49
May 28.70 156.05 0.91 19.49 47.36 0.69 23.15 88.29 0.51
June 27.79 86.11 0.89 19.79 55.30 0.67 22.57 85.86 0.58
July 27.62 31.34 0.88 18.68 318.65 0.79 20.78 220.60 0.67
August 28.17 42.40 0.89 17.63 279.18 0.86 20.75 111.51 0.67
September
29.00 109.61 0.88 17.15 97.72 0.83 21.03 90.39 0.63
October 29.95 132.64 0.88 16.81 28.10 0.66 21.05 22.36 0.50
November
29.54 113.07 0.82 16.75 13.46 0.59 20.48 4.50 0.47
December 30.67 35.98 0.72 15.97 1.17 0.54 19.34 1.61 0.41
The climatic data (rainfall, average temperature, and relative humidity) of the meteo-
rological stations in each of the cities (Arba Minch, Ziway, and Debre Birhan) had a total of
6% missing data as estimated by using grid data from the Climate Data Tool (CDT). The
quality of the data was checked to identify any outliers and eliminate data contamination.
The outliers were detected for each station for each month by using a standardization of
the mean square method. The reason for standardizing the infit and outfit mean square
statistics is to allow their statistical significance (or p-values). A familiar method to use for
this purpose is the Z-score or the standard scores [
28
]. Homogeneity testing was performed
using XLSTAT software in which the standard normal homogeneity test (SNHT) of rainfall,
mean temperature, and relative humidity for each required meteorological station were
plotted against time in order to detect the change in series of the variables.
3.2. Water Consumption Data
Nine years (2012–2020) of monthly water consumption data are collected from the
water supply and sewerage offices of the three study sites. The data collected are comprised
of the monthly consumed or billed water used by customers, as it is impossible to obtain
daily data for water consumption in Ethiopia. The data collected are limited to indoor
residential water used and do not include the outdoor water consumption such as that
used for gardening, swimming pools, or backyard fountains.
3.3. Methods
3.3.1. Multiple Linear Regression Model
A multiple linear regression is an extension of a simple linear regression, and it is a
statistical tool that evaluates the relationship between a dependent variable and one or
more independent variables. Two or more independent variables are used to predict or
explain the variance in one dependent variable. It examines whether there is a statistical
Water 2022,14, 1722 6 of 20
relationship between the independent variables and, if there is a relationship, how good it
is. A general multiple regression model equation is given as [29]:
ˆ
y=b0+ b1x1+ b2x2+ b3x3+· · · + bnXn(1)
where
ˆ
y is the predicted value of the dependent variable, and x is the independent variable
of the nth observation. The independent x value input or predictor will fluctuate with y
value output or response in its unique way. The regression coefficients (bn) of independent
variables are respective changes in y per unit change of x
n
. The regression constant (b
0
) is
the y intercept when all predictors or x values are zero. The following regression model
was developed to analyze the factors which influence water demand in the study sites
while taking into consideration the three climate parameters (temperature, precipitation,
humidity). Where
ˆ
y is water demand of the town depending on the variables (m
3
/year),
x
1
is an average monthly temperature of the town (
◦
C); x
2
describes an average monthly
precipitation (mm); x
3
is an average monthly relative humidity of the town; b
0
is the
regression constant; and b
1
to b
3
are regression coefficients for x
1
to x
3
, respectively. The
regression analysis was conducted for the period of 2012–2020, which coincided with the
water consumption data availability of the case study towns.
We reported significant findings by formatting the results as F (Regression df, Residual
df) = F-Ratio, p= Sig. In this case, the 95% confidence interval for the coefficients shown
by many regression packages gives the same information. We have 95% confidence that
the real, underlying value of the coefficient that we are estimating falls somewhere in that
95% confidence interval. So, if the interval does not contain 0, the p-value will be 0.05 or
less. The regression analysis of the model was determined with the SPSS statistical package.
Origin 2018 version was used to draw a better resolution graph. SPSS statistical package
version 26 was used for the principal component analysis (PCA).
3.3.2. Predictor Variable
Principal Component Analysis
Principal component analysis (PCA) is one of the variable selection methods most fre-
quently used in multivariate analysis. PCA is a technique used to examine the interrelations
among a set of variables in order to identify the underlying structure of those variables.
PCA assumes the dataset to be linear combinations of the variables, and regression deter-
mines a line of best fit to a dataset, while factor analysis determines several orthogonal
lines of best fit to the dataset. Locations along each component (or eigenvector) are then
associated with values across all variables. This association between the components and
the original variables is called the component’s eigenvalue. PCA reduces the dimension
of the multivariate data via replacement of a number of variables by a smaller number of
derived variables. Thus, it is a dimension reduction technique by combining features. The
number of principal components to be used is always less than the number of features or
predictors. Those components having an eigenvalue above 1 will be selected, and their
percentage of the variance in the matrix will be identified [30].
PC1 =b11X1+b12X2+· · · +b1kXk=
k
∑
j=1
b1jXj(2)
PC1 =b21X1+b22X2+· · · +b2kXk=
k
∑
j=1
b2jXj(3)
where X
1
, X
2. . .
X
k
represents the original variables in the data matrix, and b
ij
represents
the eigenvectors. The component loading values which are greater than 0.75 (>0.75) are con-
sidered as “strong”; the values ranging from 0.50 to 0.75 (
0.50 ≥component loading ≥0.75
)
are considered “moderate”; and the values ranging from 0.30 to 0.49 (0.30
≥
component
loading ≥0.49) are considered “weak” component loadings [31,32].
Water 2022,14, 1722 7 of 20
A recent research study (Mohammed Gedefaw et al. [
33
]) focusing on optimal predic-
tor variable selection used a multiple regression model for the water demand forecasting
performed on Gonder Town, Ethiopia. The study summarizes that PCA (principal compo-
nent analysis) is a more appropriate method for predicting water demand of urban towns
and, thus, can be used for other towns in Ethiopia. Therefore, following the recommen-
dation of Mohammed Gedefaw et al., we also adopt the PCA method in our study as a
variable predictor model to identify the variable.
4. Results and Discussion
4.1. Water Demand/Consumption
Drinking water scarcity is one of the pressing issues in developing countries such
as Ethiopia. Researchers state that climate variability and uneven spatial and temporal
distributions of hydrological parameters have an impact on drinking water demand [
34
,
35
].
Since drinking water source availability is one of the major limiting factors of economic
growth [
4
], focusing on the demand management programs is one of the best solutions
to reduce upcoming problems. The three selected case study areas are representative
locations of the varying climate characteristics across all Ethiopian towns. Consequently,
the residents of the three study areas have similar living standards.
