An optimization model for the allocation of water resources

Journal of Cleaner Production 164 (2017) 994e1006

Contents lists available at ScienceDirect

Journal of Cleaner Production journal homepage: www.elsevier.com/locate/jclepro

An optimization model for the allocation of water resources Dunia Abdulbaki a, Mahmoud Al-Hindi b, *, Ali Yassine a, Majdi Abou Najm c a

Department of Industrial Engineering & Management, American University of Beirut, Beirut, Lebanon Department of Chemical & Petroleum Engineering, American University of Beirut, Beirut, Lebanon c Department of Civil & Environmental Engineering, American University of Beirut, Beirut, Lebanon b

a r t i c l e i n f o

a b s t r a c t

Article history: Received 22 September 2016 Received in revised form 23 March 2017 Accepted 4 July 2017 Available online 6 July 2017

Critical water shortages, triggered by increasing demands and decreasing supplies, are growing in frequency and spatial extent pausing major challenges for water resources managers around the world. This paper presents an integer linear programming decision support model for the optimal treatment and allocation of water resources. The model seeks to minimize the total water cost which includes the economic cost of treatment and distribution, as well as the associated environmental costs. The model is unique in its ability to account for spatially distributed water supply and demand nodes, as well as multiple water supply (seawater, surface, ground and wastewater) and demand (irrigation, potable, and industrial) types and qualities. It accommodates various treatment technologies, different energy recovery levels, and resource availabilities or capacities. The optimal solution yields volumes of water transported from each supply source to each treatment plant and treated by an appropriate technology in order to satisfy multiple water demands at different required water qualities with the lowest overall economic and environmental costs. The model is applied to a case study. Results showed that the distance of brackish water sources and the environmental cost, observed in terms of carbon savings only, had limited impact on the optimal solution with the demand for the base case being met through a combination of conventional water and wastewater treatment and brackish water reverse osmosis. Sensitivity analysis is performed to determine the effects of variations in demand/supply volumes as well as variable distances and environmental cost. Sensitivity analysis showed that increased demand under limited resources can be met through the introduction of seawater desalination plants, initially through multi effect distillation combined with residual thermal energy then augmented with seawater reverse osmosis plants with further increase in demand. © 2017 Elsevier Ltd. All rights reserved.

Keywords: Water resource management Decision support system Integer linear programming Desalination Water reuse

1. Introduction Scarcity of fresh water resources is becoming increasingly critical in different regions around the world due to growing populations, increasing consumption patterns, rising anthropogenic activities, and climate change (Schwarzenbach et al., 2010; Oelkers et al., 2011; Bagatin et al., 2014). Water scarcity already affects every continent and more than 40% of the world population (WWAP, 2012). By 2025, 1.8 billion people will be living in regions with absolute water scarcity, and two-thirds of the people in the world could be living under water stressed conditions (WWAP, 2012). In order to cope with increasing demands under limited supplies, degrading qualities, and increased treatment options, water

* Corresponding author. E-mail address: [email protected] (M. Al-Hindi). http://dx.doi.org/10.1016/j.jclepro.2017.07.024 0959-6526/© 2017 Elsevier Ltd. All rights reserved.

resource management decisions are becoming increasingly complex (Bagatin et al., 2014). Water, depending on its quality and source, undergoes one or more treatment processes such as, but not limited to, desalination, filtration, and disinfection, to render it suitable for the intended use. Thus, decision makers need to determine the optimal amount of water from each supply source to be transported to each plant and treated by an appropriate technology in order to satisfy the demands, while meeting water quality constraints, at the lowest possible overall economic and environmental cost. The use of decision support systems (DSS) for water resource management has received growing attention in the past few years. Several problem formulations have been devised and operations research tools such as linear programming (LP), multi criteria decision analysis (MCDA), and cost-benefit analysis have been used to solve these formulations (Al-Zaharani et al., 2016; Atilhan et al., 2012; Ghassemi and Danesh, 2013; Molinos-Senante et al., 2015;

D. Abdulbaki et al. / Journal of Cleaner Production 164 (2017) 994e1006

Ruiz-Rosa et al., 2016). Afify (2010) used Multi-criteria Decision Analysis (MCDA) to compare all desalination alternatives for Egypt, considering the use of desalinated water, source of feed water, desalination technology, locations of the plants, and their capacities. Al-Zahrani et al. (2016) used a multi-objective goal programming approach to simulate water distribution from multiple sources to multiple users for the city of Riyadh, Saudi Arabia over a thirty five year period. Atilhan et al. (2012) developed an optimization-based approach for the design of water desalination and distribution networks to satisfy the demands of the various water-consuming sectors. Chung et al. (2008) presented a general water supply planning tool comprised of modular components including water sources, users, recharge facilities, and water and wastewater treatment plants. They concluded that these modules should be linked with optimization routines for more reliable results. Ghassemi and Danesh (2013) developed an integrated two-step model based on the fuzzy-analytic hierarchy process (AHP) and technique of order preference by similarity to ideal solution (TOPSIS) methods for the selection of the optimum desalination technology. Joksimovic et al. (2006) developed a simulation model to be used in combination with an integrated optimization engine for the evaluation and selection of optimal treatment and distribution alternatives in water reuse projects. Molinos-Senante et al. (2012) developed an approach for implementing efficient and effective policies for wastewater treatment by integrating economic and environmental benefits from wastewater using an environmental decision support system (EDSS). More recently, Molinos-Senante et al. (2015) used a different solution approach, namely the analytic network process approach, to solve the problem formulated in their earlier research. Morais and Almeida (2007) dealt with the allocation of resources for water supply in order to choose the city in which a water supply system project will be implemented. They applied the elimination and choice expressing reality (ELECTRE) method, a multi-criteria decision-aid support tool. They compared their model results to decisions based on intuitive judgments and concluded that the use of their method improved the quality of the decision making process. Ruiz-Rosa et al. (2016), on the other hand, used a cost management model to determine the most suitable water source/supply option, amongst the available sources of wastewater reuse, surface water, ground water and desalination. Sadr et al. (2015) developed a fuzzy logic based multi-criteria group decision making tool for the selection of membrane assisted treatment technologies in four different water reuse scenarios. Sudhakaran et al. (2013) created a decision support system (DSS) based on MCDA to compare processes for organic micropollutant removal using the following criteria: treatability, costs, technical considerations, sustainability and time. Their proposed DSS can be used as a screening tool for experimental planning or a feasibility study preceding the main treatment system selection and design. Zoltay et al. (2010) developed a generic decision support system to screen a range of technical, economic and policy management options for watershed management; the model was applied to the Ipswich River basin, USA, and was solved using linear programming to yield optimal water allocation and management solutions. It is clear from the above literature that several DSSs have been developed in order to select optimal water allocation or treatment strategies under given sets of criteria. However, the vast majority of these studies have not included all the water supply, treatment and demand options that are considered in this work, nor have they considered environmental impacts (such as carbon footprint, amongst others). In this paper, a mixed integer non-linear program (MINLP),

