Interbasin water transfer: economic water quality-based model

Interbasin Water Transfer: Economic WaterQuality-Based Model

Mohammad Karamouz, F.ASCE1; S. Ali Mojahedi2; and Azadeh Ahmadi3

Abstract: The interbasin water transfer project is an alternative to balance the nonuniform temporal and spatial distribution of waterresources and water demands, especially in arid and semi arid regions. A water transfer project can be executed if it is environmentallyand economically justified. In this study, the feasibility of two interbasin water transfer projects from Karoon River in the western part ofIran to the central part of the country is investigated. An optimization model with an economic objective function to maximize the netbenefit of the interbasin water transfer projects is developed. The planning horizon of the model is 23 years �the length of historical data�;and it is solved using genetic algorithm. In order to consider environmental impacts of water transfer projects, a water quality simulationmodel has been used. Then, an Artificial Neural Network model is trained based on the simulation results of a river water quality modelin order to be coupled with the optimization model. The outputs of the optimization model are the value of economic gain of the sending�Karoon� basin to offset the loss of agricultural income and environmental costs. The optimal polices for water transfer during theplanning horizon has been generated using the coupled simulation-optimization model. Then, operating rules are developed using a KNearest Neighborhood model for the real time water transfer operation. The results show the significant value of using the proposedalgorithm and economic evaluation for water transfer projects.

DOI: 10.1061/�ASCE�IR.1943-4774.0000140

CE Database subject headings: Water transfer; Water quality; Simulation; Algorithms; Economic factors; Neural networks; Aridlands; Water demand.

Author keywords: Interbasin water transfer; Water quality simulation; Genetic algorithm; Economic assessment; Artificial neuralnetwork; K Nearest Neighborhood.

Introduction

Iran is located in an arid and semiarid region of the world. It hasa nonuniform temporal and spatial distribution of water resourcesand water demands. There is enough water in some basins whilein some other basins there is water scarcity. The periodic droughtsand water deficits cause the migration of the habitants in certainregions. In addition, considering the rate of the increasing popu-lation and the improving economy, it is necessary to have long-term planning to balance the supply and the demands distribution.

The interbasin water transfer project is an alternative to bal-ance the nonuniform temporal and spatial distribution of waterresources and water demands. Transferring water from an areamay cause a variety of negative impacts, social and environmen-tal impacts. But a water transfer project can be executed if it is

environmentally and economically justified. When water is in-tended to be used in another basin, water rights could be tradedfor financial resources. In the national arena, water is equity forall. Equity for those who are in need of water and do not haveaccess to water and those who actually have the water rights andmay have a surplus that is wasted in a variety of ways. To analyzethe above issues, tangible and nontangible costs and benefitsshould be evaluated.

Several investigators have emphasized the need for economicand environmental assessment of interbasin water transfer plans.Lund and Israel �1995� presented the application of multi stagelinear programming for estimation of the least-cost integration ofseveral water marketing opportunities with water conservationand traditional water supplies. Draper et al. �2003� developed aneconomic based optimization model for California’s major watersupply system. They noted that optimization models driven byeconomic objective functions are practical for assessing the de-velopment project. Feng et al. �2007� developed a decision sup-port system �DSS� for assessing the social-economic impact ofChina’s South-to-North Water Transfer project. The DSS providesdecision support through simulation with an embedded watercomputable general equilibrium model. Gupta and Zaag �2008�have assessed the interbasin water transfers from a multidisci-plinary perspective, and attempted to answer whether such trans-fers are compatible with the concept of integrated water resourcesmanagement and the criteria for assessing such transfers. Mateteand Hassan �2005� developed an analytical framework that can beapplied to integrate environmental sustainability aspects into eco-nomic development planning in the case of exploiting water re-sources through interbasin water transfers.