According to the Ethiopian federal guidelines, in order to estimate a town’s actual
water the base water demand is multiplied by the mean annual precipitation and altitude
difference of the town. For example, if the mean annual precipitation of the town is less
than 600 mm, multiplied by 1.1; if between 601–900 mm, by 1.0; and multiplied by 0.9 if
the precipitation is above 901 mm [
36
]. However, using the same climatic factor for two
different towns located at the same altitude or having equivalent annual mean precipitation
amount does not seem right, as they might have different climatic situations and socio-
economic conditions. Water demand forecasting should be performed using a suitable
mathematical formula that incorporates compatible water demand models with different
climatic variables and has an affiliation with water consumption [
37
]. An Ethiopian Ministry
of Water Resources report states that population size, economic, social, and climatic factors
as well as the town’s mode of service should be considered while evaluating urban water
demand [36].
Multiple regression models were used to assess and calculate the residential water
demand by calculating average temperature, precipitation, and relative humidity of nine
consecutive years in relation to dependent monthly billed water consumption data. Results
of this study are expected to provide important information which will help in developing
an effective water demand management system in urban Ethiopian towns. Additionally,
this research study will produce guidelines to more accurately calculate water demand by
incorporating the impacts of climate variables on residential water demand. The results of
the impacts of the selected climate variables on the water demand in each study area town
will be discussed. The results of demand analysis and impact prediction output for each
study area will be illustrated in detail.
4.2. Seasonal Water Consumption
The evolution of water consumption for Arba Minch shows an increasing trend
throughout the study period (2012–2020). To verify the seasonal impact, the seasonal
component of the residential water consumption was investigated. The plot shows the
total water consumption in m
3
(y-axis) with corresponding years (x-axis). The analysis
resulted in several important observations. The highest consumption occurs during the
Belg season (Feb–May), which is the short rainy season of the country, with an average
value of 34.5% of total annual consumption. The lowest water consumption (32.2% of
the total) occurs during the Kiremt (October–January), while water demand in the Bega
season (June–September) represents 33.3% of the total annual consumption. The residential
water consumption shows an increasing trend throughout the study period, as shown in
Water 2022,14, 1722 8 of 20
Figure 2. The maximum water consumption, seasonal based component occurs in March
and a minimum in August.
Water 2022, 14, x FOR PEER REVIEW 8 of 20
occurs during the Kiremt (October–January), while water demand in the Bega season
(June–September) represents 33.3% of the total annual consumption. The residential water
consumption shows an increasing trend throughout the study period, as shown in Figure
2. The maximum water consumption, seasonal based component occurs in March and a
minimum in August.
Seasonal average water consumption of Debre Birhan showed a rising trend in aver-
age water consumption ranging between 0.4 Mm3 and 0.7 Mm3 during the study period.
The highest levels of water consumption came in the year 2018 during the Belg season.
This change can be attributed to water demands from newly emerged brewery industries
and factories relying on municipal water supply sources (information confirmed from the
town water supply office) and temperature increases during the same period. Interannual
monthly water consumption of the town showed that less water consumption occurred
during the months of October to January, which is the peak rainy season. Overall, the
highest consumption occurred during the Belg season (February–May), with an average
value of 35.8% of total annual consumption. The lowest water consumption (30.7% of the
total) occurred during the Kiremt season (October–January), while water demand in the
Bega season (June–September) represented 33.4% of the annual consumption.
Ziway town is a moderate climate area that can be classified by its temperate climate
conditions. The seasonal water consumption throughout the study period for the town
was almost flat, as shown in Figure 2. The highest water consumption occurred during
the Kiremt season (October–January), with an average value of 34.0% of total annual con-
sumption. The lowest water consumption (32.5% of the total) occurred during the Bega
(June–September), while water consumption in the Belg season (February–May) repre-
sented 33.5% of the annual consumption.
Figure 2. Seasonal water consumption distribution.
Figure 2. Seasonal water consumption distribution.
Seasonal average water consumption of Debre Birhan showed a rising trend in average
water consumption ranging between 0.4 Mm
3
and 0.7 Mm
3
during the study period. The
highest levels of water consumption came in the year 2018 during the Belg season. This
change can be attributed to water demands from newly emerged brewery industries and
factories relying on municipal water supply sources (information confirmed from the
town water supply office) and temperature increases during the same period. Interannual
monthly water consumption of the town showed that less water consumption occurred
during the months of October to January, which is the peak rainy season. Overall, the
highest consumption occurred during the Belg season (February–May), with an average
value of 35.8% of total annual consumption. The lowest water consumption (30.7% of the
total) occurred during the Kiremt season (October–January), while water demand in the
Bega season (June–September) represented 33.4% of the annual consumption.
Ziway town is a moderate climate area that can be classified by its temperate climate
conditions. The seasonal water consumption throughout the study period for the town
was almost flat, as shown in Figure 2. The highest water consumption occurred during
the Kiremt season (October–January), with an average value of 34.0% of total annual
consumption. The lowest water consumption (32.5% of the total) occurred during the
Bega (June–September), while water consumption in the Belg season (February–May)
represented 33.5% of the annual consumption.
According to the analysis, the percentage of seasonal water consumption in Arba
Minch and Debre Birhan are correlative, and the highest and lowest seasonal water con-
Water 2022,14, 1722 9 of 20
sumption occurs during the months with maximum temperature and minimum precip-
itation, respectively. This result looks similar to other studies which found a significant
relationship between annual total water consumption, temperature, and precipitation.
Bailing and Gober [
13
] found a negative relation with precipitation and positive relation
with temperature. Furthermore, it is obvious that people consume more water during
hot weather and less water during high precipitation weather. This finding is similar to
the results of Kumudu Rathnayaka et al. [
38
], who found that weather is shown to be a
significant determinant of shower water use. We may infer that weather differences are
one of the determinants that cause high water demand in Arba Minch: the frequency of
taking showers in hotter climates such as that of Arba Minch is not the same as areas with
cold weather such as Debre Birhan.
4.3. Per Capita Water Consumption
According to a 2019 WHO/UNICEF JMP report, only 41% of the Ethiopian population
has access to safely managed drinking water. For urban populations, 51% of the total supply
is accessed by city pipes and water infrastructure. However, unaddressed socioeconomic,
demographic, and hydrological changes have created unreliable and insufficient water
demand across the country. The water supply service level varies from place to place
depending on economic status and other factors. While developed countries use up to
500 L per capita per day of water, in less developed countries such as Ethiopia this figure
is out of reach, and they are challenged to achieve even 15 L per capita per day [
39
]. As
such, Ethiopia has set a target towards achieving middle income status by 2025 with 40 to
100 L per capita per day (LCD) in urban areas and 25 LCD at maximum distance of 1 km
access of supply point in rural areas [
36
]. The monthly average distribution of per capita
water consumption for the three case studies was shown in Figure 3. This consumption
rate shows only the households which have access to the town’s water supply services.
