Natural Resources Research (Ó 2018)

https://doi.org/10.1007/s11053-018-9440-1

Original Paper

Optimal Allocation of Water Resources Using a Two-Stage

Stochastic Programming Method with Interval and Fuzzy
Parameters

Tooraj Khosrojerdi,1 Seyed Hamed Moosavirad ,1,3 Shahram Ariafar,1

and Mahnaz Ghaeini-Hessaroeyeh2

Received 26 July 2018; accepted 30 November 2018

Efficient water allocation is one of the most prominent issues in water resources manage-
ment. In this research, a two-stage interval-parameter stochastic fuzzy programming with
type 2 membership functions was used to allocate water resources optimally to different
users under uncertainty. This method can handle uncertainties expressed as probability
distributions, discrete intervals, and fuzzy sets. The model considers treated wastewater as an
allocable water resource in a scenario, in addition to water that is extracted from wellheads,
springs, and qanats. Moreover, the loss rate of water during distribution, surplus water in the
reservoir in the previous and the next period, and treated wastewater parameters have been
incorporated into the model. This model was applied to a case study of water resources
allocation within a multi-user and multi-reservoir context in the Zarand region of Kerman,
Iran. The results indicate that reasonable solutions have been generated, in the form of
interval and fuzzy information under different scenarios that could help managers provide
optimal water resources allocation plans. The results also demonstrate that establishing a
wastewater treatment station increases the system net benefits, surplus water in wellheads
for the next period, and system reliability (level of satisfying the fuzzy goal and constraints),
and decreases encountering water shortages.
KEY WORDS: Water resources management, Optimal water allocation, Mathematical programming
model, Optimization, Uncertainty.

INTRODUCTION and even under the ground, and beyond the earth in
clouds and the atmosphere. However, 97.5% of this
Water, a blessed gift bestowed by God that amount is comprised of saline waters and thus only
covers  70% of our planetÕs surface, is found all 2.5% is fresh, and the latter is not distributed equally
around the earth, in lakes, rivers, oceans, icebergs across the planet (Shiklomanov 1998). Moreover,
only 0.4% of the total volume of freshwater avail-
1
Department of Industrial Engineering, Faculty of Engineering, able to humans is in the form of lakes and rivers, and
Shahid Bahonar University of Kerman, Kerman, Islamic unlike other resources such as oil, which can be
Republic of Iran. utilized in different forms, there are no substitutes
2
Department of Civil Engineering, Faculty of Engineering, Shahid (Guo et al. 2010).
Bahonar University of Kerman, Kerman, Islamic Republic of
Water scarcity is a serious concern in many
Iran.
3
To whom correspondence should be addressed; e-mail: countries that limits the development of the indus-
[email protected] trial and agricultural sectors as well as the overall

Ó 2018 International Association for Mathematical Geosciences

 T. Khosrojerdi et al.

aspects of human life (Li et al. 2015). Water shortage decision is made before the realization of random
occurs due to the population growth, global eco- variables; then, the second decision can be deter-
nomic development, urbanization and industrializa- mined to minimize penalties after the random event
tion, climate change, water infrastructure has taken place, and their values are known. The
destruction, and poor water quality (Sophocleous process of making the first decision is called the first
2004; Xie et al. 2013). Today, many parts of the stage, and the relevant variables are called first-stage
planet face scarcity of water resources. Based on variables. The process of making other decisions is
statistics released by the United Nations,  700 called the second stage, and the relevant variables
million people in 43 different countries are suffering are called second-stage variables.
from water shortage. It is predicted that by 2025, 1.8 Pereira and Pinto (1985) proposed a stochastic
billion people who live in regions and countries optimization approach for the planning of a multi-
faced with water shortages and 3–4 million people reservoir hydroelectric system under uncertainty.
die annually due to water shortage or related dis- Wang and Adams (1986) introduced a two-stage
eases (Wang and Huang 2015). Moreover, it is re- optimization framework for optimal water reservoir
ported that the agriculture sector will likely face the operations. In their research, periodic Markov pro-
problem of reduced water availability and the need cesses were used to describe seasonality of water
to produce more food in the near future (Dogra resources inflows. Ferrero et al. (1998) examined the
et al. 2014). hydrothermal scheduling of multi-reservoir systems
Over the past decades, optimal and efficient using a two-stage dynamic programming approach.
water allocation has challenged many water resource Seifi and Hipel (2001) proposed a method for the
managers. Ineffective allocation of water resources long-term planning of reservoirs with stochastic in-
and farmland further aggravated the conflict among flows. They used TSP and an interior-point method
the water users (Dong et al. 2018). This ineffective for optimizing reservoir operation. Ahmed et al.
allocation exacerbated the problem of water pollu- (2003) also used a limited branch and boundary
tion (Dong et al. 2018). Water allocation planning is algorithm to solve the two-stage stochastic planning.
an essential component of managing water uses, and Previous research on the TSP method does not
it involves deciding how much water is available reflect the dynamic changes in system conditions
from a particular resource and how much water can (especially in high-volume systems). Thus, multi-
be taken (Sophocleous 2004; Shao et al. 2011). In stage stochastic programming (MSP) method has
other words, equity in water allocation is a signifi- been developed by the recent researchers (Yin and
cant challenge that needs to be considered by Han 2015; Hu and Hu 2018; Zahiri et al. 2018). In
authorities and decision makers (Shukla and Gedam MSP, recourse actions are permitted for each period
2018). based on uncertainties related to specified values.
Optimization techniques play a vital role in The most prominent advantage of MSP is its flexi-
helping the decision makers optimally allocate water bility in the decision-making and scenario-setting
resources to different users. The variety of uncer- process, although period information should be
tainties existing in water resources management may investigated. Among the conducted research, quan-
intensify the complexity in decision-making process; titative researchers have implemented this method
therefore, conventional optimization techniques independently in the field of water resources man-
such as linear programming, quadratic program- agement. For example, Pereira and Pinto (1991)
ming, and integer programming would become introduced a multistage stochastic optimization
ineffective (Wang and Huang 2015). method in hydrological energy system planning.
Among the variety of methods used in water Watkins et al. (2000) presented an MSP model to
resources management for dealing with uncertain- utilize water from mountain lakes. Modeling with
ties, two-stage stochastic programming (TSP), as a MSP to manage water resources is identical to the
kind of stochastic optimization method, was widely two-stage approach.
used for dealing with randomness in such systems The role of wastewater in todayÕs world where
and has the ability to take corrective actions after a the demand for water is rising sharply, and the
random event occurs (Brige and Louveaux 1988; earthÕs water resources are consequently decreasing,
Tajeddini et al. 2014; Parisio and Jones 2015). TSP is is becoming more and more prominent. Under such
a scenario-based approach and is useful in analyzing circumstances, the use of unconventional water
long-term and midterm schemes. In TSP, at first, a including sewage from refinement plants in various
Optimal Allocation of Water Resources Using a Two-Stage Stochastic Programming Method