995

reduced to a mixed integer linear program, is proposed for selecting the optimum water treatment technologies and water resource allocation. The objective of the linear program is to minimize the overall economic and environmental cost of the water treatment and distribution system, subject to technical, economic and environmental constraints. The model is sufficiently flexible to consider any number of supply and demand nodes, water quality types, economic and environmental cost structures, as well as treatment options, simultaneously. To achieve this formulation, data on economic and environmental costs of different water and wastewater treatment options are collected from a variety of sources, and are correlated into appropriate cost functions. The result is a decision support system (DSS) that can aid decision makers in the optimal selection of water treatment and distribution systems while keeping the economic cost and environmental damage under control. 2. Material and methods Achieving model objective requires an abstraction of a generalized system of interest as a fundamental step prior to establishing the mathematical formulation of the model. This includes the different types of water demands, supplies, and treatments. Furthermore, the development of the objective function and constraints in the DSS requires definition of the underlying criteria and cost structure. 2.1. Water demand, supply and treatment processes Water demand is affected by the size of the population, type of the community, potential and actual use of water (agricultural, industrial, residential, recreational), level of economic development, and local climate conditions (Miller, 2003). In this work, the water demand options are classified into three main categories: (1) domestic, (2) agricultural, and (3) industrial. With respect to supply, a wide range of possible water supply sources are considered in this work, including conventional fresh water sources such as surface water (rivers and lakes) and groundwater (aquifers and wells) as well as non-traditional supply sources such as seawater and wastewater (Leverenz et al., 2011). Processes and technologies employed for water treatment vary depending on the quality of available supply and desired water output which is often dictated by demand and the type of end use (Miller, 2003). Conventional water treatment of low salinity surface water includes processes such as acid addition, coagulant/flocculant addition, filtration and disinfection (Viessman et al., 2008). Desalination, on the other hand is a process that removes dissolved minerals from seawater, high salinity surface/ground water, or treated wastewater (Shatat and Riffat, 2014; Wenten and Khoiruddin, 2016). All desalination processes involve three liquid streams: the feed water, the low-salinity product water, and a very saline concentrate (known as brine or reject water) which requires disposal (Perez-Gonzalezet al., 2012; Wenten and Khoiruddin, 2016). The amount of product water generated from the original feed stream (termed recovery) will depend on the type of water and technology used and varies between 45 and 90% (Saavedra et al., 2013; Wenten and Khoiruddin, 2016). Desalination technologies are often divided into two main categories: thermal and membrane processes (Elimelech and Phillip, 2011). Thermal processes employ distillation, where saline water is heated to produce water vapor, which is then condensed to produce freshwater (Khawaji et al., 2008). The most widespread distillation processes used to produce potable water include Multi-stage Flash (MSF) distillation, Multiple-effect Distillation (MED), and mechanical vapor compression (MVC) (Miller, 2003; Khawaji et al., 2008). In

996

D. Abdulbaki et al. / Journal of Cleaner Production 164 (2017) 994e1006

membrane processes (reverse osmosis (RO) or electrodialysis (ED)), water is extracted from the saline water by passing it through a membrane and applying pressure (RO) or an electric current (ED) for effective separation into the product and brine streams (Miller, 2003; Shatat and Riffat, 2014; Greenlee et al., 2009; Wenten and Khoiruddin, 2016). The treatment of wastewater is carried out in three consecutive stages (Metcalf and Eddy, 2013). Primary treatment removes suspended solids and organic matter by physical operations. Secondary treatment employs biological processes to convert the organic matter into settleable solids with processes including activated sludge (AS), oxidation ditches, and membrane bioreactor (MBR). Tertiary treatment purifies the water by removing nitrogen, phosphorus, metals, biodegradable organics, bacteria, and viruses (Metcalf and Eddy, 2013). 2.2. Criteria used for the derivation of the mathematical model Water resource management decisions have to contend with several technical, financial, environmental and social criteria (Lund, 2015). The selection of the appropriate, but often competing, criteria depends on a large number of factors such as, but not limited to, availability of information, level of stakeholder involvement, geographical location, financial budgets and technical expertise (Lund, 2015; Saloranta et al., 2003). In this work the selection of criteria was based solely on the availability of quantitative information. However, it should be noted that the “optimal” choice of water allocation will also depend on several other criteria, technical and non-technical, such as reliability of plants (Ghassemi and Danesh, 2013), public acceptance of wastewater reuse schemes (Ross et al., 2014) and desalination plants (Dolnicar et al., 2011; Gibson et al., 2015) and will, almost invariably, differ with different stakeholders and stakeholder priorities. 2.2.1. Financial criteria Economic criteria measure the net cost (in $ terms) of the project. These costs include the cost of initial design and construction of the water and wastewater treatment plants, and transportation systems from water/wastewater supply source to the plant, the cost of operation and maintenance, in addition to the cost of transporting the produced water to the demand location (Al-Karaghouli and Kazmerski, 2013; Plappally and Lienhard, 2012). The cost structure adopted in this work is divided as follows: i. Capital cost: equipment, installation, construction ii. Operating cost: energy requirements, service and maintenance, supplies and parts, labor, pre-treatment of feed water and post-treatment of output water iii. Pumping and transportation costs: source nodes to treatment plants, and treatment plants to demand nodes.

2.2.2. Environmental impact criterion During the construction and operation of the proposed system, there are several factors contributing to environmental damage, most notably the greenhouse gas emissions and the waste produced. The criteria employed in this work is based on the life cycle assessment (LCA) and measures the environmental cost of the process through its carbon footprint which is an estimate of the amount of greenhouse gases (GHGs) released into the environment as a result of the utilization of a particular treatment process over some period of time (Chen et al., 2012; Zhou et al., 2014; Shahabi et al., 2014). The LCA methodology has been used over the past decades to assess the environmental impacts resulting from water transportation (Amores et al., 2013), treatment (Barrios et al.,

2008), desalination (Raluy et al., 2006) and wastewater treatment (Mahgoub et al., 2010) processes in various locations. While other factors can cause relevant environmental or health costs, only carbon footprint was considered in our example due to the lack of comprehensive data on other factors, but such numbers can be incorporated in the model as this field develops. 2.2.3. Water quality criteria A variety of factors come into play when selecting the appropriate technology for a particular water supply/demand situation. On the supply side, the available feed water quality is a function of degree of salinity, TDS, pH, hardness, toxicity, microbial content, and chemical content (Viessman et al., 2008). On the demand side, required water quality is determined based on end use requirement (Leverenz et al., 2011). The guidelines and recommendations for drinking water, irrigation water, and water for industrial use are summarized in Table 1 (WHO, 2008). 2.3. Process treatment costs: data and correlations The capital and operating costs are primary parameters used by decision makers to select the appropriate water treatment technology for a project. Cost data are site-specific, differ with technologies, and typically are not consistent even for similar-sized facilities. In this paper, costs from several references were synthesized in order to come up with cost functions (Al-Karaghouli and Kazmerski, 2013; Plappally and Lienhard, 2012; Ruiz-Rosa et al., 2016; Younos, 2005). Capital cost, often referred to as CAPEX or Capital Expenditure, includes all expenditures associated with the implementation of a given water treatment project from the time of its conception, through design, permitting, financing, construction, and commissioning. Land cost depends on the contract agreement and may vary from zero to an agreed sum depending on the site characteristics. Operating cost, referred to as OPEX or Operating Expenditure, is also site-specific and consists of fixed costs and variable costs. Fixed costs include insurance and amortization costs. The primary variable operating costs include the cost of labor, energy, consumables such as chemicals, parts (such as pumps and membranes) replacement, and maintenance. 2.3.1. Pumping and transportation costs The cost of transporting surface water is mainly dependent on distance, elevation difference, and soil conditions, as well as labor, electricity, and spare parts costs. The variation of pumping uphill and flow due to gravity are other aspects that can influence the variation in the cost of water. It is therefore difficult to compare pipeline construction costs from one location to another (Plappally and Lienhard, 2012). Acknowledging the above constraints, data from the Lamei et al. (2008) study (from the MENA region) is used in our case study with the recommendation that such relationships need to be complemented with local data when the model is used in other regions. Lamei et al. (2008) performed a multiple linear regression analysis to estimate unit production costs for longdistance water supply. Data from different long-distance water piping projects was used to correlate unit capital and unit production costs with distance of transfer, L, in km, and with volumetric flow rate (capacity), C, in m3/day through the pipelines. The cost relationships for the transportation costs are summarized in Table 2. 2.3.2. Desalination costs and correlations Desalination capacity has rapidly increased in the last decade triggered by increased water demands and incentivized by sharp reduction in desalination cost as a result of significant technological

