Science of the Total Environment 622–623 (2018) 1590–1602

Contents lists available at ScienceDirect

Science of the Total Environment

journal homepage: www.elsevier.com/locate/scitotenv

Designing water demand management schemes using a socio-technical

modelling approach
Sotiria Baki a,⁎, Evangelos Rozos a, Christos Makropoulos a,b
a
Department of Water Resources and Environmental Engineering, School of Civil Engineering, National Technical University of Athens, Athens, Greece
b
KWR Watercycle Research Institute, Nieuwegein, The Netherlands

H I G H L I G H T S G R A P H I C A L A B S T R A C T

• Addressing the socio-technical nature of

water conservation via technology
adoption.
• Integrated model built combining sys-
tem dynamics and urban water cycle
modelling.
• Socio-economic variability captured
through scenarios for main system
drivers.
• Importance of combining policies for
penetration of not widely used technol-
ogies.
• Flexible modelling tool for designing re-
silient water conservation strategies.

a r t i c l e i n f o a b s t r a c t

Article history: Although it is now widely acknowledged that urban water systems (UWSs) are complex socio-technical systems
Received 30 April 2017 and that a shift towards a socio-technical approach is critical in achieving sustainable urban water management,
Received in revised form 2 October 2017 still, more often than not, UWSs are designed using a segmented modelling approach. As such, either the analysis
Accepted 4 October 2017
focuses on the description of the purely technical sub-system, without explicitly taking into account the system's
Available online 6 November 2017
dynamic socio-economic processes, or a more interdisciplinary approach is followed, but delivered through rel-
Editor: Simon Pollard atively coarse models, which often fail to provide a thorough representation of the urban water cycle and hence
cannot deliver accurate estimations of the hydrosystem's responses. In this work we propose an integrated
Keywords: modelling approach for the study of the complete socio-technical UWS that also takes into account socio-eco-
Urban water system nomic and climatic variability. We have developed an integrated model, which is used to investigate the diffusion
Integrated modelling of household water conservation technologies and its effects on the UWS, under different socio-economic and cli-
Water conservation matic scenarios. The integrated model is formed by coupling a System Dynamics model that simulates the water
System dynamics technology adoption process, and the Urban Water Optioneering Tool (UWOT) for the detailed simulation of the
Technology adoption
urban water cycle. The model and approach are tested and demonstrated in an urban redevelopment area in Ath-
Socio-economic scenarios
ens, Greece under different socio-economic scenarios and policy interventions. It is suggested that the proposed
approach can establish quantifiable links between socio-economic change and UWS responses and therefore
assist decision makers in designing more effective and resilient long-term strategies for water conservation.
© 2017 Elsevier B.V. All rights reserved.

1. Introduction

⁎ Corresponding author at: Department of Water Resources and Environmental

Engineering, School of Civil Engineering, National Technical University of Athens,
Managing water demand is critical for ensuring the urban water
Heroon Polytechneiou 5, Zographou GR-157 80, Greece. system's (UWS's) sustainability in the long-term and a good level of ser-
E-mail address: [email protected] (S. Baki). vice under increasing environmental and social pressures, such as

https://doi.org/10.1016/j.scitotenv.2017.10.041
0048-9697/© 2017 Elsevier B.V. All rights reserved.
 S. Baki et al. / Science of the Total Environment 622–623 (2018) 1590–1602 1591

climatic changes and droughts, limited new unpolluted freshwater re- developed an integrated SD model of physical and socio-economic
sources, population increase and shifts, as well as changes in lifestyles catchment processes and Prodanovic and Simonovic (2010) created
(Butler and Memon, 2006). Demand management measures and poli- an integrated water resources model by coupling a hydrologic with a
cies can achieve considerable water conservation, thus tackling the ef- socio-economic model under a common SD modelling framework.
fects of water shortages, but even more importantly, contributing in Mereu et al. (2016) assessed a reservoir's operational resilience under
this way towards the longer term goal of sustainable water resources four socio-economic scenarios, while Kotir et al. (2016) developed an
management. The adoption of decentralised water-aware technologies SD model that incorporates both bio-physical and socio-economic pro-
is an effective demand side measure to achieve water conservation, cesses for analysing the effect of different policies. In the second case,
while at the same time it offers additional benefits, such as the decrease Baki et al. (2012) for example, have proposed a hybrid modelling frame-
of stormwater runoff and wastewater generation (Makropoulos and work, in which an ABM model for social simulation is linked with an
Butler, 2010). However, the implementation of decentralised options urban water model through an SD modelling platform, to investigate
introduces into the UWS a significant uncertainty, since the extent and the UWS's response to different socio-economic policies. Wang et al.
rate of the diffusion of such technologies cannot be estimated by tradi- (2014) assess the water ecological carrying capacity through an inte-
tional engineering methods, as it is closely related to water users' be- grated SD model that includes neural networks and a cellular automata
haviour and individual decisions, as well as the overall socio-economic (CA) – Markov chain model. Rozos et al. (2016) have coupled a CA and
context. an SD model to estimate the penetration of water-aware technologies
It is acknowledged that UWSs are complex socio-technical systems under different scenarios.
where all diverse elements are bound together (Sofoulis, 2005) and The aim of this work is the investigation of the adoption of water
that a shift towards a socio-technical approach is critical in achieving conservation technologies by domestic water users and the assessment
sustainable urban water management (Bos and Brown, 2012). Hence, of the associated UWS's response, under varying climatic and socio-eco-
the UWS cannot be considered and studied as a purely technical system, nomic conditions, as well as related policy interventions. To this end, we
isolated from the socio-economic elements, or even by only just includ- have developed an integrated simulation methodology and are explor-
ing static representations of the socio-economic context and processes. ing the use of SD as an integrating platform for simulating the urban
To be able to study a complex socio-technical system, such as the UWS, water socio-technical system. An innovation of this work with regards
in order to design effective policies and support decision makers more to other models of water technology adoption lies in the fact that the
effectively, an interdisciplinary approach is clearly required. Urban SD environment is coupled with a dedicated urban water management
water simulation tools focus exclusively on the representation of the tool that is able to provide an in-depth analysis and response of the UWS
technical sub-system without explicitly taking into account the system's and the process is investigated under varying socio-economic and cli-
dynamic socio-economic processes that are driving the entire system matic conditions. Specifically, the integrated model is a result of the cou-
(Baki et al., 2012). There is a need for integrated modelling tools that pling of i) an SD model that specifically simulates the water technology
are able to capture adequately the complex feedbacks and interrelation- adoption process by including associated socio-economic variables, in-
ships between the diverse but complementary system elements across terrelationships and processes and ii) the Urban Water Optioneering
the socio-technical systems divide (Makropoulos, 2014). Such an ap- Tool (UWOT) (Rozos and Makropoulos, 2013) that simulates the
proach based on the coupling of different models and tools covering UWS's purely technical component by delivering a detailed representa-
both technical and social components of the UWS would allow for ex- tion of the urban water cycle at the household level. Different socio-eco-
plicitly taking into account socio-economic drivers, capturing (to the ex- nomic scenarios are considered for exploring different possible futures
tent possible) their dynamic nature and complex interrelationships, and essentially providing system boundaries (Makropoulos et al.,
instead of just using static representations usually derived from litera- 2008). The scenarios are developed using an innovative methodology;
ture or expert judgement (Baki et al., 2012). Recently, a shift towards predefined evolution profiles are assigned to the identified socio-eco-
component-based modelling approaches is observed (Safiolea et al., nomic system drivers, the SD model's exogenous variables, which es-
2011) and several integrated socio-technical modelling applications sentially represent top-level national scale indicators. The proposed
have been emerging in the field of water management, explicitly ac- methodology is applied as a proof of concept to Eleonas redevelopment
counting for the socio-economic dimensions of the UWS. Koutiva and urban area in Athens, Greece, for investigating the adoption of water-
Makropoulos (2016) for example, have developed the Urban Water saving technologies in the residential sector under socio-economic
Agent Behaviour (UWAB) model, by integrating an Agent Based Model and climatic uncertainty. The diffusion of two different water technolo-
(ABM) that simulates water users' behaviour with an urban water man- gy bundles is investigated under specific policy interventions. The total
agement tool. Galán et al. (2009) built a hybrid ABM model, which also water conservation potential is quantified by the integrated model and
includes a technological diffusion sub-model, and coupled it with a geo- finally a sensitivity analysis is performed on selected SD model
graphic information system (GIS), while Schwarz and Ernst (2009) have variables.
simulated the diffusion of water-saving technologies and the impact of
different intervention policy tools. 2. Methodology
Within this context, of particular interest are approaches employing
a System Dynamics (SD) approach. SD models are especially attractive 2.1. Integrated model description
for the simulation of complex dynamic systems that are dominated by
complex interrelationships and feedback processes between different The integrated model created for the analysis of the socio-technical
system components and can combine elements from different scientific UWS is developed through the coupling of an SD model, which simu-
domains; for this reason they have been widely used for modelling a va- lates the adoption of water-saving technologies, and UWOT, which
riety of environmental systems (Ford, 2010). SD has also been used ex- simulates the urban water cycle. The penetration of water-saving tech-
tensively to simulate the adoption of innovation products including, for nologies under different socio-economic scenarios is estimated via the
example, the adoption of new technologies in the agricultural sector SD model, the results of which are then fed into UWOT in the form of
(Fisher et al., 2000) and the diffusion of renewable energy innovative time series with the number of households that adopt the technologies,
technologies in small islands (Gravouniotis, 2010). In water manage- in order to calculate the total potable water demand for each scenario.
ment, SD modelling has been employed in a wide range of applications Both models have a 30-year time horizon that is considered appropriate
(Winz et al., 2009) and has been used either as a unifying platform for for studying long-term socio-economic changes, as well as the technol-
both the social and technical subsystems or as the social system simula- ogy adoption process. For the SD model a yearly time step has been se-
tor of the water system. In the first case, Sušnik et al. (2012) for instance, lected, since the driving forces behind the model change at a relatively
1592 S. Baki et al. / Science of the Total Environment 622–623 (2018) 1590–1602