sectors—particularly for the agricultural industry

that uses the majority of water—is inevitable. The " #
X
m X
m
year 2017 was named the Year of Wastewater by the max f ¼ NBi Ti  E Ci DiQ ð1Þ
World Water Forum, which indicates the impor- i¼1 i¼1
tance of wastewater about water reuse and quality of
wastewater discharges. Therefore, this research aims subject to:
to optimally manage and allocate water resources by
considering the uncertainty and vague parameters. X
m  
This research focuses on the city of Zarand, Kerman Q Ti  DiQ ð2Þ
province, Iran, as a case study (which consists of i¼1
three kinds of consumers and three water resources
where no wastewater refinement is carried out). In Ti max  Ti  DiQ  0 ð3Þ
addition, in the form of a scenario, refined wastew-
ater was considered as an allocable water supply and In this model, NBi is the net benefit of consumer i
then compared with the case where refined per unit of water allocated, Ti is allocation target of
wastewater was not considered as a water supplier. water that is promised to consumer i (cubic meter),
Hence, a designated wastewater refinement plant Ci is reduction in net benefit to user i per unit of
with specified volume was considered and applied water not allocated, DiQ is amount by which water
within the model where wastewater that can be re- allocation target is not met when seasonal flow is Q,
fined was collected from all areas, and subsequent E [0] is the expected value of a random variable, f is
refinement operations were carried out and re- net system benefit, m is number of water consumers,
transferred to applicant sections. Q is random variable equal to total water available,
and Ti max is maximum allowable allocation amount
for consumer i.
WATER ALLOCATION MODELING Equation 1 is an objective function that maxi-
mizes the expected value of a regionÕs economic
The world today is witnessed to increasing de- activity such that if the promised amount is deliv-
mand for water due to population growth, industrial ered, it will lead to the net benefit for the local
and agricultural development, urbanization, and economy. However, if not delivered, it leads to pe-
climate changes (Wang et al. 2016). In order to de- nalties in the local economy. Equation 2 shows the
velop the agricultural and industrial sectors, these first constraint that water allocated from different
disparate groups of water users need to be aware of resource is less than the amount of water available;
how much water they can expect for their activities and Eq. 3 shows the second constraint that the
and economic investments. This information is ex- maximum allowable allocation amount is more than
tremely necessary for planning because, if the pro- the initial allocation target and the optimal shortage
mised water cannot be delivered, they will have to of consumers.
obtain water from higher priced alternatives or In order to solve this model by linear pro-
make negative changes in their development plans. gramming, the probability distribution of Q must be
In other words, there is inherent complexity and approximated by a discrete function. If Q takes q
uncertainty in water resource decision making that values with probability P, then:
placed them beyond the conventional deterministic
optimization methods (Maqsood et al. 2005). " # !
X
m X
m X
n
E Ci DiQ ¼ Ci Pj Dij ð4Þ
i¼1 i¼1 j¼1
Water Allocation Modeling by TSP
where Dij is water shortage for consumer i, when
A region is considered that contains a number seasonal flow is qj with probability of Pj , n is number
of water resources and water managers, which are of different flow levels (low, medium, high), Pj is
responsible for allocating water to different con- probability of occurrence flow level j (low, medium,
sumers including agricultural, industrial, and high).
municipal sectors. This problem can be formulated By using Eq. 4 in Eq. 1, the TSP is formulated
as (Loucks et al. 1981): as:
 T. Khosrojerdi et al.

!
X
m X
m X
n x ¼ ½x ; xþ  ¼ ft 2 xjx  t  xþ g ð8Þ
max f ¼ NBi Ti  Ci Pj Dij ð5Þ
i¼1 i¼1 j¼1 where x , xþ represent the lower and upper bounds
of x , respectively.
subject to:
Huang and Loucks (2000) presented a TSP
X
m   method with interval parameters or inexact two-
qj  Ti  Dij 8j ð6Þ stage stochastic programming (ITSP) approach to
i¼1
allocate water to different consumers. Their re-
search has been used extensively for water resources
Ti max  Ti  Dij  0 8i:j ð7Þ
management and environmental planning, and pro-
TSP is not easily implemented in the real-world vides solutions for generating decision alternatives
cases of water resources management because it is and identifies and analyzes significant factors that
difficult for a planner to determine a deterministic affect a systemÕs performance (Huang and Loucks
value of parameters such as allocation targets Ti , the 2000). ITSP is an effective measure for addressing
net benefit of water allocated NBi or the penalty of problems where an analysis of policy scenarios is
water not delivered Ci . In addition, the inadequate desired periodically over time and uncertain
quality of the information to present them as prob- parameters have probability distribution functions
ability distribution has made TSP a challenging and interval numbers (Xie et al. 2018). Huang et al.
method (Birge and Louveaux 1988). Therefore, (1993, 1995) proposed an IPP to manage water re-
incorporating interval-parameter programming sources in an agricultural environment. Li and
(IPP) within a TSP framework is a practical ap- Huang (2008) also combined TSP and IPP in a
proach to better reflecting uncertainties. IPP was nonlinear programming framework. Their model
introduced by Huang et al. (1995), and it has been considers several water resources, demand regions
widely used in various fields, especially in water re- and a bunch of water end users, and responses as
sources management. Huang and Loucks (2000) interval parameters maximize economic benefit and
have used IPP in the TSP framework since 2000. The minimize failure risks. Xie et al. (2013) applied ITSP
reasons for using such method in water resources by considering multiple water resources and several
management can be summarized as follows: consumers. Various scenarios corresponding to dif-
ferent river inflow levels were evaluated. The results
 The quality of data obtained from uncertain indicated that different inflow levels could lead to
parameters is not satisfactory enough to be different water allocation schemes with variation in
used to obtain the probability distribution system benefit and system failure risk. Zhang and Li
function. Besides, even if these distributions (2014) used ITSP for sustainable development and
are available, their reflection in large-scale water resources management and applied it to an
TSP models can be extremely challenging area in China that includes multiple water resources
(Huang and Loucks 2000). In other words, in and some regions and users. Fu et al. (2016) pre-
order to obtain the probability distribution sented an ITSP method based on adaptive water
function of an uncertain parameter, accurate resource management (AWRM). In their model, the
and reliable information is required (Fan cost of water exchange between different regions is
et al. 2011). considered, and results indicated that optimistic
 It is hard to solve the TSP model with all water policies lead to higher income but may be
uncertain parameters being expressed as a subject to higher risk of system failure, and the re-
probability density function (Fan et al. 2015). sults can be used by managers to adjust investment
 Using and presenting information as interval activities and avoid making inappropriate decisions.
and fuzzy parameters is more favorable than Ji et al. (2017) applied the ITSP model to manage
deterministic parameters (Fan et al. 2011). water resources in Tianjin, China (a coastal city
facing severe water shortage). In their optimization
If x is a closed set of real numbers, then x is a model, water supply cost and sewage treatment cost
gray number with upper and lower bounds but with were considered besides water utilization benefits
unknown distribution; thus, and water shortage penalty. Liu et al. (2017) pro-
Optimal Allocation of Water Resources Using a Two-Stage Stochastic Programming Method

posed a two-stage regional multi-water source allo- subject to:

cation model by considering surface, ground, and
m 
X 
transit water. By solving the model of Liu et al.
q
j  Ti þ DTi yi  D
ij 8j ð15Þ
(2017), optimized water supply target and shortage, i¼1
optimized water allocation, and satisfaction of sup-
ply targets in each user sector have been obtained. Timax  Ti þ DTi yi  D
ij  0 8i:j ð16Þ