D. Abdulbaki et al. / Journal of Cleaner Production 164 (2017) 994e1006

997

Table 1 Guidelines for water quality parameters. Water Use

Domestic Agricultural Industrial

Guideline Values Total Dissolved Solids, TDS (mg/L)

pH

Hardness (mg/L)

Sodium, Na (mg/L)


500
1000
700

6.5e8.5 4.5e9.0 7.0e9.0

50e100 NA NA

NA
70 NA

Table 2 Cost functions used in the mathematical model. Process Transportation MSF MED SWRO BWRO RF WWT-Conv WWT-MBR

Investment cost functions 49 þ 18.5L - 0.04C 8995.2C0.162 6756.5C0.222 7,2029C0.294 13,584C0.17 250.56C 17,599C0.359 24,313C0.31

R2 N/A 0.7994 0.6479 1 1 N/A 0.9986 0.9336

Operating cost functions 0.04 þ 0.01Le10 3.4241C0.193 2.4051C0.255 17.214C0.297 3.1725C0.17 0.23025C 4.2652C0.307 0.375C0.103

5

C

R2 N/A 0.9997 0.9993 1 0.9996 N/A 0.9543 0.8541

C is the plant production capacity or volume of water pumped in m3/day and L is the pipe length in km. MSF: multi-stage flash; MED: multi-effect distillation; SWRO: seawater reverse osmosis; BWRO: brackish water reverse osmosis; RF: river filtration; WWT-Conv: conventional wastewater treatment; WWT-MBR: membrane bio-reactor wastewater treatment.

advances, growth rate, plant capacity, competition with other technologies, and the vast improvements in RO systems (Elimelech and Phillip, 2011; Ghaffour et al., 2013). Thermal methods are more expensive because of the large quantities of fuel required to vaporize salt water. In general, thermal energy contributes to half the cost of the thermal desalination process (Karagiannis and Soldatos, 2008). However when residual thermal energy is used or when renewable energy sources are utilized to drive these processes, their cost is greatly reduced and competes with that of RO (Al-Karaghouli and Kazmerski, 2013). Fig. 1 plots the unit capital investment cost (Fig. 1(a)) and operating cost (Fig. 1(b)) in $/m3 for the most commonly used thermal and membrane desalination processes as a function of the plant capacity, C, in m3/day. The plots are based on data reported by Wittholz et al. (2008) and Loutatidou et al. (2014). The shape of the curves indicates that the system cost follows the pattern of economies of scale; i.e. the system cost decreases with the increase in plant capacity. The cost functions and their corresponding R2 values obtained from regressions following power functions are shown in Table 2. The high R2 values validate the choice of representing the cost functions in the form of ACn where A and n are constants.

2.3.3. Wastewater costs and correlations Factors that influence the investment cost of wastewater treatment plants (WWTP) include plant capacity, design criteria, treatment process, land cost, location of construction and weather conditions, as well as competition among bidders and suppliers and stability of the local and national economic conditions (Molinos-Senante et al., 2015). Fig. 2 shows the unit capital investment cost (Fig. 2(a)) and operating cost (Fig. 2(b)) in $/m3 for the extended aeration activated sludge wastewater treatment technology WWTConventional and for the membrane bioreactor wastewater treatment technology WWT-MBR as a function of the plant capacity, C, in m3/day. The data for WWT-MBR cost curve was obtained from ^ te  et al. (2004), Brepols et al. (2009) and DeCarolis et al. (2007). Co The data used in forming the WWT-Conventional cost curve and function was obtained from Gonzalez-Serrano et al. (2005). The shape of the curves indicates that the system cost follows the pattern of economies of scale. Table 2 includes the corresponding

cost functions and the R2 values. The high R2 values validate the choice of representing the cost functions in the form of ACn. 2.3.4. Conventional water treatment costs Conventional water treatment systems (of river or lake water) usually consist of several treatment steps and will often depend on the quality of the surface water treated. In general most conventional systems consist of physical-chemical processes such as flocculation/coagulation, sedimentation/settling, filtration (sand/ multi-media/membrane) and disinfection (Viessman et al., 2008). In this work it is assumed that the conventional water treatment system consists of a lamella settling system, depth filtration, ultrafiltration, and disinfection and the costs functions used by Iglesias et al. (2010) are employed and are summarized in Table 2. 2.3.5. Environmental impact costs In this work, the carbon footprint associated with each treatment system referred to as the global warming potential (GWP) and expressed in kilograms carbon dioxide equivalents (kg CO2-e) is considered as an “environmental cost”. A summary of the environmental costs collected from the reviewed literature is presented in Table 3 which shows that the thermal desalination technologies coupled to conventional energy sources have the highest environmental impact. These observations have been further corroborated by a number of recent investigations (Uche et al., 2015; Barau and Al Hosani, 2015). The monetary cost of green-house gases (GHGs) emissions has been the subject of ongoing debate with most workers opting for the marginal abatement cost (Morris et al., 2008; Yohe et al., 2007). Costs vary from $10/ton -$350/ton with more recent estimates placing the figure in the range of $200/ton e $2000/ton (Moore and Diaz, 2015). In this work, and for the base case run, a cost of $0.2/kg CO2-e will be used. 3. Theory and mathematical model Consider a geographic area that has known, deterministic, and static water demands for specific uses (demand for water quality type m) at different specific locations (demand node n). Those demands are satisfied by various water supply sources (source node i)

998

D. Abdulbaki et al. / Journal of Cleaner Production 164 (2017) 994e1006

Fig. 1. Unit cost of desalination vesrus plant capacity for different desalination processes (Seawater reverse osmosis (SWRO), brackish water reverse osmosis (BWRO), multi-stage flash (MSF) and multi-effect distillation (MED)): (a) capital cost and (b) operating cost.

of different water types (source type j) within the same area considered. Furthermore, to improve source water quality j to satisfy a demand quality m, the model can allocate water to multiple treatment plants, (plant k), each employing different treatment processes (treatment technology l). The goal is to optimize the water allocation from different water sources, i, and their corresponding qualities, j, to different treatment plants, k, utilizing different technologies, l, to satisfy the local demands, n, for corresponding water quality type, m. The primary objective is to minimize the overall economic and environmental costs. A schematic representation of the problem is shown in Fig. 3. 3.1. Mathematical model The general form of the objective function considers the

Fig. 2. Unit cost of wastewater treatment (WWT) vesrus plant capacity for conventional (WWT-Conv) and membrane bioreactor (WWT-MBR) technologies: (a) capital cost and (b) operating cost.

Table 3 Environmental impacts of various desalination and wastewater treatment technologies. Process

GWP (kg CO2-e/m3)

Reference

Transportation MSF

0.635 23.41 conventional energy 1.98 residual energy 18.05 conventional energy 1.11 residual energy 1.81 0.624 0.829 1.35

Amores et al. (2013) Raluy et al. (2006)

MED SWRO BWRO WWT-Conv WWT-MBR

Raluy et al. (2006) Tarnacki et al. (2012) Tarnacki et al. (2012) Pasqualino et al. (2011) Hospido et al. (2012)

MSF: multi-stage flash; MED: multi-effect distillation; SWRO: seawater reverse osmosis; BWRO: brackish water reverse osmosis; RF: river filtration; WWT-Conv: conventional wastewater treatment; WWT-MBR: membrane bio-reactor wastewater treatment.