slow pace, in periods greater than one month. On the other hand the identified. Specifically, gross domestic product (GDP) per capita, sus-
urban water cycle model has been set up with a daily time step so tainability culture and price of water are defined externally by the
that it can capture variations in hydroclimatic conditions, i.e. in rainfall socio-economic scenarios and essentially provide input to the technolo-
and temperature, and therefore more accurately calculate the amount gy adoption model. Sustainability culture is a variable that acts as an in-
of water demand satisfied by household rainwater harvesting systems. dicator of the society's social values regarding the environment; in
The yearly time series of adopter and non-adopter households that are essence it reflects the environmental awareness of the population. The
calculated by the SD model are converted into daily time series for rest of the variables associated with the water technology adoption pro-
input into the urban water cycle model, assuming that the number of cess are considered to be endogenous system variables, i.e. they are de-
adopters remains constant throughout each calendar year. Climatic var- rived from within the system via interactions with other variables.
iability is taken into account by using synthetic hydroclimatic time se- Four general scenarios, within the lines of the work carried out by
ries for the simulation of the urban water cycle. A sensitivity analysis Makropoulos et al. (2008) and Casal-Campos et al. (2015), have been
is then performed on selected SD model parameters in order to account considered in this analysis. The time horizon of all four scenarios is
for the uncertainty in their estimation. The modelling process is 30 years, which is considered sufficiently long to study the evolution
summarised in Fig. 1. of water-saving technology penetration in the household sector. The
three first scenarios are the classical type scenarios literature usually
2.2. Socio-economic scenarios & modelling technology adoption considers, in which gradual and smooth transitions to the final state
were assumed. In contrast to these, in the last scenario sudden changes
The overall socio-economic context and policies in place play a crit- were incorporated, since it was considered quite interesting to analyse
ical role in the formation of households' socio-economic characteristics the effects of such transitions that are closer to current real-world
and environmental attitudes, and in general influence households' deci- socio-economic conditions, relevant to the case study, and monitor
sions on the adoption of water-saving technologies (Millock and the system's response when sudden shocks are applied to it. The main
Nauges, 2010; Grafton et al., 2011). For this reason, the penetration of qualitative characteristics of the four socio-economic scenarios in
such technologies in the domestic sector has been studied under the ef- terms of the model's exogenous variables are shown in Table 1 and
fect of different socio-economic scenarios. The scenarios used for the are analysed in detail within Appendix A.
current analysis do not represent forecasts, but rather a range of possi- Following the socio-economic scenario process, the aim was to iden-
ble futures in order to facilitate the exploration of a “possibility” space tify all other critical factors and drivers influencing the water technology
(Makropoulos et al., 2008). adoption process, as well as the interactions and causal relationships be-
The most significant socio-economic variables relevant to the water- tween them. In order to gain a more in-depth understanding of the
saving technology adoption process were highlighted and three main complex multi-parameter process of water-saving technologies pene-
socio-economic drivers, the system's main exogenous variables, were tration in households, as well as utilise as much as possible available in-
formation from literature, a more analytical approach was followed for
this work in which the process has been broken into three separate
stages: i) willingness to adopt, describing population transition towards
potential adopters, ii) willingness to pay, describing the transition of po-
tential adopters towards potential adopters with willingness to pay and
iii) technology adoption, describing the final transition towards
adopters. This breakdown can assist in understanding which influenc-
ing factors are most critical at each stage of the process and can there-
fore facilitate the design of appropriate and effective strategies and
policy interventions, the application of which can remove potential bar-
riers and thus speed up the diffusion of water-saving technologies.
The process modelled in this work via the SD methodology is the tran-
sition of households within a specific urban area from conventional water
technologies to water-aware technologies. At the start of the simulation it
is assumed that almost the entire household population has conventional
water technologies, apart from a few households that represent early
adopters of specific water conservation technologies. Fig. 2 shows the en-
tire causal loop diagram (CLD) developed for the specific technology
adoption process. The diagram is divided into three sections, representing
each of the three separate stages of the adoption process. The system's
exogenous variables, through which the socio-economic scenarios are
defined, have been marked in the CLD. The associated stock-and-flow di-
agram of the water technology adoption process is shown in Fig. 3. As it
can be observed, three of the CLD's variables form the stocks of the sys-
tem; potential adopters (Pot_Adopters), potential adopters with willing-
ness to pay (Pot_Adopters_Afford) and adopters (Adopters) and these are
supplied and drained by three flows (Pot_Adoption_rate, adopt_rate_1
and adopt_rate_2). The SD model was developed with the use of Vensim
DSS software (Ventana Systems, 2014).
The three stages of the water technology adoption process are de-
scribed in detail in the following paragraphs.

2.2.1. Willingness to adopt

The first stage of the adoption process describes the transition of a
part of the total population towards potential adopters (Pot_Adopters),
Fig. 1. Flow chart of integrated model. i.e. households that have a willingness to adopt a particular water-
 S. Baki et al. / Science of the Total Environment 622–623 (2018) 1590–1602 1593

Table 1
Developed socio-economic scenarios – exogenous model variables.