Accordingly, Arba Minch consumed an average of around 70 L per day, Ziway 50 L, and
Debre Birhan uses 40 L on average.
Water 2022, 14, x FOR PEER REVIEW 9 of 20
According to the analysis, the percentage of seasonal water consumption in Arba
Minch and Debre Birhan are correlative, and the highest and lowest seasonal water con-
sumption occurs during the months with maximum temperature and minimum precipi-
tation, respectively. This result looks similar to other studies which found a significant
relationship between annual total water consumption, temperature, and precipitation.
Bailing and Gober [13] found a negative relation with precipitation and positive relation
with temperature. Furthermore, it is obvious that people consume more water during hot
weather and less water during high precipitation weather. This finding is similar to the
results of Kumudu Rathnayaka et al. [38], who found that weather is shown to be a sig-
nificant determinant of shower water use. We may infer that weather differences are one
of the determinants that cause high water demand in Arba Minch: the frequency of taking
showers in hotter climates such as that of Arba Minch is not the same as areas with cold
weather such as Debre Birhan.
4.3. Per Capita Water Consumption
According to a 2019 WHO/UNICEF JMP report, only 41% of the Ethiopian popula-
tion has access to safely managed drinking water. For urban populations, 51% of the total
supply is accessed by city pipes and water infrastructure. However, unaddressed socio-
economic, demographic, and hydrological changes have created unreliable and insuffi-
cient water demand across the country. The water supply service level varies from place
to place depending on economic status and other factors. While developed countries use
up to 500 L per capita per day of water, in less developed countries such as Ethiopia this
figure is out of reach, and they are challenged to achieve even 15 L per capita per day [39].
As such, Ethiopia has set a target towards achieving middle income status by 2025 with
40 to 100 L per capita per day (LCD) in urban areas and 25 LCD at maximum distance of
1 km access of supply point in rural areas [36]. The monthly average distribution of per
capita water consumption for the three case studies was shown in Figure 3. This consump-
tion rate shows only the households which have access to the town’s water supply ser-
vices. Accordingly, Arba Minch consumed an average of around 70 L per day, Ziway 50
L, and Debre Birhan uses 40 L on average.
Figure 3. Average per capita water consumption distribution of 2012 to 2020 study period.
Figure 3. Average per capita water consumption distribution of 2012 to 2020 study period.
4.4. Annual Mean Temperature
Debre Birhan is the coldest place in Ethiopia, with an average mean temperature
of
18 ◦C
throughout the study period (2012–2020). According to Figure 4, the water
consumption trend is compatible with the mean annual temperature (with the exception
of an abrupt change in 2017 which might be due to a data recording mistake). The graph
Water 2022,14, 1722 10 of 20
shows that when the temperature increases, the water consumption also increases. Average
monthly water consumption of Debre Birhan during the months of February to May
increased with a rise in mean temperature during the same months. The implication of
this finding confirms that outside temperature plays a very strong role in increasing or
reducing seasonal water demand [
40
]. Figure 4shows that the minimum monthly mean
temperature value for Arba Minch (around 20.75
◦
C) area is greater than the highest mean
monthly temperature for Debre Birhan. Arba Minch and Debre Birhan have a trend of
increasing mean annual temperature, while Ziway does not follow this upward trend. This
may be due to the study area’s location at mid-latitude, where there are more variabilities of
temperature across years and within the same year. However, the total water consumption
variation is compatible with the mean annual temperature fluctuations.
Water 2022, 14, x FOR PEER REVIEW 10 of 20
4.4. Annual Mean Temperature
Debre Birhan is the coldest place in Ethiopia, with an average mean temperature of
18 °C throughout the study period (2012–2020). According to Figure 4, the water con-
sumption trend is compatible with the mean annual temperature (with the exception of
an abrupt change in 2017 which might be due to a data recording mistake). The graph
shows that when the temperature increases, the water consumption also increases. Aver-
age monthly water consumption of Debre Birhan during the months of February to May
increased with a rise in mean temperature during the same months. The implication of
this finding confirms that outside temperature plays a very strong role in increasing or
reducing seasonal water demand [40]. Figure 4 shows that the minimum monthly mean
temperature value for Arba Minch (around 20.75 °C) area is greater than the highest mean
monthly temperature for Debre Birhan. Arba Minch and Debre Birhan have a trend of
increasing mean annual temperature, while Ziway does not follow this upward trend.
This may be due to the study area’s location at mid-latitude, where there are more varia-
bilities of temperature across years and within the same year. However, the total water
consumption variation is compatible with the mean annual temperature fluctuations.
Figure 4. Total annual water consumption vs. annual mean temperature of the study area.
4.5. Average Relative Humidity
Relative humidity plotted with annual average consumption (Figure 5) shows a trend
of compatibility for all three study. The Arba Minch plot shows a compatible relationship
between average relative humidity and water consumption, while in Debre Birhan and
Ziway it varies and fluctuates from year to year. This also proves that average relative
humidity and average annual water consumption have perfect compatibility with each
other such that when relative humidity increases, water consumption also increases. The
relative humidity plot of Ziway was increasing starting from 2015 until the peak in 2016.
This may be because the drought season occurred around this area at the end 2015 until
Figure 4. Total annual water consumption vs. annual mean temperature of the study area.
4.5. Average Relative Humidity
Relative humidity plotted with annual average consumption (Figure 5) shows a trend
of compatibility for all three study. The Arba Minch plot shows a compatible relationship
between average relative humidity and water consumption, while in Debre Birhan and
Ziway it varies and fluctuates from year to year. This also proves that average relative
humidity and average annual water consumption have perfect compatibility with each
other such that when relative humidity increases, water consumption also increases. The
relative humidity plot of Ziway was increasing starting from 2015 until the peak in 2016.
This may be because the drought season occurred around this area at the end 2015 until
2016. Arba Minch and Debre Birhan plots of the water consumption and relative humidity
have a good compatibility trend. Generally, relative humidity has a significant relationship
with air temperature and rainfall. Therefore, relative humidity plays a crucial role in spatial
patterns of urban water consumption.
Water 2022,14, 1722 11 of 20
Water 2022, 14, x FOR PEER REVIEW 11 of 20
2016. Arba Minch and Debre Birhan plots of the water consumption and relative humidity
have a good compatibility trend. Generally, relative humidity has a significant relation-
ship with air temperature and rainfall. Therefore, relative humidity plays a crucial role in
spatial patterns of urban water consumption.
Figure 5. Average relative humidity vs. water consumption for the study areas.
4.6. Average Annual Precipitation
Rainfall plots throughout the study period (2012–2020) show huge fluctuation in all
case study areas. The absence of rainfall around the year 2015 is indicative of a major
drought which impacted large areas of the Horn of Africa. The 2015 drought was the worst
to occur in decades, affecting around 10 million people living across 30% of the country
[41]. It confirms that the drop in rainfall for Ziway area in 2015 is the cause of the drought.