Modeling Optimal Water Resources Allocation

0  yi  1 ð17Þ
Using ITSP
where yi is the decision variable. On this basis, when
Water resources allocation within ITSP frame- Ti reaches its upper bound ( yi ¼ 1), higher benefits
work is modeled as follows: will be obtained while the risk of unsatisfying pro-
mised allocation targets reaches upper bound. Con-
X
m m X
X n versely, if Ti reaches its lower bound ( yi ¼ 0), there
max f  ¼ NB 
i Ti  Pj Ci D
ij ð9Þ will be a lower benefit but with lower risks of the
i¼1 i¼1 j¼1
system.
subject to:
m 
X  Modeling Optimal Water Resources Allocation
q
j  Ti  D
ij 8j ð10Þ Using Interval Fuzzy TSP
i¼1

Although ITSP has been used extensively in

many studies (Huang and Loucks 2000; Xie et al.
Timax  Ti  D
ij  0 8i:j ð11Þ 2018), IPP is not applicable in many real-world

problems, because vague information may exist in
where Timax :Ci :D 
ij :NBi :Ti are interval variables the optimization model (objective function and

and parameters. If Ti represents promised water constraints). Moreover, in IPP, uncertainties in
allocation targets for different consumers, which are parameters are presented as discrete intervals; thus,
considered as uncertain parameters, it would be the upper and lower bounds of these intervals may
difficult to determine which parameter bounds ( Tiþ be known with uncertainty that results in dual
as the upper bound or Ti as the lower bound) will uncertainties. An effective way to overcome these
correspond to the upper bound of system benefit ( complexities is to incorporate fuzzy programming
f þ ). In these circumstances, the existing methods (FP) with type 2 membership function within the
will not be applicable for solving the interval linear framework of ITSP (Maqsood et al. 2005) that leads
programming problems (Huang and Loucks 2000). to an interval fuzzy TSP (IFTSP).
Therefore, an optimized set of allocation targets can FP is applicable when the information of the
be obtained by including yi in the model as decision problem is vague. In other words, the system
variables such as below: impreciseness can be addressed by FP. It is also
useful to be used in decision-making problems with
Ti ¼ Ti þ DTi yi ð12Þ the fuzzy objective function and constraints (Zim-
mermann 1985). If G is fuzzy objective and C is
DTi ¼ Tiþ  Ti ð13Þ fuzzy constraint in X decision alternative, then fuzzy
By incorporating Eqs. (12) and (13) in Eq. (9), decision set D will be the intersection between G
ITSP can be formulated: and C (Fig. 1) and is shown by the following equa-
tions:
X
m    Xm X
n D¼G \C ð18Þ
max f  ¼ NB
i Ti þ DT y
i i  Pj Ci D
ij
i¼1 i¼1 j¼1

ð14Þ lD ¼ minflG ; lC g ð19Þ

 T. Khosrojerdi et al.

µ the water resource management where many of its

information have uncertainties. Zeng et al. (2017)
Constraint C Goal G applied IFTSP for water resources allocation and
water quality management in a river basin in China
by considering sever water deficit and water quality
degradation. In their model, water allocation pat-
Decision D tern, pollution mitigation scheme, and system ben-
efit under various scenarios were analyzed.
Optimal water resource allocation can be
modeled using FP within the ITSP framework as
Figure 1. Fuzzy decision theory (Wang and Huang 2013). follows:

max a ð21Þ

subject to:
where lD ; lG ; lC are fuzzy decision membership
functions, fuzzy goal, and fuzzy constraints, respec- X
m    Xm X
n
NB
i Ti þ DTi y i  Pj Ci D
ij
tively. The desired decision is the one with the i¼1 i¼1 j¼1
highest lD value: fa
 f  þ Df ð22Þ
max lD ¼ max minflG :lC g ð20Þ
m 
X 
Ordinary fuzzy sets are not applicable if the q gj a 
 Dq Ti þ DTi yi  D 8j ð23Þ
j ij
membership functions are given imprecisely. i¼1
Therefore, in practical problems, the appropriate
choice of membership functions for fuzzy goal and Timax  DTi max a  Ti þ DTi yi ð24Þ
constraints is one of the critical issues. To reflect
such issues, the concept of type 2 fuzzy sets is Ti þ DTi yi  D
ij 8i:j ð25Þ
effective in addressing such uncertainties. The con-
cept of type 2 fuzzy sets is shown in Figure 2. D
ij  0 8i:j ð26Þ
Maqsood et al. (2005) proposed an IFTSP to
manage water resources for allocating water opti- 0  yi  1 8i ð27Þ
mally to different consumers. In their optimization
model, type 2 fuzzy membership function has been 0a1 ð28Þ
used. Wang and Huang (2013) used IFTSP to reflect
uncertainties in the objective function and right- In this model, a is degree of satisfaction of the
hand side of constraints as fuzzy and interval fuzzy objective or constraints (optimal system relia-
parameters. The results indicate that several interval bility), f  is lower bound of net system benefit where it
solutions can be obtained under various scenarios, 
is expressed as intervals, f is upper bound of net
which enhances the diversity of solutions for sup- system benefit, q is upper bound of water inflow, q
porting decisions of water resources allocation. 
Fan et al. (2015) also used IFTSP in their study is lower bound of water inflow, f is lower interval of
and obtained a series of fuzzy interval solutions upper-bound net system benefit, f  is lower interval
þ
under different a-cut levels. Zhou et al. (2016) pro- of lower-bound net system benefit, f is upper inter-
posed an IFTSP approach to support water re- val of upper-bound net system benefit, f þ is upper
sources management under dual uncertainties. They interval of lower-bound net system benefit. Equa-
took into account uncertainties as fuzzy parameters tion 29 is established for this model; thus,
and variability of a-cut levels. Their results showed h i h i
that any change in a-cut level could affect the solu- f ¼ f   f  ¼ f ; f þ  f ; f þ :
Df
tions. Therefore, fuzzy sets are highly applicable in h  þ
i
¼ f  f þ; f  f  ð29Þ
Optimal Allocation of Water Resources Using a Two-Stage Stochastic Programming Method

Constraint C Goal G

Decision D̂

− + − + − + − +
G G C C C C G G X
± ± ± ±
G C C G
̅± and ±
: upper and lower bounds of the type-2 membership fuzzy goal
̅ ± and ±
: upper and lower bounds of the type-2 fuzzy constraint
: fuzzy decision

Figure 2. Fuzzy decision theory with type 2 membership function (Wang and Huang 2013).

IFTSP Solution Method Optimized targets allocation values Tiopt can be

obtained as:
To solve modelÕs objective ðmax a Þ, we should
transform it into two deterministic sub-models. Ini- Tiopt ¼ Ti þ DTi yi ð38Þ
tially, the sub-model corresponding to the most
The next step is to solve the sub-model corre-
desirable system objective value ðf þ Þ is solved, and
sponding to f  :
then, the results are used to solve the sub-model
corresponding to f  . The sub-model to find f þ is:
max a ð39Þ
max aþ ð30Þ subject to:
subject to: X
m    Xm X
n
NB
i T i þ DT y
i i opt  Pj Ciþ Dþ
ij
X
m    Xm X
n i¼1 i¼1 j¼1
NBþ T þ DT y  Pj Ci D fa
i i i i ij  f  þ Df ð40Þ
i¼1 i¼1 j¼1
fa
 f þ þ Df ð31Þ m  
X
q g Ti þ DTi yi opt  Dþ
j  Dqj a  ij 8j ð41Þ
m 
X  i¼1
qþ g Ti þ DTi yi  D
j  Dqj a  ij 8j ð32Þ
i¼1 Timax  DTi max a  Ti þ DTi yi opt ð42Þ