Exogenous variable Scenario 1 Scenario 2 Scenario 3 Scenario 4

GDP per capita Zero growth – steady-state Sustainable growth High growth (high) Volatile economy, periods of
economy (low) (medium) growth and deep recessions
Sustainability Culture / Environ. Awareness Very high sustainability High sustainability Medium-low sustainability Low sustainability
Water Price Medium high High Low Medium low

saving technology. The two key model variables for this transition are zero, but not on the right-hand side, suggests a positively skewed distri-
environmental awareness (Env_Awareness) and technology acceptabil- bution with a tale on the right-hand side. The percentage of the popula-
ity (Tech_Acceptability). The mean environmental awareness of the pop- tion, %_pop_EA, that has an environmental awareness above the set
ulation is defined by the broader sustainability level of the society that is threshold of EA_threshold, corresponds to the exceedance probability
described by the variable Sustainability_Culture. A household is consid- that the random variable of environmental awareness is higher than
ered to have the ability to become a potential adopter provided it has the specified threshold.
high environmental awareness (Millock and Nauges, 2010), higher The flow towards the stock of potential adopters, Pot_Adoption_rate
than a specified threshold (EA_threshold). In addition to environmental (par), is determined by the difference in population with willingness to
awareness, households form the intention to adopt a specific water adopt between two successive time steps of the SD model and is de-
technology when the technology is acceptable by them, according to scribed by Eq. (1).
their lifestyles and perceptions; this is especially important in the case
of alternative water sources. Social acceptance of a particular technolo- CCDF ðEAT; EAt Þ  TAct −CCDF ðEAT; EAt−1 Þ  TAct−1
par ¼  TPop ð1Þ
gy is influenced by the availability of specific information, related for ex- dt
ample to technology effectiveness and associated benefits, as well as by
the exposure to the technology itself (Hurlimann et al., 2009). There- where:
fore, not all environmentally aware households have willingness to
adopt a particular technology, but a fraction of them that is determined TPop total population (number of households) (Total_Population in
by the variable Tech_Acceptability. It is assumed that the acceptability of Fig. 3 in SD terminology)
a particular technology is influenced by the adopter ratio that indicates EAT lower threshold of Env_Awareness above which a household
the level of exposure; the more the households that have adopted the is considered to have high environmental awareness
technology, the more acceptable it is. This interrelationship can be (EA_threshold in Fig. 3)
seen in the CLD (Fig. 2) as a positive reinforcing loop connecting EAt population's mean environmental awareness at time t
Adopters and Tech_Acceptability. This is actually a “learning loop” driving (Env_Awareness in Fig. 3)
the system as the adoption process proceeds (Gravouniotis, 2010). TAct acceptability of a specific technology at time t
It is assumed that the environmental awareness of households fol- (Tech_Acceptability in Fig. 3)
lows a lognormal probability distribution with a mean value m that is CCDF exceedance probability of EA_threshold for lognormal distri-
defined by the value of the Env_Awareness variable. The lognormal dis- bution
tribution has been selected, since it has been found to fit closely data in dt time step of the simulation (year) (TIME STEP in Fig. 3)
many cases of natural and social systems, when values cannot be nega-
tive and there are indications for a non-symmetrical, skewed distribu- Tech_Acceptability is a function of Adopter_ratio and the WTA variable
tion (Limpert et al., 2001). The fact that the distribution is bounded by represents the households that are willing to adopt a specific

Fig. 2. CLD for the adoption process of water conservation technologies with policy intervention (price subsidy).
1594 S. Baki et al. / Science of the Total Environment 622–623 (2018) 1590–1602

Fig. 3. Stock-and-flow diagram for the adoption process of water conservation technologies.

technology; it is essentially a fraction of the households with high envi- been taken into account, as it is considered the most significant factor
ronmental awareness. for the particular stage of the process.
The interrelationships between variables influencing the second
stage of the water technology adoption process can be seen in the CLD
WTAt ¼ TAct  perEA  TPop ð2Þ of Fig. 2, as well as in Fig. 3. For potential adopters to have the necessary
willingness to pay for a specific technology, their disposable income
(HH_Disposable_Income) needs to result in a willingness to pay higher
where:
than the market price of the specific technology (Tech_price). The dis-
posable income of the household is a function of the society's economic
perEA percentage of the population with high environmental
development, which can be represented by the GDP per capita
awareness (%_pop_EA in Fig. 3)
(GDPcapita_change). The disposable income financial indicator is gener-
ally closer to the concept of income and is considered to be a more rep-
resentative indicator of the population's material well-being and
2.2.2. Willingness to pay
purchasing power than GDP per capita (OECD, 2014).
During the second stage of the adoption process the transition of po-
The flow towards potential adopters with willingness to pay,
tential adopters (Pot_Adopters) towards potential adopters with will-
adopt_rate_1 (ar1), depends on the difference in population with will-
ingness to pay (Pot_Adopters_Afford) is described. Willingness to pay
ingness to pay higher than the technology's market price between two
can be defined as the monetary value a person sets on a product or a ser-
successive time steps of the SD model and is described by Eq. (3). As
vice (Piper, 2014). A social group's willingness to pay for a water-saving
in the case of environmental awareness, it is assumed that the willing-
technology may be influenced by various socio-economic factors, such
ness to pay of the households with high environmental awareness fol-
as the general social and demographic characteristics of the population
lows a lognormal probability distribution with a mean value m that is
(age, gender, educational level, environmental awareness, etc.), as well
defined by the value of the WTP variable. The percentage of the environ-
as other factors including income, ownership status, perceptions on
mentally aware population, %_pop_affords, that has willingness to pay
drought and water conservation, perception and information availabili-
higher than the technology price (Tech_price), corresponds to the ex-
ty on specific technologies, retail price of water, etc. (Berk et al., 1995;
ceedance probability that the random variable of willingness to pay is
Tapsuwan et al., 2014; Piper, 2014; Garcia Alcubilla and Lund, 2006). Ac-
higher than the specified threshold.
cording to recent econometric studies that have been carried out on the
adoption of household water-saving technologies (Piper, 2014;
Tapsuwan et al., 2014) the correlation between a household's income CCDF ðTPr t ; WTP t Þ  WTAt −CCDF ðTPr t−1 ; WTPt−1 Þ  WTAt−1
ar1 ¼ ð3Þ
and its willingness to pay is statistically significant. In the current anal- dt
ysis only the effect of the average household disposable income
(HH_Disposable_Income) on the average willingness to pay (WTP) has where:
 S. Baki et al. / Science of the Total Environment 622–623 (2018) 1590–1602 1595

WTAt population with willingness to adopt (number of house- to the left, which describes the effect of market saturation, is a negative
holds) (WTA in Fig. 3) feedback loop that gives a balancing effect to the system slowing down
WTPt mean willingness to pay of the environmentally aware popu- the adoption rate as the pool of potential adopters with willingness to
lation (euro) (WTP in Fig. 3) pay is gradually depleted by technology adoption. The flow towards
TPrt market price of a particular water-saving technology (euro) the adopters stock, adopt_rate_2 (ar2), which in essence simulates the
(Tech_price in Fig. 3) final technology adoption decision, is described by Eq. (7).