D. Abdulbaki et al. / Journal of Cleaner Production 164 (2017) 994e1006

999

Fig. 3. Schematic representation of the problem.

minimization of the overall cost of the system as stated below. The notation for decision variables and parameters used below are presented in the nomenclature section.

Minimize Z ¼

J X I X K X L X M X N n X

E ¼ CFT;ijk þ ð1  dT Þð1  dl ÞCFT;kmn þ ð1  dT Þ2 ð1  dl ÞCFl Vijklmn f1 ðB; O; EÞ

i¼1 j¼1 k¼1 l¼1 m¼1 n¼1

þ aijklmn f2 ðB; O; EÞ

for losses during treatment (1-dl) and transportation (1-dT); and (3) treatment using technology l (CFl), and is given by:

o

(1)

Where Z is the total cost. B, O, and E are the investment, operating, and environmental costs, respectively. f1(B,O,E) is a function that includes the variable terms of the investment, operating, and environmental cost; while f2(B,O,E) is a function that includes the fixed terms of the investment, operating, and environmental costs. Finally, aijklmn is a binary decision variable. B is the capital investment cost associated with transportation and treatment. It is the sum of investment costs in transportation (from source i to plant k, and from plant k to demand n) and treatment and is given by: c c B ¼ Tijk þ Cjklm þ Tkmn

(4)

The transportation cost is mainly a function of the flow rate and the distance travelled, but is also impacted by elevation and type of materials used in the distribution system. Similarly, process costs depend primarily on the flow rate, but can also be a function of energy consumption and type of energy available, process configuration, feed and product water quality, water temperature, and reject disposal methods. Exact expressions for B, O, and E used in this work were presented in Section 2 and are summarized in the case study (Section 3.3). There are several constraints used in this model. Constraint (5) is specific to the part of the objective function relating to the fixed cost. It uses a binary variable that indicates when the fixed cost is incurred, and a sufficiently large number M to create redundancy (M ¼ 1,000,000 will suffice in our case). So aijklmn is 1 only when Vijklmn > 0, and 0 otherwise.

(2)

On the other hand, O is the operating cost. It is the sum of operating costs which includes: (1) transporting water of type j o ); (2) transporting water of type m from source node i to plant k (Tijk o ) after accounting for losses from plant k to demand node n (Tkmn during treatment (1-dl) and transportation (1-dT); (3) treatment of water of type j in plant k using technology l to satisfy demand for quality type m (Ojklm ); and (4) unwanted material disposal in technology l (Ll), and is given by:

o o O ¼ Tijk þ ð1  dT Þð1  dl ÞTkmn þ ð1  dT ÞOjklm þ ð1  dT Þdl Ll

(3) Finally, E is the environmental cost. It is the sum of the environmental costs of carbon footprint from: (1) transporting water of type j from source node i to plant k (CFT;ijk ); (2) transporting water of type m from plant k to demand node n (CFT;kmn ) after accounting

Vijklmn
Maijklmn K X L X M X N X

ci; j; k; l; m; n

Vijklmn
Sij

(5)

ci; j

(6)

k¼1 l¼1 m¼1 n¼1

ð1  dT Þ2

J X I X K X L X

ð1  dl ÞVijklmn ¼ Dmn

cm; n

(7)

i¼1 j¼1 k¼1 l¼1

The basic constraints in the model are those related to satisfying supply and demand. On one hand, constraint (6) ensures the volume of water taken must be less than or equal to the capacity of supply available from the sources considered (Sij). On the other hand, constraint (7) balances the demands by equating the volume of water taken (less the losses due to transportation and technology) to the demand required (Dmn).

1000

J X I X K X L X N X

D. Abdulbaki et al. / Journal of Cleaner Production 164 (2017) 994e1006

ðTDSl  TDSmH Þð1  dl ÞVijklmn
0

function and p is an index for interval number. The additional constraints are:

cm

p

i¼1 j¼1 k¼1 l¼1 n¼1

(8) J X I X K X L X N X

ðpHl  pHmH Þð1  dl ÞVijklmn
0

cm

(9)

aijklmn ¼ aijklmn P X

Vijklmn ¼

ci; j; k; l; m; n

p Vijklmn

(15)

ci; j; k; l; m; n

(16)

p¼1

i¼1 j¼1 k¼1 l¼1 n¼1 J X I X K X L X N X

pþ1

ðpHl  pHmL Þð1  dl ÞVijklmn  0

cm

(10)

i¼1 j¼1 k¼1 l¼1 n¼1 J X I X K X L X N X

0
wpijklmn
ðhdl  hdmH Þð1  dl ÞVijklmn
0

cm

p



VHp

ci; j; k; l; m; n p ¼ 1; …; P  1

(17)

ci; j; k; l; m; n p ¼ 1; …; P  1

p

(18)

p

p

p

aijklmn ; aijklmn ; wijklmn 2f0; 1g; 0
Vijklmn
VH ; Vijklmn  0 ðhdl  hdmL Þð1  dl ÞVijklmn  0

cm

(12)

i¼1 j¼1 k¼1 l¼1 n¼1 J X I X K X L X N X



p

wijklmn

p Vijklmn

(11)

i¼1 j¼1 k¼1 l¼1 n¼1 J X I X K X L X N X

pþ1

0
Vijklmn
VH

ðNal  NamH Þð1  dl ÞVijklmn
0

cm

(13)

i¼1 j¼1 k¼1 l¼1 n¼1

Constraints (8)e(13) ensure that the water quality guidelines specific to each kind of use are applied. This is done by making sure that the values of TDS, pH, hardness, and Na present in the water treated to satisfy specific demand type m is within the limits (upper and lower) specified by the guidelines for that particular use. J X I X K X L X M X N X

ð1  dl ÞCFl Vijklmn
CFmax

(14)

i¼1 j¼1 k¼1 l¼1 m¼1 n¼1

Constraint (14) is added to ensure that even while the carbon footprint is minimized in the objective function; it should also remain less than a certain specified ceiling value in order to limit the environmental damage of the system.

3.2. Solution approach It should be noted that the mathematical model presented above is a general model that can be applied to any situation. However, it is a mixed integer nonlinear program (MINLP). Such problems are intrinsically more difficult to solve than linear programming (LP) problems since MINLP may have multiple feasible regions and multiple locally optimal points within these regions. As such, there is no simple way to determine with certainty that the problem is infeasible, that the objective function is unbounded, or that an optimal solution is the global optimum. Specialized nonlinear solvers (such as Knitro and BMIBNB) can be used, but it would still be extremely difficult to guarantee a global optimal solution for large MINLP problems. Most of these solvers implement a branch-and-bound approach and use linear programming relaxations and convex envelope approximations to ensure global optimality (Belotti et al., 2013). Accordingly, the solution approach employed in this work is based on transforming these non-linear functions into step-wise linear cost functions across p intervals (Bradley et al., 1977; Tsai and Lin, 2008). This is done by introducing a number of additional decision variables and constraints. The additional decision p variables are apijklmn , Vijklmn and wpijklmn , where P is the number of piecewise linear intervals used for approximating the nonlinear