Tiþmax  DTi max a  Ti þ DTi yi ð33Þ Ti þ DTi yi opt  Dþ 8i:j ð43Þ
ij

Ti þ DTi yi  D 8i:j ð34Þ

ij Dþ 
ij  Dij opt 8i:j ð44Þ
D  0 8i:j ð35Þ
ij
0  a  1 ð45Þ
0  yi  1 8i ð36Þ In this sub-model, Dþ
is the decision variable;
ij
thus, all optimal solutions for modelÕs objective
0  aþ  1 ð37Þ ðmax a Þ are as follows:
h i
In this sub-model, D
ij and yi are decision vari- D ¼ D 
; D þ
ð46Þ
ij opt ij opt ij opt
ables, and aþ is upper system reliability bound.
 T. Khosrojerdi et al.

h i jor assigned water to this region. In recent years, this

  þ
fopt ¼ fopt ; fopt ð47Þ region faced water shortage due to the leak of pre-
cipitation besides the agricultural and a rise in
industrial sectorsÕ water demand. Statistical reports
A  
ij opt ¼ Ti opt  Dij opt ð48Þ gathered by Kerman regional water company show
that the annual precipitation of Zarand city was less
where D ij opt is optimal shortage for consumers un-
than 130 mm. In this research, according to the
der different flow levels, fopt
is optimal objective conditions of the case study and based on the
opinions of water resources experts in Kerman
function value that is net system benefit, and A
ij opt is province, a mathematical programming model is
optimal water resources allocation to consumers presented using the IFTSP method. In this region,
under different water flow levels. there are three water resources including wellhead,
In summary, IFTSP can be overcome by the spring, and qanat. The water consumers are the
uncertainty embodied in water resources manage- agriculture, industry, and municipal sectors. The
ment by using the interval and vague parameters. model proposed in this study is described in fol-
Table 1 represents the comparison among TSP, lowing subsections.
ITSP, and IFTSP.

Problem Formulation
CASE STUDY
Objective Function
Problem Statement
This problem can be formulated as an IFTSP
This research aims to use IFTSP for optimally model:
managing and allocating water resources. In this
optimization model, treated wastewater was consid- max a ð49Þ
ered as allocable water. Therefore, the related
parameters have been incorporated into the model. In
addition, loss rate of water during transportation and Constraints
surplus water in the reservoirs have been considered.
In order to develop the agricultural and industrial
sectors, these disparate groups of water users need to  Desirable system benefit constraint
be aware of how much water they can expect for their
activities and economic investments. This informa-
tion is essential for planning because, if the promised
water cannot be delivered, they will have to obtain
water from higher priced alternatives or make nega-
tive changes in their development plans. X
m    Xm X
n
NB
i Tai þ ðDTai :yai Þ  Paj :Ci :Da
ij
i¼1 i¼1 j¼1
X
m   m X
X n
Study Area Overview þ 
NBi Tb Pbj :Ci :Db
i þ ðDTbi :ybi Þ  ij
i¼1 i¼1 j¼1
In this research, Zarand city located in the X
m m X
X n
 
Kerman province of Iran was selected as a case þ NBi Tc

i þ ðDTci :yci Þ  Pcj :Ci :Dc
ij
study. This area is part of the Bafq swamp southern i¼1 i¼1 j¼1
basin that has been surrounded from the north and X
m   
east by Kerman hills, from the west by Davaran hills, þ NB
i Twi þ ðDTwi :ywi Þ
and from the south by Kerman plain. The area is i¼1

about 8420 square meters. It is located in a hot and X

m X
n
 Pwj :Ci :Dw f 
dry region that has mining sites and industries be- ij  a Df  f ;
i¼1 j¼1
sides huge pistachio gardens. Therefore, these
industrial and agricultural sectors consume the ma- ð50Þ
Optimal Allocation of Water Resources Using a Two-Stage Stochastic Programming Method

Table 1. Comparison of the TSP, ITSP, and IFTSP

Vague parameters Interval parameters Probabilistic parameters Deterministic parameters

4 4 TSP
4 4 4 ITSP
4 4 4 4 IFTSP

Equation 50 shows that the net benefit for various

Sa1 
j ¼ qa"
j  aDfqa j
consumers minus the penalties (because of not #
m h
X i
meeting the promised amount of water) should be   
Tai þ ðDTai :yai Þ  Daij ð1 þ eÞ
higher than the desired system benefit. i¼1
þ Sa0
j
 Available water quantity constraints 8j ¼ 1:2:3
ð54Þ
m h
X i Equation 54 expresses that the surplus water in the
Ta 
i þ ðDTai :yai Þ  Daij ð1 þ eÞ þ aDf
qa j resource in the present period is equal to total
i¼1
available water in the reservoir plus the surplus
 qa 
j þ sa0j
water in resource in the previous period minus the
8j ¼ 1:2:3 allocated water to various consumers.
ð51Þ
 Wastewater treatment constraints
m h
X i m h
X i
Tb þ ð DTb :yb Þ  Db  f
ð1 þ eÞ þ aD qb
i i i ij j Tw 
i þ ðDTwi :ywi Þ  Dwij
i¼1
 i¼1
 qbj h 
ð 1 þ e Þ  h1 b 1 Ta 
1 þ ðDTa1 :ya1 Þ  Da1j
8j ¼ 1:2:3
 
ð52Þ þ Tb 1 þ ðDTb1 :yb1 Þ  Db1j


 
m h
X i þ Tc þ ð DTc :yc Þ  Dc 
1 1 1 1j
Tc 
i þ ðDTci :yci Þ  Dcij ð1 þ eÞ þ aDf
qc j  i
i¼1 þ Tw 1 þ ðDTw1 :yw1 Þ  Dw1j

 qcj h 
8j ¼ 1:2:3 þ h2 b2 Ta2 þ ð DTa 2 :ya 2 Þ  Da
2j
 
ð53Þ
þ Tb 2 þ ðDTb2 :yb2 Þ  Db2j


 
Equations 51–53 illustrate that the total water allo- þ Tc þ ð DTc :yc Þ  Dc 
2 2 2 2j
cated to all water users should be less than the  i
available water flows of the water resources in order þ Tw2 þ ðDTw2 :yw2 Þ  Dw

2j
to balance the relationship between water supply h 
and demand. þ h3 b3 Ta3 þ ð DTa 3 :ya 3 Þ  Da
3j
 
 
 Constraint of available water in wellhead þ Tb3 þ ðDTb3 :yb3 Þ  Db3j
 
þ Tc 3 þ ðDTc3 :yc3 Þ  Dc3j


 i
þ Tw 3 þ ð DTw 3 :yw 3 Þ  Dw 
3j

g1
 Swt1  aD Swt 8j ¼ 1:2:3
ð55Þ
 T. Khosrojerdi et al.