The population's mean willingness to pay for a water-saving tech- ar2 ¼ ar  PAA  adr  TAt ð7Þ
nology is calculated according to the following equation based on a lin-
ear relationship between natural logarithms (Piper, 2014; Jacobsen and where:
Hanley, 2008).
ar adoption rate of the technology from potential adopters
WTP t ¼ DIa1  ea0 ð4Þ (year−1) (Adoption_rate in Fig. 3)
PAA potential adopters with willingness to pay (number of house-
where: holds) (Pot_Adopters_Afford in Fig. 3)
adr ratio of adopters to the total population (Adopter_ratio in Fig.
a0, a1 coefficients 3)
DI mean disposable household income calculated according to TAt attractiveness of the technology (Tech_Attractiveness in Fig.
Eq. (5) (euro) (HH_Disposable_Income in Fig. 3) 3); a function of Payback_period (PP)

The payback period of a specific water technology is calculated ac-

DI ¼ ð1 þ DIcÞ  RDI ð5Þ
cording to Eq. (8):
where:
TRr t
PP ¼ ð8Þ
RDI mean reference disposable household income at the begin- AWS  WP
ning of the simulation (euro) (Ref_Disp_Income in Fig. 3)
where:
DIc household disposable income change calculated according to
Eq. (6) assuming a linear relationship between GDP and
AWS annual water savings at a household level due to the imple-
household income (HH_Disposable_Income_change in Fig. 3).
mentation of a specific water-saving technology (m3/year)
(Annual_water_savings in Fig. 3)
b1  ð100 þ GDPcc  100Þ−100 WP mean price of water per unit volume (euro/m3) (Water_price
DIc ¼ ð6Þ in Fig. 3)
100

where: In addition to the adoption model described, the effects of a policy

intervention, providing subsidies that lower the technology price paid
GDPcc GDP per capita annual change (GDPcapita_change in Fig. 3) by consumers, were also studied. Specifically, when the percentage of
b1 coefficient adopters is low the subsidy is activated reducing thus technology
price and as a result more households are able to afford to implement
the intervention (Fig. 2). As the adoption process progresses and the
2.2.3. Technology adoption adopter ratio increases above a certain threshold, the subsidy policy
The final adoption stage is in line with the model developed by Bass ends or is substituted by a lower price subsidy. As a result, the adoption
(1969) to describe the adoption and diffusion of a product via innova- process slows down and this is also demonstrated by the additional
tion and imitation mechanisms. The last part of the SD model focuses negative feedback loop in the CLD.
on imitation via word-of-mouth, whereas technology adoption via in-
novation is not modelled explicitly. During this last stage, the final deci- 2.3. Modelling the urban water cycle
sion to go through or not with the adoption of a particular water-saving
technology is taken by the households in the second stock of the model. The urban water cycle at the household level is simulated by UWOT,
This transition from potential adopters with willingness to pay a tool that follows a metabolism modelling approach and has the ability
(Pot_Adopters_Afford) to actual adopters (Adopters) is driven by the to simulate the entire urban water cycle, including water use, abstrac-
information transmission and imitation mechanism, and is also influ- tion from water sources, water conveyance and distribution, water
enced by the attractiveness of the specific technology. The attractive- and wastewater treatment, water reuse and discharge (Rozos and
ness of a specific technology (Tech_Attractiveness) is a function of its Makropoulos, 2013; Bouziotas et al., 2015). UWOT is a demand-orient-
price (Tech_price), the price of water (Water_price), as well as of the an- ed tool; instead of actual water flows being simulated, demand signals
nual household water savings as a result of the implementation of the for potable water, runoff and wastewater are generated by the various
technology (Annual_water_savings), all of which determine the payback model components and are then aggregated and transmitted back to
period of the original investment (Payback_period). The smaller the pay- the water source, receiving water body or treatment plant, accordingly.
back period is, the more attractive a specific investment is. Water consumption within the household is simulated at the water
As shown in Fig. 2 in the last part of the CLD two feedback loops are appliance level and all the required information for water appliances
formed that are representative of the adoption process as described by and technologies is stored in the “technology library”, the tool's data-
Sterman (2001). The feedback loop to the right, which describes the in- base (Rozos and Makropoulos, 2013). UWOT can incorporate both con-
formation transmission (word-of-mouth) and imitation process, is a ventional and water-saving appliances, as well as centralised and
self-reinforcing feedback loop that drives the system and gives it the distributed water technologies (Rozos et al., 2010), such as rainwater
power to grow. An increase in the number of adopters, results in a harvesting and water-recycling schemes (Rozos et al., 2013). It is there-
higher adoption rate since encounters rise and word-of-mouth becomes fore ideal for the analysis of the effects on the UWS stemming from the
stronger; this leads to a further increase in adopters. The feedback loop adoption of different water-aware technologies.
1596 S. Baki et al. / Science of the Total Environment 622–623 (2018) 1590–1602

In the current work that investigates the penetration of water con- The model was run for the period 2015–2045 with a daily time step.
servation technologies in the domestic sector, UWOT is used to simulate The hydroclimatic data used in the model include daily rainfall and tem-
the household demand within the urban water cycle. Adopter and con- perature data originating from synthetic time series for the specific area
ventional time series calculated by the SD adoption model feed into the created via the stochastic weather generator WeaGETS (Chen et al.,
water efficient and conventional household multipliers (see Fig. 5), 2012). The model was also simulated with only conventional house-
respectively. holds, i.e. assuming zero penetration of water conservation technolo-
gies, in order to obtain the business as usual (BAU) scenario for
3. Integrated model application comparison reasons and estimation of the resulting potable water
savings.
3.1. Case study general description
3.3. SD & socio-economic parameters
The integrated modelling methodology was applied in Eleonas
urban area in Athens as a proof of concept. The area of Eleonas is consid- The exogenous and endogenous parameters of the SD water tech-
ered an “urban void” and has been in the centre of redevelopment plans nology adoption model have been estimated according to various differ-
during recent years. The area has significant potential for urban and res- ent assumptions. The main parameter values and assumptions are listed
idential development and numerous urban planning design proposals in summary in Table 2 and described in detail in Appendix A.
have been carried out to this end (Rozos and Makropoulos, 2014). For The policy intervention was designed specifically for the promotion
the current modelling experiment an existing urban development mas- of the second technology bundle, which includes the installation of the
ter plan was selected that is characterised by substantial residential de- rainwater harvesting system, due to the high cost of the particular mea-
velopment. Only the domestic sector has been taken into account and sure. Specifically, a price subsidy of 1500 euros is activated when the
modelled in UWOT, since the aim of this analysis is the investigation adopter ratio is lower than 20% of the total household population and
of the penetration of water conservation technologies in households. a subsidy of 1000 euros is activated when the adopter ratio is lower
The selected urban development plan includes different household than 50%. When the adopter ratio is higher than 50% the subsidy is
types that are modelled explicitly in both the urban water cycle model deactivated. Additionally, the policy includes the full funding of certain
and the SD model. households for the installation of the water conservation bundle at the
beginning of the simulation. The initial adopters in the case of the sin-
3.2. Urban water cycle model schematisation & parameters gle-detached households have been set equal to 3.5% of the total popu-
lation. It should be noted that this percentage represents a threshold,
The urban water cycle model developed in UWOT includes only do- below which the adoption process in the particular model does not
mestic water users that reside in four different household types. The pick-up. This is in itself a significant benefit of the model, as it demon-
total 1959 households are distributed as follows: i) 289 single-detached strates its ability to provide useful insights in designing effective
houses, ii) 1152 large enclosed courtyard apartments, iii) 192 small policies.
enclosed courtyard apartments and iv) 326 tower apartments. The
same household size has been assumed for all different household
4. Results and discussion
types and the occupancy has been set equal to 2.5 residents per house-
hold, which is representative for the Athens area.
The integrated model was run in total for the period 2015–2045,
For each household type different sets of interventions aiming at po-
with a yearly time step for the SD model and a daily time step for the
table water conservation were selected. For the three household types
urban water cycle model. Below are presented initially the intermediate
corresponding to apartments, that have a relatively limited outside
results from the SD model, concerning technology adoption, and sec-
space, the proposed intervention includes the substitution of selected
ondly the final results of the integrated model on water consumption
conventional household appliances with low-water consumption
and accomplished water conservation levels under all socio-economic
ones. Specifically, the first intervention bundle includes:
scenarios, as calculated by UWOT.
• Dual-flush conversion mechanism (application to existing conven-
tional toilets) 4.1. SD model results
• Water efficient faucets for the wash basin and kitchen sink
• Efficient shower head The final output of the SD model is the penetration level of each
• Drip irrigation system water-saving technology bundle in the relevant household types
throughout the 30-year horizon and under the four socio-economic sce-
narios. These results are presented below for each intervention bundle.
The second intervention bundle, applicable to single-detached hous-
es that have large outside areas and gardens, as well as big roof surfaces, 4.1.1. First intervention bundle
additionally includes the installation of small/medium sized rainwater Fig. 6a shows the penetration of water efficient appliances in Eleonas
harvesting systems, which are used by the households as an alternative apartment households under the four socio-economic scenarios. It can
water source for both garden irrigation and toilet flushing. be observed that the resulting adoption curve has the form of the char-
Two alternative household categories for each of the four household acteristic S-curve or S-shaped growth, a system behaviour that is very
types have been created in the Eleonas domestic sector urban water common in natural systems (Ford, 2010). The first part of the curve re-
cycle model; a conventional household and a household incorporating sembles exponential growth, where there is a mild increase in the early
the corresponding intervention bundle. Conventional households that stages, but as adoption progresses the positive loops of the system gain
only include conventional water appliances and technologies corre- more strength and adoption accelerates. Then the system encounters
spond to users that have not yet adopted a water conservation interven- specific limitations, as the number of potential adopters that afford the
tion, whereas the second category corresponds to adopters of water technology gradually decreases and the word-of-mouth mechanism be-
conservation technologies. The resulting overall model schematisation comes less effective. Adoption slows down in the second part of the
in UWOT is shown in Fig. 4, whereas Fig. 5 shows a more detailed curve, as the negative feedback loop starts dominating the system and
view of the model schematization for the single-detached houses with the positive feedback loop weakens. The system gradually moves to-
water efficient appliances and a rainwater collection system. wards a state of equilibrium in which the technology adoption is in
 S. Baki et al. / Science of the Total Environment 622–623 (2018) 1590–1602 1597