Other urban towns also confirm that there was less rainfall recorded around the years
2015 and 2016. It is clear from Figure 6 that Debre Birhan has more rainfall compared to
the other study areas. Arba Minch’s rainfall amount ranged between 700 and 900 mm,
with less fluctuation throughout the study period, while for Ziway and Debre Birhan the
amount of fluctuation in precipitation was relatively large year to year. As identified in
Figure 6, the general compatibility of water consumption with annual precipitation for all
study areas was fairly low. Thus, they also have a negative correlation with each other.
This means that when precipitation increases, annual water consumption decreases and
vice versa.
Ziway and Debre Birhan have similar trends in precipitation, with a single modal
where the highest peak rainfall occurs in the months from July to October for both study
areas (Figure 7). Annual rainfall of Debre Birhan ranges from 814 to 1080 mm with about
70% rain falling between June and September [42], while Arba Minch has a different pre-
cipitation trend with bimodal rainfall type. Arba Minch receives the highest peak rainfall
at two times during the year, April and October, though other areas receive less rainfall
Figure 5. Average relative humidity vs. water consumption for the study areas.
4.6. Average Annual Precipitation
Rainfall plots throughout the study period (2012–2020) show huge fluctuation in all
case study areas. The absence of rainfall around the year 2015 is indicative of a major
drought which impacted large areas of the Horn of Africa. The 2015 drought was the worst
to occur in decades, affecting around 10 million people living across 30% of the country [
41
].
It confirms that the drop in rainfall for Ziway area in 2015 is the cause of the drought. Other
urban towns also confirm that there was less rainfall recorded around the years 2015 and
2016. It is clear from Figure 6that Debre Birhan has more rainfall compared to the other
study areas. Arba Minch’s rainfall amount ranged between 700 and 900 mm, with less
fluctuation throughout the study period, while for Ziway and Debre Birhan the amount of
fluctuation in precipitation was relatively large year to year. As identified in Figure 6, the
general compatibility of water consumption with annual precipitation for all study areas
was fairly low. Thus, they also have a negative correlation with each other. This means that
when precipitation increases, annual water consumption decreases and vice versa.
Ziway and Debre Birhan have similar trends in precipitation, with a single modal
where the highest peak rainfall occurs in the months from July to October for both study
areas (Figure 7). Annual rainfall of Debre Birhan ranges from 814 to 1080 mm with about
70% rain falling between June and September [
42
], while Arba Minch has a different
precipitation trend with bimodal rainfall type. Arba Minch receives the highest peak
rainfall at two times during the year, April and October, though other areas receive less
rainfall during these months. Even though there is plenty of rainfall in the months of March
and April, the water consumption was high due to the temperature and humid air.
Water 2022,14, 1722 12 of 20
Water 2022, 14, x FOR PEER REVIEW 12 of 20
during these months. Even though there is plenty of rainfall in the months of March and
April, the water consumption was high due to the temperature and humid air.
Figure 6. Annual average precipitation vs. average water consumption.
Figure 7. Total precipitation vs. mean temperature of the study areas (2012–2020).
Figure 6. Annual average precipitation vs. average water consumption.
Water 2022, 14, x FOR PEER REVIEW 12 of 20
during these months. Even though there is plenty of rainfall in the months of March and
April, the water consumption was high due to the temperature and humid air.
Figure 6. Annual average precipitation vs. average water consumption.
Figure 7. Total precipitation vs. mean temperature of the study areas (2012–2020).
Figure 7. Total precipitation vs. mean temperature of the study areas (2012–2020).
4.7. Correlation between Climate Variables and Monthly Water Consumption
The correlation between the total water consumption in Arba Minch and recorded
mean temperature shows weak correlation. Relative humidity and precipitation do not
Water 2022,14, 1722 13 of 20
have any correlation with average monthly water consumption. The percent variation in
the consumption rates for each 1
◦
C increase in average temperature is computed based on
the trend line in Figure 8. Accordingly, the slope shows that a mean temperature increase of
1
◦
C causes an average increase in the rate of water consumption by 5.28%. The trendline
slope shows that a 1% increase in humidity caused a 0.16% decrease in water consumption,
and a 1 mm increase of precipitation caused a 0.02% decrease of water consumption.
The R-square value for relative humidity and precipitation parameters shows that the
predictor explains less than 1% of the variability in the dependent variable, whereas 13%
of the variance in water consumption is explained by the variability of mean temperature.
Therefore, mean temperature has a relatively positive correlation with water consumption
when compared with the other analyzed parameters.
Water 2022, 14, x FOR PEER REVIEW 13 of 20
4.7. Correlation between Climate Variables and Monthly Water Consumption
The correlation between the total water consumption in Arba Minch and recorded
mean temperature shows weak correlation. Relative humidity and precipitation do not
have any correlation with average monthly water consumption. The percent variation in
the consumption rates for each 1 °C increase in average temperature is computed based
on the trend line in Figure 8. Accordingly, the slope shows that a mean temperature in-
crease of 1 °C causes an average increase in the rate of water consumption by 5.28%. The
trendline slope shows that a 1% increase in humidity caused a 0.16% decrease in water
consumption, and a 1 mm increase of precipitation caused a 0.02% decrease of water con-
sumption. The R-square value for relative humidity and precipitation parameters shows
that the predictor explains less than 1% of the variability in the dependent variable,
whereas 13% of the variance in water consumption is explained by the variability of mean
temperature. Therefore, mean temperature has a relatively positive correlation with water
consumption when compared with the other analyzed parameters.
Figure 8. Monthly water consumption trendline of the climate variables of Arba Minch (2012–2020).
According to the scatter plot shown for Ziway (Figure 9), mean temperature, precip-
itation, and relative humidity are not correlated with monthly average water consump-
tion. The Pearson’s correlation coefficient value “r” is almost zero for all of the variables,
which means that they do not have any relationship with each other. The R-square value,
which indicates the proportion of the variation in the dependent variable, shows that pre-
cipitation and mean temperature explain less than 1% of the variability in water consump-
tion. Alternatively, relative humidity variation affects monthly water consumption pat-
tern variability by 3%. Reduced into single increments, each 1 °C increase in mean tem-
perature causes an average increase of water consumption by 0.175%; each 1% increase in
relative humidity causes a decrease of water consumption by 0.18%; and each 1 mm in-
crease of precipitation decreases the monthly water consumption by 0.04%.
Figure 8.
Monthly water consumption trendline of the climate variables of Arba Minch (2012–2020).
According to the scatter plot shown for Ziway (Figure 9), mean temperature, precipi-
tation, and relative humidity are not correlated with monthly average water consumption.