Equation 55 demonstrates that total wastewater allocated and greater than shortages of each con-
allocated for various consumptions should be less sumer under all water inflow levels.
than the obtained wastewater from various uses and
the total capacity of the wastewater treatment sta-  Technical constraints
tion. 0  a  1 ð60Þ

 Water demand, water shortage, and nonneg- 0  yai :ybi :yci :ywi  1 8i ¼ 1:2:3 ð61Þ
ativity constraints
  Equations 60–61 are technical constraints.
Ta  
i max  Tai þ ðDTai :yai Þ  Daij  Da
ij  0 In this model, Ta i is allocation target for con-
8j ¼ 1:2:3 sumer i from wellheads, Tb i is allocation target for
ð56Þ consumer i from springs, Tc i is allocation target for
consumer i from the qanats, Tw i is allocation target
 
Tb  
 Db for consumer i from treated wastewater,
i max  Tbi þ ðDTbi :ybi Þ  Dbij ij  0
yai :ybi :yci :ywi are decision variables with zero to one
8j ¼ 1:2:3 values from well, springs, qanats, treated wastewater
ð57Þ to determine the optimal set of allocation targets,
  Da   
ij ; Dbij ; Dcij , Dwij are optimal water shortages
Tc
i max  Tc 
i þ ð DTc i :yc i Þ  Dc 
ij  Dc
ij  0 for consumer i from wellheads, springs, qanats, and
8j ¼ 1:2:3 refined wastewater under water inflow level j,
Pwj :Pcj :Pbj :Paj are probability of occurrence of flow
ð58Þ level for wellheads, springs, and treated wastewater,
  e is loss rate of water during distribution,
Tw  
i max  Twi þ ðDTwi :ywi Þ  Dwij  Dw
ij  0

qa :qb :qc are lower interval of the upper-bound
8j ¼ 1:2:3 wellhead, spring, and qanat water flow, qaþ :qbþ :qcþ
ð59Þ are upper interval of the lower-bound wellhead,
þ
spring, and qanat water flow, qaþ :qb :qcþ are upper
Equations 56–59 show that the maximum allowable interval of the upper-bound wellhead, spring, and
allocation target should be greater than the water qanat water flow, qa :qb :qc are lower interval of

Table 2. Description of water flow distribution

Flow level Water availability and associated probabilities

qf
a Probability f
qb Probability f
qc Probability

j¼1 ((180,182), (208,210)) 0.49 ((2, 2.5), (3.5, 4)) 0.15 ((10, 10.5), (11.5, 12)) 0.12
j¼2 ((210, 212), (258, 260)) 0.31 ((4, 4.5), (4.75, 5.25)) 0.55 ((12, 13), (15, 16)) 0.11
j¼3 ((260, 264), (276, 280)) 0.20 ((5.25, 5.75), (6.75, 7.25)) 0.30 ((16, 16.5), (17.5, 18)) 0.77

Table 3. Water resources and economic data

Users Allocation targets Economic data

Tai max (106 Tbi max (106 Tci max (106 Ta (106 Tb (106 Tc (106 NB (Rials) C (103
m3) m3) m3) m3) m3) m3) Rials)

Agriculture 220 4 15 (193.5, (3.15, 3.85) (12.6, 15.4) (11.31, 17.34) (70, 80)
223.5)
Industry 6 0 0 (2.7, 3.3) 0 0 (1650432, (80, 90)
1941685)
Municipal 20 0 0 (13.5, 16.5) 0 0 (71.58, 131.161) (3000, 5000)
Optimal Allocation of Water Resources Using a Two-Stage Stochastic Programming Method

the lower-bound wellhead, spring, and qanat water  Scenario 1:

flow, sa0
j is surplus water remaining in wellhead in
the previous period, Sa1 Taking into account the current conditions of
j is surplus water in well-
the case study (all parameter models except the
head for the next period. hi is discharge of wastew-
parameters related to wastewater treatment)
ater per unit of consumption for different
consumers, Swtþ is upper interval of upper-bound  Scenario 2:
wastewater treatment station capacity, Swt is lower
interval of lower-bound wastewater treatment sta- Considering treated wastewater as an allocable
tion capacity, Swtþ is upper interval of lower-bound resource (all model parameters)
wastewater treatment station capacity, Swt is lower Scenario 1
interval of lower-bound wastewater treatment sta- In this scenario, the current conditions of Zar-
tion capacity, Tai max :Tbi max :Tci max . Twi max are and city are considered. There are three consumer
maximum allowable allocation target from well- sectors—agricultural, industrial, and munici-
heads, springs, qanats, and treated wastewater to pal—that utilize water from wellheads, springs and
different consumers, bi is rate of recycling per unit of qanats. The information required for scenario 1 is
wastewater created. i is water consumers (1: presented in Tables 2 and 3, respectively. By con-
municipal, 2: industrial, 3: agricultural), j is various sidering the water loss rate ðeÞ to be 30% and using
water flow levels (1: low, 2: medium, 3: high). the data of Tables 2 and 3 and according to the
section IFTSP Solution Method, the model is con-
structed and solved in GAMS software.
Model Scenarios Scenario 2
This scenario improves the current conditions of
In fact, this research answers the question of the case study by establishing a wastewater treatment
‘‘how to allocate water resources to various con- station. All data are same as scenario 1, except the
sumers in order to obtain maximum economic ben- allocation target of wellhead ðTa i Þ and maximum
efits for the region.’’ This study examined two allowable allocation target from wellhead ðTa i max Þ.
scenarios:
Moreover, some parameters related to wastewater
are incorporated into the model. The modelÕs
parameters are represented in Tables 4 and 5.
Table 4. Information required for scenario 2

Users Allocation targets

RESULTS
Ta
i Tw
i Ta
i max Tw
i max
Results of Scenario 1
Agricultural (173.5, 203.5) (20, 22) 205 22
Industrial (1, 2) (2, 3) 2 2.5
Municipal (12, 15) (1, 2) 15 2 Results of the sub-model corresponding to f þ in
scenario 1 are presented in Table 6.

Table 5. Parameters related to wastewater treatment

Parameter Quantity

Discharge of wastewater per unit of consumption for municipal sector h1 70%

Discharge of wastewater per unit of consumption for industry h2 70%
Discharge of wastewater per unit of consumption for agriculture h3 10%
The municipal rate of recycling per unit of wastewater created b1 80%
The industrial rate of recycling per unit of wastewater created b2 80%
The agricultural rate of recycling per unit of wastewater created b3 80%
g
Swt ((28, 30), (31, 33))
1
 T. Khosrojerdi et al.

Table 6. Solution of the sub-model corresponding to maximum net benefit

Activity Optimal allocation targets Decision variables

Ta
i opt Tb
i opt Tc
i opt yai ybi yci

Municipal (i = 1) 13.5 3.3 193.5 0 1 0

Industry (i = 2) 0 0 3.15 0 0 0
Agriculture (i = 3) 0 0 12.6 0 0 0

Table 7. Overall results of scenario 1

Activity Users

Municipal ði ¼ 1Þ Industry ði ¼ 2Þ Agriculture ði ¼ 3Þ

Shortage Da
ij
Low ðj ¼ 1Þ 0 0 60.647
Medium ðj ¼ 2Þ 0 0 32.513
High ðj ¼ 3Þ 0 0 (0, 1.213)
Shortage Db ij
Low ðj ¼ 1Þ 0 0 (0.589, 0.895)
Medium ðj ¼ 2Þ 0 0 0
High ðj ¼ 3Þ 0 0 0
Shortage Dcij
Low ðj ¼ 1Þ 0 0 (3.885, 4.191)
Medium ðj ¼ 2Þ 0 0 (1.324, 1.936)
High ðj ¼ 3Þ 0 0 0
Optimal resources allocation amount from wellhead Aa ij
Low ðj ¼ 1Þ 13.5 3.3 132.852
Medium ðj ¼ 2Þ 13.5 3.3 160.987
High ðj ¼ 3Þ 13.5 3.3 (192.286, 193.5)
Optimal resources allocation amount from spring Ab ij
Low ðj ¼ 1Þ 0 0 (2.254, 2.560)
Medium ðj ¼ 2Þ 0 0 3.15
High ðj ¼ 3Þ 0 0 3.15
Optimal resources allocation amount from qanat Ac ij
Low ðj ¼ 1Þ 0 0 (8.408, 8.714)
Medium ðj ¼ 2Þ 0 0 (10.663, 11.275)
High ðj ¼ 3Þ 0 0 12.6
Surplus water in the wellhead for the next period Sa1 j
Low ðj ¼ 1Þ (14.952, 17.452)
Medium ðj ¼ 2Þ (28.376, 30.876)
High ðj ¼ 3Þ (5.687, 8.61)
System benefit and reliability
System net benefit f  (2178604.361, 3578184.443) million Rials
System reliability a (0.2843, 0.6712)