Fig. 4. UWOT schematisation of Eleonas domestic urban water cycle model.

balance with its “carrying capacity” in the social and economic environ- with sufficient willingness to pay. The SD model under Scenario 4 has
ment (Sterman, 2001). a quite different behaviour than the rest scenarios due to the distinct
Technology penetration has a quite similar behaviour under Scenar- form of the socio-economic scenario variables. The mean environmental
ios 1 and 2, in which technology adoption picks up quite early and high awareness is decreasing and the GDP per capita is characterised by
percentages of adoption are achieved by the end of the simulation peri- abrupt changes instead of gradual transitions; as a result the first two
od (Table 3). In both these scenarios there is a rapid uptake in the first flows get negative values. Consequently, final adoption levels are signif-
time steps of the simulation due to the high increase of environmental icantly lower under this scenario (Table 3).
awareness, which causes a significant flow towards potential adopters
that then stabilises. The flow towards potential adopters that afford 4.1.2. Second intervention bundle
gradually decreases as GDP stabilises. The final results in the two scenar- Fig. 6b shows the penetration of water efficient appliances together
ios are quite similar because the different characteristics of each scenar- with the rainwater harvesting system in Eleonas single-detached
io eventually balance out. In Scenario 3 there are significantly less households under the four socio-economic scenarios. It can be observed
potential adopters due to the decreased environmental awareness. that the resulting adoption curve resembles the early stages of the S-
However, even though technology uptake progresses at a slower pace shaped curve, but not the entire curve is formed within the simulation
and its final value is lower than the first two scenarios (Table 3), adop- period, since the positive loop is not as strong as in the case of the first
tion is still relatively high due to the strong economic growth and low intervention bundle. The SD model showed that without the presence
technology price, which result in an increased number of households of the subsidy policy tool the adoption process is so slow that this

Fig. 5. UWOT schematisation of the single-detached household (conventional & efficient).

1598 S. Baki et al. / Science of the Total Environment 622–623 (2018) 1590–1602

Table 2
Main values and assumptions of Eleonas SD model exogenous and endogenous variables.

Variable name Values Source/assumptions

SD exogenous variables (socio-economic scenarios)

Sustainability Culture 5.70 (Baseline) Koutiva and Makropoulos (2015) - baseline
7.20 (Scenario 1) Harju-Autti and Kokkinen (2014) - scenarios
6.40 (Scenario 2)
5.90 (Scenario 3)
5.35 (Scenario 4)
GDP per capita change (cumulative) (%) 37.9 (Scenario 1) OECD (2015a) & Eurostat (2015) - Scenario 3; OECD (2012) - Scenarios 1&2;
81.5 (Scenario 2) The World Bank (2015) - Scenario 4
153.2 (Scenario 3)
62.4 (Scenario 4)
Water price (euro/m3) 1 (Baseline) EYDAP (2015) - baseline; OECD (2010) - scenarios
2.6 (Scenario 1)
3.2 (Scenario 2)
1.4 (Scenario 3)
2.2 (Scenario 4)
SD endogenous variables
Household Disposable Income (euro/hh) 28,000 (apartments) OECD (2015b) - reference value
50,000 (single-detached houses)
OECD (2015c) - relationship between disposable income and GDP per capita
Eq. (A.1) (Appx. A) and Fig. A.2 (Appx. A)
Willingness to pay (euro) Eq. (A.2) (Appx. A) (first intervention bundle) Piper (2014) – a1
Eq. (A.3) (Appx. A) (second intervention bundle) Tapsuwan et al. (2014) – a0 (1st intervention bundle)
Vernardakis (2014) - a0 (2nd intervention bundle)
Technology price (euro) 230 (1st intervention bundle) Current market prices and Tapsuwan et al. (2014)
2800 (2nd intervention bundle)
Annual water savings (m3) 54 (1st intervention bundle) UWOT calculations
133 (2nd intervention bundle)
Tech_Acceptability (%) 85–100 (1st intervention bundle) Assumptions & Vernardakis (2014)
30.8–100 (2nd intervention bundle)
Tech_Attractiveness 0.2–1.0 Assumptions
Adoption_rate (year−1) 0.65 Bass (1969)
Initial_adopters (%) 2.5% (1st intervention bundle) Rogers (2003)
3.5% (2nd intervention bundle) Assumption (policy)

particular intervention bundle would never actually pick up under any water prices, which result in a low technology attractiveness. In the
socio-economic scenario due to its high cost. case of Scenario 4 the reduction in environmental awareness combined
As can be seen in Fig. 6b the SD model has a similar response under with the economic instability and long recession periods lead to a very
Scenarios 1 and 2. Technology penetration is quite lower (Table 3) than slow technology uptake.
for the first technology bundle, mainly due to the technology's low initial
acceptability and high price. However, as adoption picks up 4.2. Urban water cycle model results
Tech_Acceptability increases considerably and so does Tech_Attractiveness
due to the rising water prices. When adoption levels reach 20% the sub- Using the above technology adoption results UWOT then estimated
sidy amount reduces but there is no noticeable effect on technology pen- the possible evolution of total potable water demand in Eleonas under
etration. This means that even a reduced subsidy is sufficient since the the four socio-economic scenarios, as well as the baseline scenario, in
penetration achieved that far has resulted in an increase of the initially which zero adoption is assumed (Fig. 7a). The temporal evolution of
low technology acceptability. When adoption levels supersede 50% of water conservation achieved in comparison to the baseline scenario
the households, the subsidy stops entirely and the sharp increase in tech- under each socio-economic scenario is shown in Fig. 7b.
nology price causes the adoption to stop altogether.
The adoption process under Scenarios 3 and 4 is quite slow and the 4.3. Climatic variability
final technology penetration rates achieved within the examined time
frame are very low in comparison to the other scenarios and technolo- The integrated model results presented above were estimated using
gies (Table 3). In Scenario 3 this slow adoption is mainly caused by a single set of hydroclimatic time series. In order to include the effect of
the relatively low environmental awareness in combination with low hydroclimatic variability on water conservation levels synthetic time

Fig. 6. Modelled adoption of a) first intervention bundle and b) second intervention bundle, under four socio-economic scenarios.
 S. Baki et al. / Science of the Total Environment 622–623 (2018) 1590–1602 1599

Table 3
Final technology adoption rates and annual water savings in 30 years under four socio-economic scenarios and sensitivity analysis.