The Pearson’s correlation coefficient value “r” is almost zero for all of the variables, which
means that they do not have any relationship with each other. The R-square value, which
indicates the proportion of the variation in the dependent variable, shows that precipitation
and mean temperature explain less than 1% of the variability in water consumption. Alter-
natively, relative humidity variation affects monthly water consumption pattern variability
by 3%. Reduced into single increments, each 1
◦
C increase in mean temperature causes an
average increase of water consumption by 0.175%; each 1% increase in relative humidity
causes a decrease of water consumption by 0.18%; and each 1 mm increase of precipitation
decreases the monthly water consumption by 0.04%.
Debre Birhan’s monthly water consumption and the three independent variable scatter
plots show that relative humidity has a weak negative correlation, while precipitation and
mean temperature have no correlation (Figure 10). A mean temperature increase of 1
◦
C
causes an average increase of 0.29% in the rate of water consumption, and a 1% increase in
relative humidity causes a 0.3% decrease in water consumption. The effect of precipitation
was not statistically significant and smaller in magnitude; a 1 mm rise in precipitation
Water 2022,14, 1722 14 of 20
results in a 0.014% drop in water consumption. The R-square value of the predictors are
close to zero. As such, 10% of the variance in water consumption can be explained by
relative humidity, while less than 1% of variance was explained by the precipitation and
mean temperature variables.
Water 2022, 14, x FOR PEER REVIEW 14 of 20
Figure 9. Monthly water consumption trendline for the climate variable of Ziway (2012–2020).
Debre Birhan’s monthly water consumption and the three independent variable scat-
ter plots show that relative humidity has a weak negative correlation, while precipitation
and mean temperature have no correlation (Figure 10). A mean temperature increase of 1
°C causes an average increase of 0.29% in the rate of water consumption, and a 1% increase
in relative humidity causes a 0.3% decrease in water consumption. The effect of precipita-
tion was not statistically significant and smaller in magnitude; a 1 mm rise in precipitation
results in a 0.014% drop in water consumption. The R-square value of the predictors are
close to zero. As such, 10% of the variance in water consumption can be explained by
relative humidity, while less than 1% of variance was explained by the precipitation and
mean temperature variables.
Figure 9. Monthly water consumption trendline for the climate variable of Ziway (2012–2020).
Water 2022, 14, x FOR PEER REVIEW 15 of 20
Figure 10. Monthly water consumption trendline for the climate variable of Debre Birhan (2012–
2020).
4.8. Multiple Linear Regressions and Principal Component Analysis
4.8.1. Arba Minch
This study was conducted to determine whether climate variables reliably predict
residential water demand. It was hypothesized that average mean temperature, precipi-
tation, and relative humidity would impact the water demand. Results show that 14% of
the variance in water demand can be accounted for by the three predictors. The sum result
of the variables showed that F (3, 80) = 4.39 and p = 0.007—looking at the unique individual
contributions of the predictors: the mean temperature (t = 3.31, p = 0.01), relative humidity
(t = 0.38, p = 0.7), and precipitation (t = −0.091, p = 0.93). Therefore, the p-value for mean
temperature is p = 0.01, which is less than 0.05, such that mean temperature is a predictor
variable. In the multiple linear regression test analyzed for the test results, the p-values
show that there is a significant relationship between the dependent and independent var-
iables. This confirms that, although precipitation and relative humidity do not have a sig-
nificant relationship with residential water demand, mean temperature does.
The general result of the regression equation is written as below:
Water demand (m3/month) = −37052 + 5935 × Mean Temp + −2.8 × Precipitation + 5307 × Humidity
(4)
The results of the PCA on the three independent variables was able to identify the
impact of the predictor variables on residential water consumption. According to the PCA
output, only one factor has an eigenvalue above 1. Though only one PC explained 58.3%
of the variability, the relative humidity and mean temperature variables (0.856 and −0.756)
are incorporated component loadings in the PC and also have high loadings. The residen-
tial water consumption changes were caused mainly by the changes of the highest com-
ponent loading value, which is relative humidity and mean temperature.
Figure 10.
Monthly water consumption trendline for the climate variable of Debre Birhan (2012–2020).
Water 2022,14, 1722 15 of 20
4.8. Multiple Linear Regressions and Principal Component Analysis
4.8.1. Arba Minch
This study was conducted to determine whether climate variables reliably predict
residential water demand. It was hypothesized that average mean temperature, precip-
itation, and relative humidity would impact the water demand. Results show that 14%
of the variance in water demand can be accounted for by the three predictors. The sum
result of the variables showed that F (3, 80) = 4.39 and p= 0.007—looking at the unique
individual contributions of the predictors: the mean temperature (t = 3.31, p= 0.01), relative
humidity (t = 0.38, p= 0.7), and precipitation (t =
−
0.091, p= 0.93). Therefore, the p-value
for mean temperature is p= 0.01, which is less than 0.05, such that mean temperature is a
predictor variable. In the multiple linear regression test analyzed for the test results, the
p-values show that there is a significant relationship between the dependent and indepen-
dent variables. This confirms that, although precipitation and relative humidity do not
have a significant relationship with residential water demand, mean temperature does.
The general result of the regression equation is written as below:
Water demand (m3/month) = −37,052 + 5935 ×Mean Temp + −2.8 ×Precipitation + 5307 ×Humidity (4)
The results of the PCA on the three independent variables was able to identify the
impact of the predictor variables on residential water consumption. According to the PCA
output, only one factor has an eigenvalue above 1. Though only one PC explained 58.3% of
the variability, the relative humidity and mean temperature variables (0.856 and
−
0.756) are
incorporated component loadings in the PC and also have high loadings. The residential
water consumption changes were caused mainly by the changes of the highest component
loading value, which is relative humidity and mean temperature.
4.8.2. Debre Birhan
The result from the regression shows that 12% of the variance in water demand can
be accounted by the three predictors, collectively: F (3, 104) = 4.86, p= 0.03. Looking
at the unique individual contributions of the predictors, the mean temperature (t = 0.06,
p= 0.95) and precipitation (t = 1.4, p= 0.16) had a value of p> 0.05 for both predictors.
The contribution of one predictor, relative humidity (t =
−
3.51, p= 0.001), was proved to
positively predict residential water demand. The p-values in the multiple linear regression
test showed that there is a significant relationship between the dependent and independent
variables. This suggests that precipitation and mean temperature variables do not have
a significant relationship with residential water demand, but relative humidity has a
significant relationship with the residential water demand.