By incorporating above results into the other because it brings the highest benefit. Municipal
sub-model, overall results based on scenario 1 ob- usage has the second priority and the agriculture
tained. The results shown in Table 7 present the sector has the last priority. In Table 7, the solutions
optimized water allocation, water shortage amount, Da  
11 ¼ Da12 ¼ Da13 ¼ 0 Da  
21 ¼ Da21 ¼ Da21 ¼ 0
and related system benefits. In case of insufficient indicate that there will be no shortage of water for
water, the industry use should be first guaranteed municipal and industrial users. However, under low
Optimal Allocation of Water Resources Using a Two-Stage Stochastic Programming Method

Table 8. Overall results of scenario 2

Activity Probability Municipal ði ¼ 1Þ Industry ði ¼ 2Þ Agriculture ði ¼ 3Þ

Daij
Low ðj ¼ 1Þ 0.49 0 0 38.126
Medium ðj ¼ 2Þ 0.31 0 0 10.205
High ðj ¼ 3Þ 0.2 0 0 0
Db ij
Low ðj ¼ 1Þ 0.15 0 0 (0.6, 0.921)
Medium ðj ¼ 2Þ 0.55 0 0 0
High ðj ¼ 3Þ 0.3 0 0 0
Dcij
Low ðj ¼ 1Þ 0.12 0 0 (3.896, 4.217)
Medium ðj ¼ 2Þ 0.11 0 0 (1.346, 1.988)
High ðj ¼ 3Þ 0.77 0 0 0
Dw ij
Low ðj ¼ 1Þ 0.26 0 0 (6.081, 6.123)
Medium ðj ¼ 2Þ 0.33 0 0 (4.044, 4.086)
High ðj ¼ 3Þ 0.41 0 0 3.286
Optimal allocation from wellhead Aa ij
Low ðj ¼ 1Þ 0.49 12 2 135.374
Medium ðj ¼ 2Þ 0.31 12 2 163.294
High ðj ¼ 3Þ 0.2 12 2 173.5
Optimal allocation from spring Ab ij
Low ðj ¼ 1Þ 0.15 0 0 (2.228, 2.549)
Medium ðj ¼ 2Þ 0.55 0 0 3.15
High ðj ¼ 3Þ 0.3 0 0 3.15
Optimal allocation from qanat Ac ij
Low ðj ¼ 1Þ 0.12 0 0 (8.382, 8.703)
Medium ðj ¼ 2Þ 0.11 0 0 10.613, 11.253)
High ðj ¼ 3Þ 0.77 0 0 12.6
Optimal allocation from treated wastewater Aw ij
Low ðj ¼ 1Þ 0.26 1 2.5 (13.876, 13.918)
Medium ðj ¼ 2Þ 0.33 1 2.5 (15.913, 15.955)
High ðj ¼ 3Þ 0.41 1 2.5 16.713
Surplus water in the wellhead for the next period Sa1 j
Low ðj ¼ 1Þ (15.315, 17.814)
Medium ðj ¼ 2Þ (29.017, 31.517)
High ðj ¼ 3Þ (33.75, 38.25)
System benefit and reliability
System net benefit f  (5270477.738, 6865883.395) million Rials
System reliability a (0.3013, 0.6851)

flow level of wellhead with a probability of 49%, system net benefit is corresponding to 0.6712, and
there should be a shortage of 60:646 106 m3 for the the lower bound of the system net benefit is corre-
agricultural sector. Even if the flow level of wellhead sponding to 0.284.
is high, the optimal shortage for the agricultural
sector will be (0, 1.213) 9 106 m3 with probability of
20%. It can be concluded that under high water flow Results of Scenario 2
level, when the available water is over 280 million
cubic meters, the shortages for all consumers will be Table 8 shows the results of solving the model in
zero. Similarly, for other sources, if there are high scenario 2 by considering all parameters and variables
levels of water flow for the qanat (with probability of of model. Results show that with 49% probability for
77%), then there will be no shortages for consumers. wellhead under low flow level, agricultural sector will
In this scenario, the system net benefit is face a shortage of 38.126 million cubic meters, while
(217604.361, 3578184.443) million Rials with relia- the other sectors will not face any shortages. In the
bility of (0.284, 0.6712). The upper bound of the previous scenario (lack of treated wastewater re-
 T. Khosrojerdi et al.

Scenario 1 Scenario 2
38.25
39
36 33.75
33 31.517
29.017 30.876
30
milion cubic meter

27 28.376
24
21 17.814
18 15.314
17.452
15 14.952
12
9 8.61
6 5.687
3
0
lower bound upper bound lower bound upper bound lower bound upper bound
Low level Medium level High level
Figure 3. Comparison of surplus water in the wellhead for the next period for scenarios 1 and 2.

Scenario 1 Scenario 2
6865883.395
70,00,000

60,00,000
5270477.738
50,00,000
Milion Rials

40,00,000
3578184.443
30,00,000

20,00,000 2178604.361

10,00,000

0
Lower bound Upper bound

Figure 4. Comparison of the net benefit resulting from the implementation of scenarios 1 and 2.

source), the optimal shortage was 60.647 million cubic Figure 3 shows that there is more surplus water
meters for the agricultural sector. Moreover, under in wellhead for the next period in scenario 2 than in
medium flow level there is lower shortage than the scenario 1. As shown in Figure 4, the system net
previous scenario, whereas under high flow level there benefit in scenario 2 is significantly higher than in
is no shortage for agricultural sector. As shown in scenario 1. This indicates that establishing a
Table 8, the total water allocated would be 173:5 wastewater treatment station in Zarand can reduce
ðAa33 ¼ 173:5Þ, and optimal allocation target is 173.5
the usage of natural water resources while maxi-
ðTa3 opt ¼ 173:5Þ; thus, there is no shortage for agri-
mizing regional economic benefits.
cultural sector under high flow level. Treated
wastewater satisfies industry and municipal sectorsÕ
demands, while the agricultural sector will face SENSITIVITY ANALYSIS OF WATER LOSS
shortage under different flow levels. There is also RATE
improvements in system net benefit and system reli-
ability ðf  ¼ ð5270477:738; 6865883:395Þ million Today, among the important and critical issues
in water resources management in Kerman province
Rials and a ¼ ð0:3013; 0:6851Þ).
Optimal Allocation of Water Resources Using a Two-Stage Stochastic Programming Method