Scenario 1 Scenario 2 Scenario 3 Scenario 4

Final adoption (%)

First technology bundle 72% 68% 62% 44%
Second technology bundle 53% 52% 28% 19%
Total technology adoption 69% 65% 57% 41%
Final adoption (%) - sensitivity analysis
First technology bundle 61%–77% 58%–71% 48%–65% 34%–50%
Second technology bundle 11%–58% 3%–56% 3%–43% 3%–32%
Total technology adoption 54%–74% 50%–69% 42%–62% 29%–47%
Eleonas annual water savings (%)
Mean 19.59% 18.56% 15.42% 10.85%
Range (climatic variability) 18.65%–22.20% 17.69%–21.04% 14.74%–17.38% 10.37%–12.22%
Range (sensitivity analysis) 13.09%–23.92% 11.85%–22.41% 9.96%–19.52% 7.02%–14.88%

series were used for rainfall and temperature. Specifically, the urban The technology adoption SD results from the sensitivity analysis
water cycle model was run 30 times using different hydroclimatic in- were converted into daily time series for input into the Eleonas urban
puts. The same time series for technology adoption were used as in water cycle model. UWOT was then run with the synthetic
the original model since the SD model remained unaffected. The final hydroclimatic time series sets to also take into account climatic variabil-
results after all synthetic model runs are shown in summary in Table 3 ity. Water conservation under all socio-economic scenarios was esti-
and in Fig. 8, in which the lines represent the mean estimated water mated via UWOT and the final results of the integrated model are
conservation and the shaded areas the confidence bounds due to the ef- presented in Table 3 and Fig. 9b. In Fig. 9b the lines represent the
fect of climatic variability. The difference in water conservation is mean estimated water conservation for the initial SD parameters and
caused mainly by the difference in rainwater collection, but also the dif- the shaded area the confidence bounds reflecting climatic variability
ference in garden irrigation needs. and the SD parameter sensitivity analysis. The final confidence bounds,
after performing the sensitivity analysis, are considerably wider in com-
5. Socio-economic parameter sensitivity analysis parison to Fig. 8, for which only climatic variability has been taken into
account.
Following the simulation of the integrated model for the residential
sector of Eleonas, a sensitivity analysis was performed on selected vari- 6. Discussion and conclusions
ables of the SD model. Many SD variables included in the model have a
considerably high degree of uncertainty associated with them, since This work presented an integrated modelling methodology for in-
these are socio-economic variables for which there is a lack of specific vestigating the diffusion of water conservation technologies in the do-
data and studies to assist in their estimation. Sensitivity analysis has mestic water sector. The study of the socio-technical UWS is facilitated
been employed in order to deal with this inherent uncertainty and by the coupling of an SD model for the representation of the socio-eco-
quantify its effect on technology adoption. It is also a useful methodolo- nomic system and simulation of the technology adoption process, with
gy for identifying the most critical model parameters that should be the an urban water cycle model for the simulation of the technical compo-
focus of further research and site specific studies and surveys, so that nent of the UWS. An important feature of the methodology employed
these can be estimated more precisely, reducing therefore model uncer- is that it involves the explicit simulation of socio-economic scenarios
tainty for future model applications. For this purpose, Vensim's built-in through the SD modelling environment, instead of using static repre-
sensitivity analysis - Monte Carlo functionality was used and five main sentations common in scenario analysis applications, by incorporating
variable categories were selected for the analysis (see Appendix B). interrelationships between national scale indicators and socio-econom-
Sensitivity analysis was performed for all four socio-economic sce- ic factors influencing the UWS. The advantage of this approach is that
narios and the corresponding final results were produced (Fig. 9a and various socio-economic scenarios can be tested, in order to quantify
Table 3). As indicated by the results (Table 3 and Appendix B), there is the UWS's response, by only modifying the model's exogenous drivers.
big uncertainty especially surrounding the penetration of the second The proposed integrated simulation methodology follows therefore a
technology bundle. In fact, the results in some scenarios indicate that top-down approach, in that socio-economic scenarios are set-up
there is the possibility that very little or even no technology penetration through high-level variables at national scale, which then drive the SD
is going to be achieved within the examined time horizon. water technology adoption module that estimates technology diffusion

Fig. 7. Temporal evolution of a) total annual water demand and b) annual water demand conservation in Eleonas under four socio-economic scenarios.
1600 S. Baki et al. / Science of the Total Environment 622–623 (2018) 1590–1602

Fig. 8. Temporal evolution of annual water demand conservation in Eleonas under four socio-economic scenarios and with hydroclimatic confidence bounds.

at the household level under different socio-economic contexts and ap- widely-used and accepted technologies, as in the case of rainwater
plied policy instruments. Outputs of the technology adoption module harvesting systems, whose diffusion would not be otherwise self-
feed into the dedicated urban water cycle model that in turn calculates sustained.
water consumption and water savings associated with the use of water- • Model results further demonstrate the important effect of water pric-
aware technologies. ing on water technology adoption by households and that low water
The integrated model was tested in Eleonas redevelopment project. prices may not provide sufficient incentives to conserve water
The developed SD model exhibits an overall reasonable behaviour, since through a not so attractive and profitable technology investment.
it produces results that we would anticipate in the actual UWS. The fol- This factor is especially critical for more expensive technologies that
lowing insights and conclusions have been reached through the inte- their implementation does not accomplish very high annual water
grated analysis of water technology adoption in UWSs: savings, since low water prices result in even longer payback periods.
• There is significantly higher uncertainty related to the diffusion of less
• The analysis highlights the importance of policy tools, such as price accepted and more expensive water technologies. This provides an in-
subsidies, as their effect can be critical for the penetration of not sight into the need for further strengthening the adoption process