Constraint (15) is used to insure that the binary decision variable aijklmn is bound to equal its linearized counterpart in the linear approximation. Constraint (16) guarantees that the sum of the linearized P intervals adds up to the defined range of the original nonlinear function. Constraints (17) and (18) are set to linearize the non-linear cost functions. They ensure that the volume variables satisfy the lower ranges before moving to the higher ranges. If p pþ1 Vijklmn < VHp then Vijklmn should be 0. This condition is ensured by p

p

p

these two constraints since if Vijklmn < VH then wijklmn ¼ 0 and pþ1

consequently Vijklmn ¼ 0. Despite having avoided the difficulty of solving a nonlinear program, this approach adds some complexity in terms of the increase in the number of decision variables and the associated constraints. In particular, the number of decision variables for the nonlinear problem is the product 2ijklmn. For the linearized problem, the number of decision variables increases to ijklmnð3p  1Þ. Increasing p provides a more accurate approximation to the nonlinear function; however, it will also add complexity to the linearized problem. Another compromise made during linearization is the margin of error that occurs as a result of the linearization process. It is assumed that by increasing the number of intervals considered, the solution to the linearized problem will be closer to the solution of the original nonlinear problem. 3.3. Case study A small sized coastal city located 5 km away from the sea with a total population of 250,000 inhabitants is considered. The inhabitants of this city require 0.2 m3/day per capita of potable water, 0.15 m3/day per capita of irrigation water, and 0.1 m3/day per capita of industrial water. This population generates 0.175 m3/day per capita wastewater. At a distance of 20 km away from the city, there is a brackish water well that can supply up to 25,000 m3/day. A river running 30 km away from the city can provide up to 70,000 m3/day. It is assumed that there are two treatment plants available, each located at a 5 km distance from the city. One of these plants is run by conventional energy sources, while the other plant is located in the proximity of power generating facilities and will tap into the residual thermal energy provided from this power plant. According to the information presented above, the following four sources are considered: the city (i ¼ 1) which provides wastewater (j ¼ 4), the sea (i ¼ 2) providing seawater (j ¼ 1), the brackish water well (i ¼ 3) providing brackish water (j ¼ 2), and the river (i ¼ 4) providing the river water (j ¼ 3). Accordingly, the supply available is S14 ¼ 43,750 m3/day, S21 ¼ 150,000 m3/day,

D. Abdulbaki et al. / Journal of Cleaner Production 164 (2017) 994e1006

S32 ¼ 25,000 m3/day, and S43 ¼ 70,000 m3/day. These sources will be used to satisfy the city's demand (n ¼ 1) of potable water (m ¼ 1), irrigation water (m ¼ 2), and industrial water (m ¼ 3). The demand required is D11 ¼ 50,000 m3/day, D21 ¼ 37,500 m3/day, and D31 ¼ 25,000 m3/day. Two treatment plants will be considered: PA (k ¼ 1) is run by conventional energy, while PB (k ¼ 2) is run by residual energy. Each plant can support seven treatment processes: MSF desalination (l ¼ 1), MED desalination (l ¼ 2), SWRO desalination (l ¼ 3), BWRO desalination (l ¼ 4), River Filtration/conventional treatment (l ¼ 5), Conventional WWT (l ¼ 6), and WWT with MBR (l ¼ 7). The linear and non-linear cost functions for the investment and operating costs to be used in this case study were discussed in section 2.3. The cost functions are linearized across nine intervals. The first volume interval (p ¼ 1) covers the range 1e5000 m3/day, the second interval (p ¼ 2) covers the range 5000e10,000 m3/day, the third interval (p ¼ 3) covers the range 10,000e15,000 m3/day, the fourth interval (p ¼ 4) covers the range 15,000e20,000 m3/day, the fifth interval (p ¼ 5) covers the range 20,000e30,000 m3/day, the sixth interval (p ¼ 6) covers the range 30,000e40,000 m3/day, the seventh interval (p ¼ 7) covers the range 40,000e50,000 m3/ day, the eighth interval (p ¼ 8) covers the range 50,000e100,000 m3/day, and finally the ninth interval (p ¼ 9) covers the range 100,000e500,000 m3/day. The linearization method used is step-wise linearization, where the cost function across each interval takes a constant value. The losses/recoveries assumed in this case study are 6% loss in transportation, 50% recovery in SWRO desalination, 75% recovery in BWRO desalination, 20% recovery in MSF and MED desalination, and no losses for WWT processes and river filtration/conventional treatment. Sensitivity analysis was performed to determine the effects of changing a number of parameters on the optimal solution. The effects of changing the volume of water (in terms of supply available and demand required) and the effect of changing the location of the brackish water well (increasing the distance between the well and plants) were performed. Data for the base case along with the parameters varied for the sensitivity analysis are summarized in Table 4. 4. Results and discussion For the case study data, cost correlations, constraints and other relevant information detailed above, the optimization model was solved using Excel premium solver. The problem formulated contained a total of 8736 decision variables and 8408 constraints excluding the variable lower and upper bound restrictions. The total cost of the system was found to be equal to $60,737,757. The optimal mapping of sources to the corresponding treatment

1001

processes to meet the required demand was as follows: - Treat 56,587 m3/day of river water by filtration at plant PB to satisfy the potable water demand. - Treat 42,440 m3/day of wastewater by conventional WWT at plant PB to satisfy irrigation water demand. - Treat 19,840 m3/day of well water by BWRO at plant PB, and 13,413 m3/day of river water by filtration at plant PB to satisfy industrial water demand. Naturally, the model first selects river water filtration treatment since it is the least expensive of the available processes and has minimum loss, provided there is enough river water to be treated. Usually there is a limited amount of river water available as compared to the unlimited amount of seawater that may be used for desalination treatment. The model also used the available wastewater treated by conventional WWT to satisfy agricultural and irrigation demands. Concerning the WWT processes, the model finds the conventional WWT optimal compared to WWT with MBR. The difference between the two processes is that WWT with MBR is more expensive than conventional WWT in the investment phase, but less expensive in the operating phase. In the case of a much higher flow rate, the optimal solution may vary to include WWT with MBR in order to satisfy the required demand. Once the available amount of river water and wastewater is used up, the model then refers to brackish water since its treatment is less expensive compared to seawater desalination to satisfy the remaining demand. It selects a combination of conventional WWT, RF, and BWRO processes in such a way as to satisfy the model constraints while minimizing the cost. It is important to note that the above solution is one of two optimal solutions since there is no difference, in terms of distance, between plants PA and PB since both plants are at the same distance from the source nodes and demand nodes, and both plants offers the same treatment processes without recourse to the expensive seawater desalination technologies. In other words, given that plants PA and PB are both located at the same distance from the proposed city, the environmental and distance constraints did not have a major impact on the solution and in view of the same power costs, results for PA and PB were almost identical and interchangeable as the solution will be optimal choosing either of them. As mentioned in section 3.3, sensitivity analysis was performed to determine the effects of changing the volume of water (in terms of supply available and demand required) and the effect of changing the location of the brackish water well (increasing the distance between the well and plants). The Parametric values for the base case and all of the sensitivity analysis scenarios are detailed in Table 5.

Table 4 Base case scenario and ranges of parameter varied during sensitivity analysis. Parameter

Baseline value

Range of variation

1. volume of water (in terms of supply available and demand required) Note: Since the volume of water flowing through the system depends on the population being served, we studied the effect of the demand volume on the total system cost and optimal combination of treatment processes by increasing the population size. 2. distance between the brackish water well and the treatment plants

population of 250,000 inhabitants Demand: 0.2 m3/day per capita of potable water (50,000 m3/day), 0.15 m3/day per capita irrigation water (37,500 m3/day), and 0.1 m3/day per capita industrial water (25,000 m3/day) Supply: 0.175 m3/day per capita wastewater (43,750 m3/day) 25,000 m3/day brackish well water 70,000 m3/day river water

Variation between 250,000 and 2,000,000 inhabitants by steps of 250,000 The demand and wastewater supply will vary accordingly.

brackish water well located at distance of 20 km from the city

Variation between 20 and 160 km distance in steps of 20.