Table 9. Overall results of a scenario considering 20% water loss rate

Activity Probability users

Municipal ði ¼ 1Þ Industrial ði ¼ 2Þ Agricultural ði ¼ 3Þ

Da
ij
Low ðj ¼ 1Þ 0.49 0 0 29.688
Medium ðj ¼ 2Þ 0.31 0 0 2.525
High ðj ¼ 3Þ 0.2 0 0 0
Db ij
Low ðj ¼ 1Þ 0.15 0 0 (1.14, 1.54)
Medium ðj ¼ 2Þ 0.55 0 0 0
High ðj ¼ 3Þ 0.3 0 0 0
Dcij
Low ðj ¼ 1Þ 0.12 0 0 (3.225, 3.924)
Medium ðj ¼ 2Þ 0.11 0 0 (0.717, 1.914)
High ðj ¼ 3Þ 0.77 0 0 0
Dw ij
Low ðj ¼ 1Þ 0.26 0 0 (3.878, 3.964)
Medium ðj ¼ 2Þ 0.33 0 0 (1.713, 1.799)
High ðj ¼ 3Þ 0.41 0 0 1.482
Optimal allocation from wellhead Aa ij
Low ðj ¼ 1Þ 0.49 12 2 143.812
Medium ðj ¼ 2Þ 0.31 12 2 170.974
High ðj ¼ 3Þ 0.2 12 2 173.5
Optimal allocation from springs Ab ij
Low ðj ¼ 1Þ 0.15 0 0 (2.009, 2.608)
Medium ðj ¼ 2Þ 0.55 0 0 3.15
High ðj ¼ 3Þ 0.3 0 0 3.15
Optimal allocation from qanats Acij
Low ðj ¼ 1Þ 0.12 0 0 (8.675, 9.274)
Medium ðj ¼ 2Þ 0.11 0 0 (10.685, 11.882)
High ðj ¼ 3Þ 0.77 0 0 12.6
Optimal allocation from wastewater Aw ij
Low ðj ¼ 1Þ 0.26 1 2.5 (16.035, 16.121)
Medium ðj ¼ 2Þ 0.33 1 2.5 (18.2, 18.28)
High ðj ¼ 3Þ 0.41 1 2.5 18.517
Surplus water remaining in the wellhead for the next period Sa1j
Low ðj ¼ 1Þ (20.126, 22.626)
Medium ðj ¼ 2Þ (37.530, 40.03)
High ðj ¼ 3Þ (52.5, 57)
System benefit and reliability
System net benefit f  (5957024.861, 7477933.959) million Rials
System reliability a (0.5444, 0.8702)

are the old water pipelines and the lack of scientific especially in net benefit and the surplus water
methods of transforming water. Currently, water remaining in the wellhead for the next period.
loss rate is estimated to be 30%. In Zarand, such
high water loss rate entails high volume of water
extraction from groundwater resources and causes DISCUSSION AND CONCLUSION
reservoir drainage as well as serious problems for
agricultural and industrial economic schemes. Today, issues regarding water resources have
Therefore, as a scenario, the water loss rate is set at become one of the most prominent challenges facing
20% and the results (Table 9) are compared with humanity, and there is a need to find improved ways
scenarios 1 and 2. As shown in Figures 5 and 6, when of making optimal uses of these resources. Water
water loss rate is set at 20%, there are improvements resources management provides a set of methods,
 T. Khosrojerdi et al.

Figure 5. Comparison of the net benefit considering reduction in water loss rate.

20% loss rate 30% loss rate (scenario 2) 30% loss rate (scenario 1)

60 56.4
52.38

50

39.88
milion cubic meters

37.86 38.25
40
31.517 33.75
29.017 30.876
28.376
30
22.54
20.52
17.814
20 15.314 17.452
14.952
8.61
10 5.687

0
Lower bound Upper bound Lower bound Upper bound Lower bound Upper bound
Low level Medium level High level
Figure 6. Comparison of surplus water remaining in wellhead for the next period considering reduction in water loss rate.

tools, and optimization models to prevent reservoir ing the importance of this issue, the aim of this study
shortages, water waste, water shortage, loss of water was based on optimal water resources allocation,
quality and to deliver optimally high-quality water and research results of a case study (Zarand region,
to consumers. One of the main issues in this field is Kerman province) were presented and discussed to
the optimal allocation of water resources to various identify the differences between scientific allocation
consumers, which becomes increasingly important methods and current management methods. In this
every year because water resources are decreasing research, a two-stage stochastic method with interval
while demands are increasing. Therefore, consider- and fuzzy parameters was used. One of the advan-
Optimal Allocation of Water Resources Using a Two-Stage Stochastic Programming Method

tages of this method is that before the identification REFERENCES

of uncertainties relevant to the model parameters
(especially water available from sources), objectives Ahmed, S., King, A. J., & Parija, G. (2003). A multi-stage
are set for various sectors, and the optimization stochastic integer programming approach for capacity
expansion under uncertainty. Journal of Global Optimiza-
model is solved upon identifying uncertainties. tion, 26(1), 3–24.
Then, each sector will be informed of the shortages Birge, J. R., & Louveaux, F. V. (1988). A multicut algorithm for
they will face and the quantity of water they will two-stage stochastic linear programs. European Journal of
Operational Research, 34(3), 384–392.
receive. Therefore, the agricultural consumer will Dogra, P., Sharda, V. N., Ojasvi, P. R., Prasher, S. O., & Patel, R.
know the quantity of water that they will receive in M. (2014). Compromise programming based model for aug-
the future to make suitable investments on farmland menting food production with minimum water allocation in a
watershed: a case study in the Indian Himalayas. Water Re-
while knowing when to compensate water shortages source Management, 28(15), 5247–5265.
from other sources or to limit plantation and culti- Dong, C., Huang, G., Cheng, G., & Zhao, S. (2018). Water re-
vation. Similarly, if the industry sector is informed of sources and farmland management in the Songhua River
watershed under interval and fuzzy uncertainties. Water Re-
future water shortages, it will reduce its production sources Management, 32(13), 4177–4200.
or plan to purchase water from the more expensive Fan, Y. R., Huang, G. H., Guo, P., & Yang, A. L. (2011). Inexact
sources. However, in current management condi- two-stage stochastic partial programming: application to
water resources management under uncertainty. Stochastic
tions, consumers are unaware of their received water Environmental Research and Risk Assessment, 26, 281–293.
quantity and cannot design economic schemes. Un- Fan, Y., Huang, G., Huang, K., & Baetz, B. W. (2015). Planning
der these conditions, facing scarcities will entail water resources allocation under multiple uncertainties
through a generalized fuzzy two-stage stochastic program-
crisis for various sectors especially municipal, agri- ming method. IEEE Transactions on Fuzzy Systems, 23(5),
cultural, and industrial consumers. Therefore, there 1488–1504.
is a need for scientific methods to become involved Ferrero, R. W., Rivera, J. F., & Shahidehpour, S. M. (1998). A
dynamic programming two-stage algorithm for long-term
in policy making and management to maximize the hydrothermal scheduling of multireservoir systems. IEEE
economic benefits of societies while preventing Transactions on Power Systems, 13(4), 1534–1540.
consumers from facing the crisis. Fu, Q., Zhao, K., Liu, D., Jiang, Q., Li, T., & Zhu, C. (2016). Two-
stage interval-parameter stochastic programming model
The model presented in this research can help based on adaptive water resource management. Water Re-
regional water managers to provide equity for dis- sources Management, 30(6), 2097–2109.
tributing water among different consumers with vari- Guo, P., Huang, G. H., Zhu, H., & Wang, X. L. (2010). A two-
stage programming approach for water resources manage-
ous limitations. Besides, the results of this study ment under randomness and fuzziness. Environmental Mod-
indicate a substantial improvement in system profits elling and Software, 25(12), 1573–1581.
and water resources conservation and maintenance. Hu, Z., & Hu, G. (2018). A multi-stage stochastic programming
for lot-sizing and scheduling under demand uncertainty.
Thus, it can be concluded that regional water managers Computers & Industrial Engineering, 119, 157–166.
have this opportunity to develop their regions from Huang, G. H., Baetz, B. W., & Party, G. G. (1993). A grey fuzzy
three perspectives including social, environmental, and linear programming approach for municipal solid waste
management planning under uncertainty. Civil Engineering
economic. For the future, this model can be applied to Systems, 10(2), 123–146.
other case studies, and the economic results can be Huang, G. H., Baetz, B. W., & Patry, G. G. (1995). Grey integer
compared with the results of this study. For extending programming: an application to waste management planning
under uncertainty. European Journal of Operational Re-
the application of this model in other regions and search, 83(3), 594–620.
making it more comfortable for the users, a decision Huang, G. H., & Loucks, D. P. (2000). An inexact two-stage
support system can be designed by providing a graph- stochastic programming model for water resources manage-
ment under uncertainty. Civil Engineering and Environmen-
ical user interface that can be developed by programs tal Systems, 17, 95–118.
such as MATLAB, C#, and Java. Moreover, models Ji, L., Sun, P., Ma, Q., Jiang, N., Huang, G. H., & Xie, Y. L.
with better accuracy can be presented by considering (2017). Inexact two-stage stochastic programming for water
resources allocation under considering demand uncertainties
costs for wastewater refinement maintenance. and response—A case study of Tianjin, China. Water, 9(6),
414.
Li, Y. P., & Huang, G. H. (2008). IFMP: Interval-fuzzy multistage
programming for water resources management under
uncertainty. Resources, Conservation and Recycling, 52, 800–
ACKNOWLEDGMENTS 812.
Li, W., Wang, B., Xie, Y. L., Huang, G. H., & Liu, L. (2015). An
inexact mixed risk-aversion two-stage stochastic program-
This research was supported by Kerman Re- ming model for water resources management under uncer-
gional Water Company under Grant 97/02. tainty. Environmental Science and Pollution Research, 22(4),
2964–2975.
 T. Khosrojerdi et al.