Fig. 9. a) Modelled adoption of all water conservation interventions under four socio-economic scenarios with sensitivity analysis confidence bounds (top graphs) and b) temporal
evolution of annual water demand conservation in Eleonas under four socio-economic scenarios with hydroclimatic and sensitivity analysis confidence bounds (bottom graphs).
 S. Baki et al. / Science of the Total Environment 622–623 (2018) 1590–1602 1601

through the use of additional policy instruments, such as information References

campaigns on the benefits, application and use of particular water
conservation technologies. Baki, S., Koutiva, I., Makropoulos, C., 2012. A hybrid artificial intelligence modelling frame-
work for the simulation of the complete, socio-technical, urban water system. 6th In-
• The analysis also identified the most critical adoption model parame- ternational Congress on Environmental Modelling and Software (iEMSs 2012), 1-5
ters that should be the focus of further research and targeted studies, July 2012, Leipzig, Germany.
in order to estimate them more precisely, and therefore reduce model Bass, F.M., 1969. A new product growth for model consumer durables. Manag. Sci. 15 (5):
215–227. https://doi.org/10.1287/mnsc.15.5.215.
uncertainty for future model applications. Specifically, it is indicated Berk, R.A., Cameron, T., Schulman, D., 1995. Willingness to Pay for Household Water
that adoption rate plays an important role for expensive and not Saving Technology in Two California Service Areas. University of California Water
widely used technologies, whereas technology acceptability seems Resources Center, California, USA.
Bos, J.J., Brown, R.R., 2012. Governance experimentation and factors of success in socio-
to be critical for all technologies. Additionally, the relationship be- technical transitions in the urban water sector. Technol. Forecast. Soc. Chang. 79
tween household disposable income and willingness to pay, as well (7):1340–1353. https://doi.org/10.1016/j.techfore.2012.04.006.
as the statistical characteristics of the employed probability distribu- Bouziotas, D., Rozos, E., Makropoulos, C., 2015. Water and the city: exploring links be-
tween urban growth and water demand management. J. Hydroinf. 17 (2):176–192.
tions (environmental awareness, willingness to pay), should be esti- https://doi.org/10.2166/hydro.2014.053.
mated via site specific surveys and econometric studies. Butler, D., Memon, F.A. (Eds.), 2006. Water Demand Management. IWA Publishing, Lon-
don, UK.
Casal-Campos, A., Fu, G., Butler, D., Moore, A., 2015. An integrated environmental assess-
ment of green and gray infrastructure strategies for robust decision making. Environ.
The proposed integrated tool could be used in future applications as a Sci. Technol. 49 (14):8307–8314. https://doi.org/10.1021/es506144f.
testing ground for different policies with the aim to compare their effec- Chen, J., Brissette, F.P., Leconte, R., Caron, A., 2012. A Versatile Weather Generator for Daily
tiveness. It could be used for example to investigate different subsidy Precipitation and Temperature. Trans. ASABE 55 (3):895–906. https://doi.org/
10.13031/2013.41522.
schemes, but also even different type of policy tools that target different Eurostat, 2015. Eurostat Database, Population projections – Main scenario. Available at:.
parts of the adoption process, such as awareness raising and information http://ec.europa.eu/eurostat/data/database, Accessed date: 15 October 2015.
campaigns promoting particular technologies or water pricing policies, EYDAP, 2015. EYDAP S.A. water tariffs. Available at:. http://www.eydap.gr/, Accessed
date: 20 October 2015.
or even policy mixes. The water technology adoption module can be Fisher, D.K., Norvell, J., Sonka, S., Nelson, M.J., 2000. Understanding technology adoption
modified accordingly to accommodate such type of analysis by including through system dynamics modeling: implications for agribusiness management.
additional variables and interrelationships. Overall, it is suggested that Int. Food Agribusiness Manag. Rev. 3 (3):281–296. https://doi.org/10.1016/S1096-
7508(01)00048-9.
the proposed methodology offers a flexible modelling tool, in that the in- Ford, A., 2010. Modeling the Environment. Second edition. Island Press, Washington DC,
tegrated model could be further expanded, both in terms of the SD and USA.
the urban water cycle sub-models, by including additional components Galán, J.M., López-Paredes, A., del Olmo, R., 2009. An agent-based model for domestic
water management in Valladolid metropolitan area. Water Resour. Res. 45 (5),
and relationships as new information and data are becoming available. W05401. https://doi.org/10.1029/2007WR006536.
In this way factors that have not been accounted for could be incorporat- Garcia Alcubilla, R., Lund, J.R., 2006. Derived willingness-to-pay for household water use
ed in the future; for example, information on water availability could be with price and probabilistic supply. J. Water Res. Plann. ASCE 132 (6):424–433.
https://doi.org/10.1061/(ASCE)0733-9496(2006)132:6(424).
taken into account for shaping environmental awareness or willingness
Grafton, R.Q., Ward, M.B., To, H., Kompas, T., 2011. Determinants of residential water con-
to adopt a water conservation technology. Even more interesting could sumption: Evidence and analysis from a 10-country household survey. Water Resour.
be the extension of the integrated methodology to include feedbacks Res. 47 (8), W08537. https://doi.org/10.1029/2010WR009685.
from the urban water cycle model back to the SD model. Gravouniotis, P., 2010. Seasonal power peaking and the diffusion of demand-side technol-
ogies: modelling socio-economic & technical dynamics in the Greek Islands. (Ph.D.
This methodology attempts to establish quantifiable links between thesis). Imperial College London, London, UK.
socio-economic drivers that influence the UWS system's response and Harju-Autti, P., Kokkinen, E., 2014. A novel environmental awareness index measured
the dynamics and feedbacks of the system's technological evolution. Al- cross-nationally for fifty seven countries. Univers. J. Environ. Res. Technol. 4 (4),
178–198.
though clearly the method does not capture the complexity of these in- Hurlimann, A., Dolnicar, S., Meyer, P., 2009. Understanding behaviour to inform water
teractions in their entirety, it is suggested that this type of analysis can supply management in developed nations – a review of literature, conceptual
assist decision and policy makers in designing more effective demand model and research agenda. J. Environ. Manag. 91 (1):47–56. https://doi.org/
10.1016/j.jenvman.2009.07.014.
management strategies. This is because such strategies, as argued Jacobsen, J.B., Hanley, N., 2008. Are there income effects on global willingness to pay for
throughout this work, typically possess both engineering and socio-eco- biodiversity conservation? Environ. Resour. Econ. 43 (2):137–160. https://doi.org/
nomic characteristics, constraints and success caveats and as such, re- 10.1007/s10640-008-9226-8.
Kotir, J.H., Smith, C., Brown, G., Marshall, N., Johnstone, R., 2016. A system dynamics sim-
quire design tools able to account for both. ulation model for sustainable water resources management and agricultural develop-
Although this specific application focuses on water conservation ment in the Volta River Basin, Ghana. Sci. Total Environ. 573:444–457. https://
technologies' adoption, the general concept and methodology are flexi- doi.org/10.1016/j.scitotenv.2016.08.081.
Koutiva, I., Makropoulos, C., 2015. Exploration of domestic water demand attitudes and
ble enough to support the investigation of a variety of urban water man-
behaviours using an online survey in Athens, Greece. 14th International Conference
agement problems and challenges at the interface between traditional on Environmental Science and Technology (CEST2015), 3-5 September 2015, Rhodes,
engineering and related socio-economic policy tools, arguing for a Greece.
more comprehensive ‘sociotechnical’ discourse of direct relevance to Koutiva, I., Makropoulos, C., 2016. Modelling domestic water demand: an agent based
approach. Environ. Model. Softw. 79:35–54. https://doi.org/10.1016/j.envsoft.2016.01.005.
an ever more customer-focused urban water sector. Limpert, E., Stahel, W.A., Abbt, M., 2001. Log-normal distributions across the sciences:
keys and clues. Bioscience 51 (5):341–352. https://doi.org/10.1641/0006-
Acknowledgements 3568(2001)051[0341:LNDATS]2.0.CO;2.
Makropoulos, C., 2014. Thinking platforms for smarter urban water systems: fusing tech-
nical and socio-economic models and tools. Model Fusion: integrating environmental
This research has been co-financed by the European Union (Europe- models to solve real world problems. Geological Society, London, Special Publications
an Social Fund - ESF) and Greek national funds through the Operational 408. https://doi.org/10.1144/SP408.4 (SP408.4).
Makropoulos, C.K., Butler, D., 2010. Distributed water infrastructure for sustainable com-
Program “Education and Lifelong Learning” of the National Strategic munities. Water Resour. Manag. 24 (11):2795–2816. https://doi.org/10.1007/
Reference Framework (NSRF) (MIS 346725) - Research Funding Pro- s11269-010-9580-5.
gram: Heracleitus II. Investing in knowledge society through the Euro- Makropoulos, C.K., Memon, F.A., Shirley-Smith, C., Butler, D., 2008. Futures: an exploration
of scenarios for sustainable urban water management. Water Policy 10 (4):345–373.
pean Social Fund. https://doi.org/10.2166/wp.2008.014.
Mereu, S., Sušnik, J., Trabucco, A., Daccache, A., Vamvakeridou-Lyroudia, L., Renoldi,
Supplementary data S., Virdis, A., Savić, D., Assimacopoulos, D., 2016. Operational resilience of reser-
voirs to climate change, agricultural demand, and tourism: a case study from
Sardinia. Sci. Total Environ. 543 (Part B):1028–1038 (Special Issue on Climate
Supplementary data to this article can be found online at https://doi. Change, Water and Security in the Mediterranean). https://doi.org/10.1016/
org/10.1016/j.scitotenv.2017.10.041. j.scitotenv.2015.04.066.
1602 S. Baki et al. / Science of the Total Environment 622–623 (2018) 1590–1602