1002

D. Abdulbaki et al. / Journal of Cleaner Production 164 (2017) 994e1006

Table 5 Parametric values for the base case and sensitivity analysis scenarios. Variation 1

Variation 2

Population

Potable Demand

Irrigation Demand

Industrial Demand

Seawater Supply

Brackish Supply

River Supply

Wastewater Supply

250,000

50,000

37,500

25,000

1,000,000

25,000

70,000

43,750

Baseline Parameters

20

500,000 750,000 1,000,000 1,250,000 1,500,000 1,750,000 2,000,000

100,000 150,000 200,000 250,000 300,000 350,000 400,000

75,000 112,500 150,000 187,500 225,000 262,500 300,000

50,000 75,000 100,000 125,000 150,000 175,000 200,000

1,000,000 1,000,000 1,000,000 1,000,000 1,000,000 1,000,000 1,500,000

50,000 50,000 50,000 50,000 50,000 50,000 50,000

120,000 120,000 120,000 120,000 120,000 120,000 120,000

87,500 131,250 175,000 218,750 262,500 306,250 350,000

Variation

40 60 80 100 120 140 160

For the range of populations considered, the results indicated that despite increasing the distance of the brackish water well from the location of the plants, it remains optimal to meet part of the water demand from the treated brackish water because of its relatively lower cost. This result also shows that the optimal solution, irrespective of the demand volume, is not affected by the location of the brackish water well. The increase of the marginal abatement cost from $0.2/kgCO2-e up to $2.0/kgCO2-e had an almost insignificant effect on the optimal solution for the base case as the magnitude of this marginal cost is relatively small compared to the economic (capital and operating) costs (for the base population of 250,00 inhabitants the “economic” cost was equal to $60 Million while the “environmental” cost, at a marginal abatement cost of $2.0/kgCO2-e, was $0.5 Million; i.e. less than 1% of the total cost). As the field of environmental valuation develops, additional environmental costs can be added. Since the volume of water flowing through the system depends on the population being served, the effect of increasing the demand volume on the total system cost and optimal combination of treatment processes is investigated by increasing the population size; the optimization was performed at marginal abatement cost of $2.0/kgCO2-e. Fig. 4 shows the variation of the economic cost and the environmental cost as a function of the population size.

Well Distance

Naturally both costs increase as the population size increases. The increase in environmental cost is almost linear, while the economic cost curve has changes in slope. This may be due to a change in the selection of processes in the optimal solution causing an increase in the system cost. It is clear that the magnitude of the environmental cost is substantially smaller than the economic costs. Fig. 5 shows that the optimal selection of process technologies, and their utilization in more than one process plant, changes as the demand volume increases for potable use, agricultural use, and industrial use. There are several observations that can be made from these figures. The first is that the dominant treatment process is river water filtration. The most obvious reason for this dominance is the low cost of this technology compared to the rest. In addition to river water filtration, brackish water desalination is always part of the optimal solution because of its moderate cost. Other observations that can be made are concerning the utilization of seawater desalination and waste water treatment technologies. When the demand for water can no longer be met by river water, well water, or treated wastewater, the difference is made up by desalination of seawater. The processes used, initially, are SWRO and MED driven by residual energy. The model selects a combination of SWRO and MED desalination processes in such a way so as to satisfy the model constraints while minimizing the total cost.

Fig. 4. The variation of economic and environmental cost for various population sizes.

D. Abdulbaki et al. / Journal of Cleaner Production 164 (2017) 994e1006

1003

Fig. 5. Optimal allocation of treatment processes and plants for potable, agricultural and industrial water for various population sizes: (a) legend, (b) 250,000 inhabitants, (c) 500,000 inhabitants, (d) 1,000,000 inhabitants, (e) 1,750,000 inhabitants and (f) 2,000,000 inhabitants.

MED driven by residual energy is less expensive than SWRO and has a lower carbon footprint, but SWRO has a higher recovery rate (SWRO recovery is 50% while MED recovery rate is only 20%). This selection is supported by the proximity of the sea and the unlimited amount of seawater available. However, and for moderately large

populations, the use of SWRO with conventional energy sources becomes inevitable. With regard to WWT, conventional WWT was always part of the optimal solution while WWT with MBR was not part of it even at large demand capacity. This may be due to the large investment

1004

D. Abdulbaki et al. / Journal of Cleaner Production 164 (2017) 994e1006

cost of WWT with MBR having a higher effect on the optimal solution than its low operating cost when compared to conventional WWT.

Cjklm

pHl pHmH pHmL Sij c Tijk

Investment capital cost of plant k which uses technology l to treat water of type j to satisfy demand m; this cost includes cost of equipment, installation and construction Carbon footprint cost for technology l Carbon footprint cost for transporting water from source node i of type j to plant k Carbon footprint cost for transporting of type m from plant k to demand node n Carbon footprint upper limit for whole system Fraction of water lost during treatment Fraction of water lost during transportation Demand of water of type m required at node n Environmental costs Estimated hardness of water treated by technology l Upper guideline hardness value of water for demand type m (domestic use) Lower guideline hardness value of water for demand type m (domestic use) Cost of transporting water, brine or other material removed by technology l and disposing it off in the appropriate manner A large integer number Estimated Sodium value of water treated by technology l Guideline Sodium value of water for demand type m (agricultural use) Total operating costs Operating cost of plant k which uses technology l to treat water of type j to satisfy demand m; this cost includes cost for energy requirements, service and maintenance, supplies and parts, and labor Estimated pH of water treated by technology l Upper guideline pH value of water for demand type m Lower guideline pH value of water for demand type m Supply of water of type j available at node i Investment cost of transporting water from source node i

o Tijk

of type j to plant k Operating cost of transporting water from source node i of

CFl CFT;ijk 5. Conclusions In this paper, an integer programming optimization model was developed as a DSS to evaluate and select the optimum combination of water source and treatment process to meet different kinds of demand. The evaluation was based on economic, technical, and environmental criteria. The mathematical model developed is a general model, and can be used for different cost functions and constraints. In this paper, the cost data collected produced nonlinear cost functions, so a stepwise linearization method across several intervals was used for overcoming the difficulty of solving a nonlinear problem. The optimization model, re-formulated as an integer linear program, was applied to a case study, the results were analyzed, and sensitivity analysis was conducted to determine the effect of variations in demand/supply volumes, distance and environmental abatement cost on the optimal solution. For the base case, the results show that filtration, BWRO, and conventional WWT are the first choices selected as an optimal solution. When the demand increases, more water is made available through desalination of sea water. The results also show that there is no dependence between the demand volume and the location of the water well. On the other hand, considering environmental cost in terms of global warming potential only revealed very limited contribution to environmental cost in the decision process as greenhouse gases contributed an almost insignificant amount to the total cost. Such finding highlights a gap in environmental valuation and the need for more research in the area of LCA and environmental valuation. The model, however, can accommodate any additional environmental costs. This optimization-based DSS can be used for water resources planning in general. However, every region has a specific set of constraints, which will be, in all likelihood, different from those considered in this work. Therefore, the model proposed will need to be modified by incorporating some additional/alternative casespecific constraints. These may be related to the type of terrain, the area available for such a project, the type of funding, among others. The model presented in this paper is the first of its kind to simultaneously consider location choices, multiple options of water supply (seawater, surface and wastewater), various treatment technologies (desalination, water reuse, conventional); in addition to accounting for economic and environmental costs. Furthermore, the use of this model, instead of intuitive judgments, could assist in improving the quality of the decision by making it more explicit, rational and efficient.