Liu, D., Liu, W., Fu, Q., Zhang, Y., Li, T., Imran, K. M., et al. zation of parameter uncertainties and their interactions: an
(2017). Two-stage multi-water sources allocation model in application to water resources management. European
regional water resources management under uncertainty. Journal of Operational Research, 240(2), 572–581.
Water Resources Management, 31(11), 3607–3625. Wang, S., Huang, G. H., & Zhou, Y. (2016). A fractional-factorial
Loucks, D. P., Stedinger, J. R., & Haith, D. A. (1981). Water probabilistic-possibilistic optimization framework for plan-
resource systems planning and analysis. Englewood Cliffs: ning water resources management systems with multi-level
Prentice-Hall. parametric interactions. Journal of Environmental Manage-
Maqsood, I., Huang, G. H., & Scott Yeomans, J. (2005). An ment, 172, 97–106.
interval-parameter fuzzy two-stage stochastic program for Watkins, D. W., McKinney, D. C., Lasdon, L. S., Nielsen, S. S., &
water resources management under uncertainty. European Martin, Q. W. (2000). A scenario-based stochastic program-
Journal of Operational Research, 167, 208–225. ming model for water supplies from the highland lakes. In-
Parisio, A., & Jones, C. N. (2015). A two-stage stochastic pro- ternational Transactions in Operational Research, 7(3), 211–
gramming approach to employee scheduling in retail outlets 230.
with uncertain demand. Omega, 53, 97–103. Xie, Y. L., Huang, G. H., Li, W., Li, J. B., & Li, Y. F. (2013). An
Pereira, M. V. F., & Pinto, L. M. V. G. (1985). Stochastic opti- inexact two-stage stochastic programming model for water
mization of a multireservoir hydroelectric system: a decom- resources management in Nansihu Lake Basin, China. Jour-
position approach. Water Resources Research, 21(6), 779–792. nal of Environmental Management, 127, 188–205.
Pereira, M. V., & Pinto, L. M. (1991). Multi-stage stochastic Xie, Y. L., Xia, D. H., Huang, G. H., & Ji, L. (2018). Inexact
optimization applied to energy planning. Mathematical Pro- stochastic optimization model for industrial water resources
gramming, 52(1–3), 359–375. allocation under considering pollution charges and revenue-
Seifi, A., & Hipel, K. W. (2001). Interior-point method for risk control. Journal of Cleaner Production, 203, 109–124.
reservoir operation with stochastic inflows. Journal of Water Yin, L., & Han, L. (2015). Risk management for international
Resources Planning and Management, 127(1), 48–57. portfolios with basket options: A multi-stage stochastic pro-
Shao, L. G., Qin, X. S., & Xu, Y. (2011). A conditional value-at- gramming approach. Journal of Systems Science and Com-
risk based inexact water allocation model. Water Resources plexity, 28(6), 1279–1306.
Management, 25, 2125–2145. Zahiri, B., Torabi, S. A., Mohammadi, M., & Aghabegloo, M.
Shiklomanov, I.A. (1998). In World water resources: A new ap- (2018). A multi-stage stochastic programming approach for
praisal and assessment for the 21st century: A summary of the blood supply chain planning. Computers & Industrial Engi-
monograph World water resources. UNESCO. neering, 122, 1–14.
Shukla, S., & Gedam, S. (2018). Evaluating hydrological re- Zeng, X. T., Li, Y. P., Huang, G. H., & Liu, J. (2017). Modeling of
sponses to urbanization in a tropical river basin: A water water resources allocation and water quality management for
resources management perspective. Natural Resources Re- supporting regional sustainability under uncertainty in an
search. https://doi.org/10.1007/s11053-018-9390-7. arid region. Water Resources Management, 31(12), 3699–
Sophocleous, M. (2004). Global and regional water availability 3721.
and demand: Prospects for the future. Natural Resources Zhang, L., & Li, C. Y. (2014). An inexact two-stage water re-
Research, 13(2), 61–75. sources allocation model for sustainable development and
Tajeddini, M. A., Rahimi-Kian, A., & Soroudi, A. (2014). Risk management under uncertainty. Water Resources Manage-
averse optimal operation of a virtual power plant using two ment, 28(10), 3161–3178.
stage stochastic programming. Energy, 73, 958–967. Zhou, Y., Huang, G., Wang, S., Zhai, Y., & Xin, X. (2016). Water
Wang, D., & Adams, B. J. (1986). Optimization of real-time resources management under dual uncertainties: a factorial
reservoir operations with markov decision processes. Water fuzzy two-stage stochastic programming approach. Stochastic
Resources Research, 22(3), 345–352. Environmental Research and Risk Assessment, 30(3), 795–
Wang, S., & Huang, G. H. (2013). An interval-parameter two- 811.
stage stochastic fuzzy program with type-2 membership Zimmermann, H. J. (1985). Fuzzy set theory and its applications.,
functions: an application to water resources management. International Series in Management Science. Operations
Stochastic Environmental Research and Risk Assessment, Research. Kluwer Boston: Nijhoff Publishing.
27(6), 1493–1506.
Wang, S., & Huang, G. H. (2015). A multi-level Taguchi-factorial
two-stage stochastic programming approach for characteri-