Millock, K., Nauges, C., 2010. Household adoption of water-efficient equipment: the role Rozos, E., Butler, D., Makropoulos, C., 2016. An integrated system dynamic – cellular au-
of socio-economic factors, environmental attitudes and policy. Environ. Resour. tomata model for distributed water-infrastructure planning. Water Sci. Technol. 16
Econ. 46 (4):539–565. https://doi.org/10.1007/s10640-010-9360-y. (6):1519–1527. https://doi.org/10.2166/ws.2016.080.
OECD, 2010. Pricing Water Resources and Water and Sanitation Services. ENV/EPOC/ Safiolea, E., Baki, S., Makropoulos, C., Deliege, J.F., Magermans, P., Everbecq, E., Gkesouli, A.,
GSP(2009)17/FINAL, Organisation for Economic Co-operation and Development, Paris, Stamou, A., Mimikou, M., 2011. Integrated modelling for river basin management
France Available at:. http://www.oecd-ilibrary.org/content/book/9789264083608-en, planning. Proc. Inst. Civil Eng. Water Manag. 164 (8):405–419. https://doi.org/
Accessed date: 19 November 2015. 10.1680/wama.2011.164.8.405.
OECD, 2012. Dataset: Economic Outlook No 91 - June 2012 - Long-term baseline projec- Schwarz, N., Ernst, A., 2009. Agent-based modeling of the diffusion of environmental in-
tions, GDPPOP variable. Available at:. http://stats.oecd.org/Index.aspx?DataSetCode= novations — an empirical approach. Technol. Forecast. Soc. 76 (4):497–511. https://
EO91_LTB, Accessed date: 17 September 2015. doi.org/10.1016/j.techfore.2008.03.024.
OECD, 2014. National Accounts at a Glance 2014, OECD Publishing, Paris, France. Avail- Sofoulis, Z., 2005. Big water, everyday water: a sociotechnical perspective. Continuum 19
able at:. https://doi.org/10.1787/na_glance-2014-en, Accessed date: 3 July 2016. (4):445–463. https://doi.org/10.1080/10304310500322685.
OECD, 2015a. GDP long-term forecast (indicator). https://doi.org/10.1787/d927bc18-en. Sterman, J.D., 2001. System dynamics modeling: tools for learning in a complex world.
Available at:. https://data.oecd.org/gdp/gdp-long-term-forecast.htm#indicator- Calif. Manag. Rev. 43 (4):8–25. https://doi.org/10.2307/41166098.
chart, Accessed date: 17 September 2015. Sušnik, J., Vamvakeridou-Lyroudia, L.S., Savić, D.A., Kapelan, Z., 2012. Integrated System
OECD, 2015b. Better Life Index - Edition 2015, Household net adjusted disposable income. Dynamics Modelling for water scarcity assessment: case study of the Kairouan region.
Available at:. http://stats.oecd.org/Index.aspx?DataSetCode=BLI, Accessed date: 17 Sci. Total Environ. 440:290–306. https://doi.org/10.1016/j.scitotenv.2012.05.085.
September 2015. Tapsuwan, S., Burton, M., Mankad, A., Tucker, D., Greenhill, M., 2014. Adapting to less
OECD, 2015c. Household Dashboard, OECD Stat. Available at:. http://stats.oecd.org/ water: household willingness to pay for decentralised water systems in urban Aus-
Index.aspx?DataSetCode=HH_DASH, Accessed date: 17 September 2015. tralia. Water Resour. Manag. 28 (4):1111–1125. https://doi.org/10.1007/s11269-
Piper, S.L., 2014. The influence of rebates on the purchase of and willingness to pay for 014-0543-0.
water conservation devices. Reclamation: Managing Water in the West. U.S. Depart- The World Bank, 2015. World DataBank, World Development Indicators, Indicator name:
ment of the Interior, Bureau of Reclamation, Technical Service Center, Denver, Colo- GDP per capita growth (annual %) (NY.GDP.PCAP.KD.ZG). Available at:. http://
rado, USA. databank.worldbank.org/data/reports.aspx?source=2&country=ARG&series=&pe-
Prodanovic, P., Simonovic, S.P., 2009. An operational model for support of integrated wa- riod=#, Accessed date: 17 November 2015.
tershed management. Water Resour. Manag. 24 (6):1161–1194. https://doi.org/ Ventana Systems, 2014. Vensim DSS software, Ventana Systems Inc. Retrieved from:
10.1007/s11269-009-9490-6. http://www.vensim.com.
Rogers, E.M., 2003. Diffusion of Innovations. 5th Edition. Simon and Schuster, New York, Vernardakis, C., 2014. [Deliverable 4.1.1: Evaluation report of the household survey for
USA. water use & alternative water sources], Hydropolis project report, Thales Research
Rozos, E., Makropoulos, C., 2013. Source to tap urban water cycle modelling. Environ. Funding Program, Athens, Greece (In Greek).
Model. Softw. 41:139–150. https://doi.org/10.1016/j.envsoft.2012.11.015. Wang, S., Xu, L., Yang, F., Wang, H., 2014. Assessment of water ecological carrying
Rozos, E., Makropoulos, C., 2014. [Deliverable 3.4.2: Pilot application of the decision sup- capacity under the two policies in Tieling City on the basis of the integrated system
port tool], Hydropolis project report, Thales Research Funding Program, Athens, dynamics model. Sci. Total Environ. 472:1070–1081. https://doi.org/10.1016/
Greece (In Greek). j.scitotenv.2013.11.115.
Rozos, E., Makropoulos, C., Butler, D., 2010. Design robustness of local water-recycling Winz, I., Brierley, G., Trowsdale, S., 2009. The use of System Dynamics simulation in water
schemes. J. Water Res. Plann. ASCE 136 (5):531–538. https://doi.org/10.1061/ resources management. Water Resour. Manag. 23 (7):1301–1323. https://doi.org/
(ASCE)WR.1943-5452.0000067. 10.1007/s11269-008-9328-7.
Rozos, E., Makropoulos, C., Maksimović, Č., 2013. Rethinking urban areas: an example of
an integrated blue-green approach. Water Sci. Technol. 13 (6):1534–1542. https://
doi.org/10.2166/ws.2013.140.