Nomenclature aijklmn

apijklmn

Binary decision variable that is 1 when water from source node i of type j is transported to plant k and treated with technology l to satisfy demand type m at demand node n (Vijklmn > 0); and 0 otherwise.

CFmax dl dT Dmn E hdl hdmH hdmL Ll

M Nal NamH O Ojklm

p Vijklmn

type j to plant k Investment cost of transporting water of type m from plant k to demand node n Operating cost of transporting water of type m from plant k to demand node n Estimated TDS value of water treated by technology l Guideline TDS value of water for demand type m Volume of water taken from source node i of type j and transported to plant k and treated with technology l to satisfy demand type m at demand node n. A decision variable Volume of water taken from source node i of type j and

wpijklmn

transported to plant k and treated with technology l to satisfy demand type m at demand node n for volume range p. p Binary variable that is 1 when Vijklmn is equal to a certain

c Tkmn o Tkmn

TDSl TDSmH Vijklmn

Z

maximum value; and 0 otherwise. Total cost: sum of investment, operating and environmental costs

Binary decision variable that is 1 when water from source node i of type j is transported to plant k and treated with technology l to satisfy demand type m at demand node n p > 0); and 0 otherwise. for volume range p (Vijklmn

B

CFT;kmn

Total investment cost

Appendix A. Supplementary data Supplementary data related to this article can be found at http:// dx.doi.org/10.1016/j.jclepro.2017.07.024.

D. Abdulbaki et al. / Journal of Cleaner Production 164 (2017) 994e1006

References Afify, A., 2010. Prioritizing desalination strategies using multi-criteria decision analysis. Desalination 250 (3), 928e935. Al-Karaghouli, A., Kazmerski, L., 2013. Energy consumption and water production cost of conventional and renewable-energy-powered desalination processes. Renew. Sustain. Energy Rev. 24, 343e356. Al-Zahrani, A., Musa, A., Chowdhury, S., 2016. Multi-objective optimization model for water resource management: a case study for Riyadh, Saudi Arabia. Environ. Dev. Sustain 18, 777e798.  n, A., Castells, F., 2013. Environmental Amores, M.J., Meneses, M., Pasqualino, J., Anto assessment of urban water cycle on Mediterranean conditions by LCA approach. J. Clean. Prod. 43, 84e92. Atilhan, S., Mahfouz, A., Batchelor, B., Linke, P., Abdel-Wahab, A., N apoles-Rivera, F., nez-Gutie rrez, A., El-Halwagi, M., 2012. A systems-integration approach to Jime the optimization of macroscopic water desalination and distribution networks: a general framework applied to Qatar's water resources. Clean Technol. Environ. Policy 14 (2), 161e171. Bagatin, R., Klemes, J.J., Reverberi, A.P., Huising, D., 2014. Conservation and improvements in water resource management: a global challenge. J. Clean. Prod. 77, 1e9. Barau, A.S., Al Hosani, N., 2015. Prospects of environmental governance in addressing sustainability challenges of seawater desalination industry in the Arabian Gulf. Env. Sci. Policy 50, 145e154. Barrios, R., Siebel, M., van der Helm, A., Bosklopper, K., Gijzen, H., 2008. Environmental and financial life cycle impact assessment of drinking water production at Waternet. J. Clean. Prod. 16, 471e476. Belotti, P., Kirches, C., Leyffer, S., Linderoth, J., Luedtke, J., Mahajan, A., 2013. Mixedinteger nonlinear optimization. Acta Numer. 22, 1e131. Bradley, S., Hax, A., Magnanti, T., 1977. Applied Mathematical Programming. Addison-Wesley, Reading, MA. Brepols, C., Sch€ afer, H., Engelhardt, N., 2009. Economic aspects of large scale membrane bioreactors. In: Workshop “Membrane Technologies for Alternative Water Resources”, Thessaloniki, Greece, pp. 29e30. Chen, Z., Ng, H.H., Guo, W., 2012. A critical review on sustainability assessment of recycled water schemes. Sci. Total Environ. 426, 413e431. Chung, G., Lansey, K., Blowers, P., Brooks, P., Ela, W., Stewart, S., Wilson, P., 2008. A general water supply planning model: evaluation of decentralized treatment. Environ. Model. Softw. 23 (7), 893e905. ^ te , P., Masini, M., Mourato, D., 2004. Comparison of membrane options for water Co reuse and reclamation. Desalination 167, 1e11. DeCarolis, J.F., Hirani, Z.M., Adham, S., 2007. Evaluation of Newly Developed Membrane Bioreactor Systems for Water Reclamation. Desalination and water purification R&D program report 147. http://www.usbr.gov/tsc/water/media/ pdfs/report147.pdf (Accessed 25 June 2016). Dolnicar, S., Hurlimann, A., Grun, B., 2011. What affects public acceptance of recycled and desalinated water? Water Res. 45, 933e943. Elimelech, M., Phillip, W.A., 2011. The future of seawater desalination: energy, technology, and the environment. Science 333, 712e717. Ghaffour, N., Missimer, T.M., Amy, G.L., 2013. Technical review and evaluation of the economics of water desalination: current and future challenges for better water supply sustainability. Desalination 309, 197e207. Ghassemi, S.A., Danesh, S., 2013. A hybrid fuzzy multi-criteria decision making approach for desalination process selection. Desalination 313, 44e50. Gibson, F.L., Tapsuwan, S., Walker, I., Randrema, E., 2015. Drivers of an urban community's acceptance of a large desalination scheme for drinking water. J. Hydrol. 528, 38e44. Gonzalez-Serrano, E., Rodriguez-Mirasol, J., Cordero, T., Koussis, A.D., Rodriguez, J.J., 2005. Cost of reclaimed municipal wastewater for applications in seasonally stressed semi-arid regions. J. Water Supply Res. Technol. AQUA 54 (6), 355e370. Greenlee, L.F., Lawler, D.F., Freeman, B.D., Marrot, B., Moulin, P., 2009. Reverse osmosis desalination: water sources, technology, and today's challenges. Water Res. 43 (9), 2317e2348. Hospido, A., Sanchez, I., Rodriguez-Garcia, G., Iglesias, A., Buntner, D., Reif, R., Moreira, M.T., Feijoo, G., 2012. Are all membrane reactors equal from an environmental point of view? Desalination 285 (0), 263e270. Iglesias, R., Ortega, E., Batanero, G., Quintas, L., 2010. Water reuse in Spain: data overview and costs estimation of suitable treatment trains. Desalination 263 (1e3), 1e10. Joksimovic, D., Kubik, J., Hlavinek, P., Savic, D., Walters, G., 2006. Development of an integrated simulation model for treatment and distribution of reclaimed water. Desalination 188 (1e3), 9e20. Karagiannis, I.C., Soldatos, P.G., 2008. Water desalination cost literature: review and assessment. Desalination 223 (1e3), 448e456. Khawaji, A.D., Kutubkhanah, I.K., Wie, J.-M., 2008. Advances in seawater desalination technologies. Desalination 221 (1e3), 47e69. Lamei, A., van der Zaag, P., von Münch, E., 2008. Basic cost equations to estimate unit production costs for RO desalination and long-distance piping to supply water to tourism-dominated arid coastal regions of Egypt. Desalination 225 (1e3), 1e12. Leverenz, H.L., Tchobanoglous, G., Asano, T., 2011. Direct potable reuse: a future imperative. J. Water Reuse Desal. 1 (1), 2e10. Loutatidou, S., Chalermthai, B., Marpu, P.R., Arafat, H.A., 2014. Capital cost estimation of RO plants: GCC countries versus southern Europe. Desalination 347,