The general result of the regression equation can be interpreted as below:
Water demand (m3/month) = 97,535 + 31 ×Mean Temp + 15 ×Precipitation −323 ×Humidity (5)
According to the PCA output, two factors have eigenvalues above 1, such that two
PCs explained around 56% and 34% of the variability. The precipitation (0.921) and relative
humidity (0.914) variables are incorporated in the first PC and have high loadings. Mean
temperature (0.994) is the only high loading independent variable incorporated in the
second principal component. The values with high loadings of the independent variables
show which variables would have more effect on water demand from the PCA. According
to the result of the regression analysis and PCA, the predicted variables have a significant
relationship with the dependent variable. Accordingly to the PCA result, component one
contains 56% of the variation of the relative humidity and precipitation variables. As such,
these two variables highly explain the variation of water demand.
4.8.3. Ziway
Results from Ziway show that 8% of the variance in water demand can be accounted
for by the three predictors, collectively: F (3, 80) = 2.44, p= 0.07. However, p> 0.05, which
Water 2022,14, 1722 16 of 20
means that the model is not significant to predict the residential water demand according
to any of the predictors: the mean temperature (t = 0.25, p= 0.79), precipitation (t = −1.79,
p= 0.076
), or relative humidity (t =
−
0.53, p= 0.59). In the multiple linear regression test, the
p-values suggest that there is no significant relationship between residential water demand
and selected variables. This indicates that none of the variables has a significant relationship
with residential water demand. The regression analysis did not show clearly the interaction
and relationship between the residential water demand and climate variables.
The general result of the regression equation can be interpreted as below:
Water demand (m3/month) = 77,768 + 151 ×Mean Temp −25 ×Precipitation −4433 ×Humidity (6)
PCA was conducted on the three independent variables to identify the predictor
variables on residential water demand. According to the PCA results, two factors have
eigenvalues above 1, such that there are two PCs that explained around 53% and 33% of the
variability. Precipitation (0.894) and relative humidity (0.894) variables are incorporated
in the first PC with high loadings. Mean temperature (0.999) is the only high loading
independent variable incorporated in the second PC. The values with high loadings of the
independent variables identify which variables would have much more effect on water
demand from the PCA.
Collectively, the climate variables used for the multiple linear regression analysis
explain the variance in residential water consumption in some case study towns. The
spatial variation of the study areas has an influence on water consumption of the towns,
which indicates that some particular variables have an impact on some specific areas, while
some do not. Residual plots were also verified in order to check the goodness of a fitted
model. Consequently, Arba Minch and Debre Birhan show random patterns with randomly
distributed data points, while Ziway data points show some non-random pattern. In the
regression analysis, the most consistently significant variables were mean temperature and
relative humidity. This indicates that it would be better for the future study to include other
non-climatic determinants such as socio-demographic influences. Basically, at seasonal and
a monthly time scale, interannual variation in the climate parameters was shown to identify
the important predictor of the seasonal water consumption. At a daily scale of water
consumption, the influence of the climate variables on water consumption could explain
with high precision the percentage of the variation in water consumption. Therefore, our
result suggests that, although a significant amount of variance can be predicted by using
monthly data, daily based data might explain the variation of water consumption in a more
precise way.
5. Conclusions and Future Work
The investigation of weather influences on seasonal and interannual water consump-
tion at three study areas is presented in this section. Three independent climate parameters
(mean temperature, relative humidity, and precipitation) were chosen to assess and de-
termine their relationship with residential water consumption rates in three study areas.
Accordingly, climate variables specifically used in this study could have a significant impact
on both seasonal and interannual water consumption.
The lowest water consumption occurs during “Kiremt” season, which is October–
January for Arba Minch and Debre Birhan, while the highest consumption occurred for
Ziway. The highest water consumption occurs during “Belg” season, which is February to
May, for Debre Birhan and Arba Minch, but the lowest consumption occurs during “Bega”
season (June to September) for Ziway. The maximum water consumption happens during
March and the minimum in August for Arba Minch. For Debre Birhan, the maximum occurs
around March to May and the minimum around October to December. The maximum
occurs around December and January for Ziway, while the minimum is during the month
of August. The geographical and altitude distribution within the country could be the
main cause of the differences between the three climate variables. So, the differences in
interannual and seasonal water consumptions are the consequence of climate variables.
Water 2022,14, 1722 17 of 20
Mean temperature increases of 1
◦
C cause an average increase in the rate of water
consumption of approximately 5.28%, 0.29%, and 0.175% for Arba Minch, Debre Birhan,
and Ziway, respectively. This shows that Arba Minch is more sensitive to mean temperature
increases, as demonstrated by the amount of water consumed relative to other case study
areas. This is because the area is classified under one of the hottest in Ethiopia or “Kola” and
found on an altitude 1300 m above sea level. Conversely, relative humidity and precipitation
have a negative impact with weak magnitude on residential water consumption in the
Arba Minch area. Ziway has a Pearson’s correlation value R close to zero for the mean
temperature variable. This implies that mean temperature does not have any relationship
with residential water consumption. According to the scatter plot of the three climate
variables and water consumption rates of Debre Birhan, relative humidity has a weak
relationship with the residential water consumption.
Multiple linear regression models were used to explain how much variation in water
consumption could be explained by weather variables for each study site. According to the
regression results, residential water consumption is primarily influenced by mean tempera-
ture in Arba Minch. Relative humidity is the only independent variable with p< 0.05 for
Debre Birhan and, thus, the dominant variable. Residential water consumption of Ziway is
primarily influenced by other determinants (likely factors such as demographics, water
availability, and socioeconomic status) rather than the climate variables under investigation
within this study, because the p-value is greater than 0.05 for all the variables assessed in
the multiple regression analysis.
The predictor variable analysis was performed using principal component analysis,
which is one of the most common and frequently used variable selection methods used
in multivariate analysis. Arba Minch has only one component factor (PC1), with mean
temperature having 0.76 loading and relative humidity having 0.86 loading. Debre Birhan
and Ziway have two component factors (PC1 and PC2) each; the first component has
precipitation and relative humidity variables, and the second component has a mean
temperature variable. Accordingly, Debre Birhan has a component loading of 0.92, 0.91,
and 0.99 for precipitation, relative humidity, and mean temperature, respectively. Ziway
has component loadings of 0.89, 0.89, and 0.99 for precipitation, relative humidity, and
mean temperature, respectively.
The findings from this study demonstrate that the analyzed climate parameters have
a significant relationship with the water consumption. Since the general objective of the
study is assessments rather than comparison of the models, the results obtained from each
method could be different. From the scatter plot and the multiple linear regression analysis,
relative humidity is correlated with residential water consumption at Debre Birhan. In
Arba Minch, mean temperature’s correlation result is similar for both scatter plot and
multiple regression. Since the number of the variables studied is small, it is not possible to
select a few parameters as the dominant variable in principal component analysis. Instead,
increasing the number of variables with other determinant variables would be a better way
to obtain a precise result. Generally, in this study, we focused solely on the three climate
variables’ influences on monthly water consumption, but water consumption depends
on other factors too, such as pricing, demographic, and conservation measures [
43
,
44
].
We found that the influence of temperature is stronger than that of relative humidity
and precipitation at lower altitude areas such as Arba Minch. The influence of relative
humidity is stronger at Debre Birhan, which might be due to the fact that relative humidity
increases with altitude in the atmospheric boundary layer [
45
]. Since higher elevations in
the troposphere usually have lower temperatures, relative humidity will be higher.
Because of the topographical distribution, Ethiopia is naturally exposed to climate
variability, making the water consumption management and ability to estimate the impacts
of climate on water consumption more complex. The main findings of this paper show
that geographical distribution and other determinants are more important determinants
of residential water demand, even though households in warmer cities consume more
water that households in colder cities. Climate change has led global water availability
Water 2022,14, 1722 18 of 20
to become more variable and unpredictable [
46
]. As such, the decision-makers and wa-
ter managers need specific information on climate variables and their impacts on water
management. Challenges to collect relevant data on water consumption include the perils
of manual recording, limited water supply coverage, the intermittent water distribution
system, and the prevalence of private hand-dug wells at the household’s backyard. Specific
and relevant information about the climate variables’ impacts on water demand would
enhance the water management system and also better evaluate and audit the water pro-
duced from the sources, since the unavailability of other data imposes it upon us to use
a few climate variables to assess the relationship between the variables and water con-
sumption. Recommendations for future research should include assessments that measure
daily or monthly climate data as well as other factors such as demographic data, water
price, availability of extra water sources in backyards (private hand-dug well), and water
management measure issues. With these changes, more precise results can be obtained to
identify which parameter is dominantly causing the changes in water consumption in each
of the study areas.
Author Contributions:
M.T.T. and B.M.B. developed the research idea and methodology. M.T.T.
collected the required data and performed the analysis and manuscript writing. M.B. followed up all
procedures, supervised the study, and edited and reviewed the manuscript. All authors have read
and agreed to the published version of the manuscript.
Funding:
This research is part of the DAAD-EECBP Home Grown PhD Scholarship Program un-der
(EECBP Homegrown PhD Program-2019). The APC was funded by the Open Access TU Publication
fund of Technical University of Berlin, Berlin.
Institutional Review Board Statement: Not applicable.
Data Availability Statement: The original data used in this study are available from the authors on
reasonable request.
Acknowledgments:
We would like to thank Arba Minch, Debre Birhan, and Ziway towns water
supply offices for providing valuable datasets for this study. The authors would like to thank the
three anonymous reviewers for their valuable comments which were helpful in improving the paper.
Conflicts of Interest: The authors have not declared any conflict of interests.
References
1.
Bates, B.; Kundzewicz, Z.; Wu, S.; Palutikof, J. Climate Change and Water; Intergovernmental Panel on Climate Change: Geneva,
Switzerland, 2008.
2.
UN Report. The Future Is Now Science for Achieving Sustainable Development. 2019. Available online: https:
//sustainabledevelopment.un.org/content/documents/24797GSDR_report_2019.pdf (accessed on 5 December 2020).
3.
Bio Intelegence. Literature Review on the Potential Climate Change Effects on Drinking Water Resources across the EU and the
Identification of Priorities among Different Types of Drinking Water Supplies. 2012. Available online: https://ec.europa.eu/
environment/archives/water/adaptation/pdf/ADWICE_FinalReport.pdf (accessed on 5 June 2020).
4.
Yahaya, O.; Nathaniel, E.; Micheal, O. Simulation and Modelling of Climate Change Ef-fects on River Awara Flow Discharge
Using WEAP Model for Effective Water Management. Int. J. Technol. Enhanc. Emerg. Eng. Res. 2014,2, 128–130.
5.
Eslamian, S.A.; Li, S.S.; Haghighat, F. A new multiple regression model for predictions of urban water use. Sustain. Cities Soc.
2016,27, 419–429. [CrossRef]
6.
Syme, G.J.; Shao, Q.; Po, M.; Campbell, E. Predicting and understanding home garden water use. Landsc. Urban Plan.
2004
,68,
121–128. [CrossRef]
7.
Gato, S.; Jayasuriya, N.; Roberts, P. Forecasting Residential Water Demand: Case Study. J. Water Resour. Plan. Manag.
2007
,133,
309–319. [CrossRef]
8. Schleich, J.; Hillenbrand, T. Determinants of residential water demand in Germany. Ecol. Econ. 2009,68, 1756–1769. [CrossRef]
9.
Alshaikhli, M.; Aqeel, S.; Valdeolmillos, N.; Fathima, F.; Choe, P. A Multi-Linear Regression Model to Predict the Factors Affecting
Water Consumption in Qatar. IOP Conf. Ser. Earth Environ. Sci. 2021,691, 124. [CrossRef]
10.
Lu, S.; Gao, X.; Li, W.; Jiang, S.; Huang, L. A study on the spatial and temporal variability of the urban residential water
consumption and its influencing factors in the major cities of China. Habitat Int. 2018,78, 29–40. [CrossRef]
11.
Hoyos, D.; Artabe, A. Regional Differences in the Price Elasticity of Residential Water Demand in Spain. Water Resour. Manag.
2017,31, 847–865. [CrossRef]
Water 2022,14, 1722 19 of 20
12.
Mayer, P.W.; DeOreo, W.B.; Opitz, E.M.; Kiefer, J.C.; Davis, W.Y.; Dziegielewski, B.; Nelson, J.O. Residential End Uses of Water.
1999. Available online: www.aquacraft.com (accessed on 7 March 2021).
13.
Balling, R.C.; Gober, P. Climate variability and residential water use in the city of Phoenix, Arizona. J. Appl. Meteorol. Climatol.
2007,46, 1130–1137. [CrossRef]
14. Sebri, M. A meta-analysis of residential water demand studies. Environ. Dev. Sustain. 2014,16, 499–520. [CrossRef]
15.
Nauges, C.; Whittington, D. Estimation of water demand in developing countries: An overview. World Bank Res. Obs.
2009
,25,
263–294. [CrossRef]
16.
Zapata, O. More Water Please, It’s Getting Hot! The Effect of Climate on Residential Water Demand. Water Econ. Policy
2015
,
1, 1550007. [CrossRef]
17.
Zubaidi, S.L.; Dooley, J.; Alkhaddar, R.M.; Abdellatif, M.; Al-Bugharbee, H.; Ortega-Martorell, S. A Novel approach for predicting
monthly water demand by combining singular spectrum analysis with neural networks. J. Hydrol.
2018
,561, 136–145. [CrossRef]
18.
Tong, X.; Brandt, M.; Yue, Y.; Ciais, P.; Rudbeck Jepsen, M.; Penuelas, J.; Wigneron, J.P.; Xiao, X.; Song, X.P.; Horion, S.; et al. Forest
management in southern China generates short term extensive carbon sequestration. Nat. Commun. 2020,11, 1–10. [CrossRef]
19.