1005

103e111. Lund, J.R., 2015. Integrating social and physical sciences in water management. Water Resour. Res. 51, 5905e5918. Mahgoub, M., van der Steen, N.P., Abu-Zeid, K., Vairavamoorthy, K., 2010. Towards sustainability in urban water: a life cycle analysis of the urban water system of Alexandria City. Egypt. J. Clean. Prod. 18, 1100e1106. Metcalf, Eddy, 2013. Wastewater Engineering: Treatment and Resource Recovery, fifth ed. McGraw Hill, New York. Miller, J.E., 2003. Review of Water Resources and Desalination Technologies. Sandia National Laboratories: Materials Chemistry Department (SAND 2003-0800). Available at: www.semide.net/media_server/files/a/p/MillerSAND2003_0800. pdf (Accessed 20 June 2016). Molinos-Senante, M., Garrido-Baserba, M., Reif, R., Hern andez-Sancho, F., Poch, M., 2012. Assessment of wastewater treatment plant design for small communities: environmental and economic aspects. Sci. Total Environ. 427e428 (0), 11e18. Molinos-Senante, M., Gomez, T., Caballero, R., Hern andez-Sancho, F., SalaGarrido, R., 2015. Assessment of wastewater treatment plant design for small communities: an analytic network process approach. Sci. Total Environ. 532, 676e687. Moore, F.C., Diaz, D.B., 2015. Temperature impacts on economic growth warrant stringent mitigation policy. Nat. Clim. Change 5, 127e131. Morais, D.C., Almeida, A.T., 2007. Water supply system decision making using multicriteria analysis. Water SA 32 (2), 229e236. Morris, J., Paltsev, S., Reilly, J., 2008. Marginal Abatement Costs and Marginal Welfare Costs for Greenhouse Gas Emissions Reductions: Results from the EPPA Model. Report No. 164, Joint Program on the Science and Policy of Global Change. MIT, Cambridge. http://globalchange.mit.edu/files/document/ MITJPSPGC_Rpt164.pdf (Accessed 30 June 2016). Oelkers, E.H., Hering, J.G., Zhu, C., 2011. Water: is there a global crisis? Elements 7 (3), 157e162. Pasqualino, J.C., Meneses, M., Castells, F., 2011. Life cycle assessment of urban wastewater reclamation and reuse alternatives. J. Ind. Ecol. 15 (1), 49e63. Perez-Gonzalez, A., Urtiaga, A.M., Ibanez, R., Ortiz, I., 2012. State of the art and review on the treatment technologies of water reverse osmosis concentrates. Water Res. 46, 267e283. Plappally, A.K., Lienhard, J.H., 2012. Costs for water supply, treatment, end-use and reclamation. Desal. Water Treat. 51 (1e3), 200e232. Raluy, G., Serra, L., Uche, J., 2006. Life cycle assessment of MSF, MED and RO desalination technologies. Energy 31 (13), 2361e2372. Ross, V.L., Fielding, K.S., Louis, W.L., 2014. Social trust, risk perceptions and public acceptance of recycled water: testing a social-psychological model. J. Environ. Manag. 113, 158e169. Ruiz-Rosa, I., Garcia-Rodriguez, F.J., Mendoza-Jimenez, J., 2016. Development and application of a cost management model for the wastewater treatment and reuse processes. J. Clean. Prod. 113, 299e310. Saavedra, E.R., Gotor, A.G., B aez, S.O., Martín, A.R., 2013. Estimation of the maximum conversion level in reverse osmosis brackish water desalination plants. Desal. Water Treat. 51 (4e6), 1143e1150. Sadr, S.M.K., Saroj, D.P., Kouchaki, S., Ilemobade, A.A., 2015. A group decisionmaking tool for the application of membrane technologies in different water reuse scenarios. J. Environ. Manag. 156, 97e108. Saloranta, T.M., Kamari, J., Rekolainen, S., Makve, O., 2003. Benchmark criteria: a tool for selecting appropriate models in the field of water management. Environ. Manag. 32 (3), 322e333. Schwarzenbach, R.P., Egli, T., Hofstetter, T.B., von Gunten, U., Wehrli, B., 2010. Global water pollution and human health. Annu. Rev. Environ. Resour. 35, 109e136. Shahabi, M.P., McHugh, A., Anda, M., Ho, G., 2014. Environmental life cycle assessment of seawater reverse osmosis desalination plant powered by renewable energy. Renew. Energy 67, 53e58. Shatat, M., Riffat, S.B., 2014. Water desalination technologies utilizing conventional and renewable energy sources. Int. J. Low Carbon Technol. 9 (1), 1e19. Sudhakaran, S., Lattemann, S., Amy, G.L., 2013. Appropriate drinking water treatment processes for organic micropollutants removal based on experimental and model studies - a multi-criteria analysis study. Sci. Total Environ. 442 (0), 478e488. Tarnacki, K., Meneses, M., Melin, T., van Medevoort, J., Jansen, A., 2012. Environmental assessment of desalination processes: reverse osmosis and Memstill®. Desalination 296 (0), 69e80. Tsai, J.F., Lin, M.H., 2008. Global optimization of signomial mixed-integer nonlinear programming problems with free variables. J. Glob. Optim. 42 (1), 39e49. Uche, J., Martínez-Gracia, A., Círez, F., Carmona, U., 2015. Environmental impact of water supply and water use in a Mediterranean water stressed region. J. Clean. Prod. 88, 196e204. Viessman Jr., W., Hammer, M.J., Perez, E.M., Chadik, P.A., 2008. Water Supply and Pollution Control, eighth ed. Pearson Education, Inc., Upper Saddle River, NJ. Wenten, I.G., Khoiruddin, 2016. Reverse osmosis applications: prospect and challenges. Desalination 391, 112e125. Wittholz, M.K., O'Neill, B.K., Colby, C.B., Lewis, D., 2008. Estimating the cost of desalination plants using a cost database. Desalination 229 (1e3), 10e20. WHO (World Health Organization), 2008. Guidelines for Drinking-water Quality, 3 edn. WWAP (World Water Assessment Programme), 2012. The United Nations World Water Development Report 4: Managing Water under Uncertainty and Risk. UNESCO, Paris.

1006

D. Abdulbaki et al. / Journal of Cleaner Production 164 (2017) 994e1006

Yohe, G.W., et al., 2007. Executive summary. In: Parry, M.L., et al. (Eds.), Perspectives on Climate Change and Sustainability. Climate Change 2007: Impacts, Adaptation and Vulnerability. Contribution of Working Group II to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change. http:// www.ipcc.ch/publications_and_data/ar4/wg2/en/ch20s20-es.html (Accessed 29 June 2016). Younos, T., 2005. The economics of desalination. J. Contemp. Water Res. Educ. 132,

39e45. Zhou, J., Chang, V.W.-C., Fane, A.G., 2014. Life Cycle Assessment for desalination: a review on methodology feasibility and reliability. Water Res. 61, 210e223. Zoltay, V.A., Vogel, R.M., Kirshen, P.H., Westphal, K.S., 2010. Integrated watershed management modeling: generic optimization model applied to the Ipswich River Basin. J. Water Resource Plan. Manag. 136 (5), 566e575.