MoWR. Ethiopian Water Sector Development Program. 2006. Available online: https://chilot.me/wp-content/uploads/2011/0
8/water-sector-development-program-vol-2.pdf (accessed on 9 December 2021).
20.
Awulachew, S.B.; International Water Management Institute. Water Resources and Irrigation Development in Ethiopia; International
Water Management Institute: Colombo, Sri Lanka, 2007.
21.
Tarfasa, S.; Brouwer, R. Estimation of the public benefits of urban water supply improvements in Ethiopia: A choice experiment.
Appl. Econ. 2013,45, 1099–1108. [CrossRef]
22.
Anteneh, Y.; Zeleke, G.; Gebremariam, E. Valuing the water supply: Ecosystem-based potable water supply management for the
Legedadie-Dire catchments, Central Ethiopia. Ecol. Process. 2019,8, 9. [CrossRef]
23.
WHO. Vision 2030: The Resilience of Water Supply and Sanitation in the Face of Climate Change. 2010. Available online:
https://apps.who.int/iris/handle/10665/70462. (accessed on 12 March 2021).
24.
Ali, M.; Terfa, A.B. State of Water Supply and Consumption in Urban Areas at Household Level: A Case Study of East Wollega
Zone, Ethiopia. Br. J. Hum. Soc. Sci. 2012,5, 2.
25.
Slavíková, L.; Malý, V.; Rost, M.; Petružela, L.; Vojáˇcek, O. Impacts of Climate Variables on Residential Water Consumption in the
Czech Republic. Water Resour. Manag. 2013,27, 365–379. [CrossRef]
26.
Praskievicz, S.; Chang, H. Identifying the Relationships between Urban Water Consumption and Weather Variables in Seoul,
Korea. Phys. Geogr. 2009,30, 324–337. [CrossRef]
27.
Renwick, M.E.; Green, R.D. Do Residential Water Demand Side Management Policies Measure Up? An Analysis of Eight
California Water Agencies. J. Environ. Econ. Manag. 2000,40, 37–55. [CrossRef]
28.
Jansen, A.; Schulz, C. Water demand and the urban poor: A study of the factors influencing water consumption among households
in Cape Town, South Africa. S. Afr. J. Econ. 2006,74, 593–609. [CrossRef]
29.
Lawens, M.; Mutsvangwa, C. Application of multiple regression analysis in projecting the water demand for the City of Cape
Town. Water Pract. Technol. 2018,13, 705–711. [CrossRef]
30.
Jadi, Z. Step by Step Explanation on Principal Component Analysis (PCA). 2021. Available online: https://builtin.com/data-
science/step-step-explanation-principal-component-analysis (accessed on 8 March 2020).
31.
Li, S.; Zhang, Q. Risk assessment and seasonal variations of dissolved trace elements and heavy metals in the Upper Han River,
China. J. Hazard. Mater. 2010,181, 1051–1058. [CrossRef]
32.
Abdi, H.; Williams, L.J. Principal component analysis: Principal component analysis. Wiley Interdiscip. Rev. Comput. Stat.
2010
,2,
433–459. [CrossRef]
33.
Gedefaw, M.; Hao, W.; Denghua, Y.; Girma, A. Variable selection methods for water demand forecasting in Ethiopia: Case study
Gondar town. Cogent Environ. Sci. 2018,4, 1537067. [CrossRef]
34.
Swain, S.S.; Mishra, A.; Sahoo, B.; Chatterjee, C. Water scarcity-risk assessment in data-scarce river basins under decadal climate
change using a hydrological modelling approach. J. Hydrol. 2020,590, 125260. [CrossRef]
35.
Zang, C.; Liu, J. Trend analysis for the flows of green and blue water in the Heihe River basin, northwestern China. J. Hydrol.
2013,502, 27–36. [CrossRef]
36.
MoWE. Urban Water Utilities Tariff Setting. Addis Abeba, Ethiopia. 2013. Available online: http://https://www.open.edu/
openlearncreate/mod/oucontent/view.php?id=80356§ion=4.1 (accessed on 8 June 2021).
37.
Mandefro, C.D.; Habtamu, K.M. Water supply and demand scenario of Dilla Town, Southern Ethiopia. Int. J. Water Resour.
Environ. Eng. 2017,9, 270–276. [CrossRef]
38.
Rathnayaka, K.; Malano, H.; Arora, M.; George, B.; Maheepala, S.; Nawarathna, B. Prediction of urban residential end-use water
demands by integrating known and unknown water demand drivers at multiple scales I: Model development. Resour. Conserv.
Recycl. 2017,117, 85–92. [CrossRef]
39.
Behailu, B.M.; Pietilä, P.E.; Katko, T.S. Indigenous Practices of Water Management for Sustainable Services: Case of Borana and
Konso, Ethiopia. Sage Open 2016,6, 2158244016682292. [CrossRef]
40.
Dimki´c, D. Temperature Impact on Drinking Water Consumption. In Proceedings of the 4th EWaS International Conference:
Valuing the Water, Carbon, Ecological Footprints of Human Activities, Online, 24–27 June 2020; MDPI: Basel, Switzerland, 2020.
[CrossRef]
Water 2022,14, 1722 20 of 20
41.
Hirvonen, K.; Sohnesen, T.P.; Bundervoet, T. Impact of Ethiopia’s 2015 drought on child undernutrition. World Dev.
2020
,
131, 104964. [CrossRef]
42.
Girma, Y.; Abebe, B. Isolation, identification and antimicrobial susceptibility of Listeria species from raw bovine milk in
Debre-Birhan Town, Ethiopia. J. Zoonotic Dis. Public Health 2018,2, 4.
43.
Kenney, D.S.; Goemans, C.; Klein, R.; Lowrey, J.; Reidy, K. Residential Water Demand Management: Lessons from Aurora,
Colorado. J. Am. Water Resour. Assoc. 2008,44, 192–207. [CrossRef]
44.
Gleick, P.H. The World’s Water Peak Water, China’s Growing Disaster, and Solutions to the World’s Water Crisis. 2009. Avail-
able online: http://www.wilsoncenter.org/event/report-launch-the-worlds-water-2008-2009-the-biennial-report-freshwater-
resources (accessed on 23 May 2022).
45.
Ao, C.O.; Waliser, D.E.; Chan, S.K.; Li, J.L.; Tian, B.; Xie, F.; Mannucci, A.J. Planetary boundary layer heights from GPS radio
occultation refractivity and humidity profiles: PBL height from GPS RO. J. Geophys. Res. Atmos. 2012,117, D16117. [CrossRef]
46. Cooper, R. Water Security beyond COVID-19; Institute of Development Studies: Brighton, UK, 2020.