1Professor, School of Civil Engineering, Univ. of Tehran, Tehran,Iran; and Research Professor, Polytechnic Institute of NYU, Brooklyn,NY 11201 �corresponding author�. E-mail: [email protected], [email protected]

2Staff Engineer, Water and Wastewater Planning Bureau, Ministry ofEnergy, Tehran, Iran. E-mail: [email protected]

3Assistant Professor, Dept. of Civil Engineering, Isfahan Univ. ofTechnology, Isfahan, Iran; formerly, School of Civil Engineering, Univ.of Tehran, Tehran, Iran. E-mail: [email protected]

Note. This manuscript was submitted on June 30, 2008; approved onJuly 8, 2009; published online on July 17, 2009. Discussion period openuntil July 1, 2010; separate discussions must be submitted for individualpapers. This paper is part of the Journal of Irrigation and DrainageEngineering, Vol. 136, No. 2, February 1, 2010. ©ASCE, ISSN 0733-9437/2010/2-90–98/$25.00.

90 / JOURNAL OF IRRIGATION AND DRAINAGE ENGINEERING © ASCE / FEBRUARY 2010

In this study, the proposed optimization model with consider-able computational complexity due to a high number of decisionvariables and nonlinear behavior of objectives and constraints aresolved with the use of a genetic algorithm �GA� method. In thelast decade, more attention has been given to soft computing tech-niques, such as evolutionary algorithms �EA� and in general par-ticularly GA in particular. Burn and Yulianti �2001� have shownthe capabilities of GAs for identifying solutions to classicalwaste-load allocation problems. They showed that GAs provide arather robust and noninferior solution for deterministic waste loadallocation in low flow conditions. Cai et al. �2001� combined GAswith linear programming approaches to solve a set of complicat-ing constraints. The results show that the GA is capable of findingquality solutions to the problems at a reasonable run time. Kera-chian and Karamouz �2006� used an algorithm combining a waterquality simulation model and a stochastic conflict resolution GA-based optimization technique for determining optimal reservoiroperation rules.

In this paper, an economic and environmental evaluation ofwater basin transfer projects is developed. Two water transferprojects in the central part of Iran, Solegan to Rafsanjan �Case A�and Koohrang III to Zayandeh-Rud �Case B� are considered asthe case studies. These two interbasin water transfer projects havethe same basin �origin� as the source of providing water. An eco-nomic model is developed to optimize the benefit for evaluatingthe quantity of the water to be transferred and the variation inquality of the remaining water. The GA method is used to solvethe optimization model for determining the flow rate to be trans-ferred in each month. The constraints include the system capacity�tunnels�, average inflow to the diversion reservoirs, continuityequation, and water allocation to the monthly demands.

The paper is organized as follows: the methodology is pre-sented in “Methodology” followed by case study characteristicsin “Case Studies.” The water quality simulation model and itsresults are presented in “Simulation Model of the Karoon River�Sending Basin�.” The optimization model formulation of inter-basin water transfer is given in the “Structure of the OptimizationModel” section. The results of optimization models and devel-oped operating rules are presented in section 6. Finally, a “Sum-mary and Conclusion” is given.

Methodology

In this study, an optimization model with an economic objectivefunction for water basin transfer projects has been developed andthe environmental impacts of water transfer from the river head-water are considered. The proposed model of water transfer isbased on attaining the maximum benefits with the minimum cost.Therefore, costs and benefits for each basin �sending basin andreceiving basins� are estimated. The benefits include an increasein the agricultural production, decrease in pumping costs �receiv-ing basin� and increase of water release in the receiving basin.The costs include an increase in the dredging cost and a decreasein hydropower energy generation. Most importantly, the decreasein the agricultural production in the sending basin, capital invest-ment and the operation and maintenance �OM� costs of watertransfer projects implementation as well as the treatment costs tomaintain the water quality standards in the Karoon River �sendingbasin�. The water quality variation is determined through acoupled Artificial Neural Network �ANN� model in the optimiza-tion model. The ANN model is developed using the results of thewater quality simulation model for the Karoon River. The Qual2k

software developed by the U.S. Environment Protection Agency�U.S. EPA� is used to simulate the water quality on a monthlytime scale. The GA-based optimization model determines themonthly water allocation to the receiving basins in each monthconsidering the benefit and cost analysis. Fig. 1 shows the pro-posed algorithm of optimization model for interbasin water trans-fer project. By using the results of the optimization model,operating rules are generated using a K Nearest-Neighborhood�KNN� model. The operating rules are used to develop a workingoperational scheme for real time operation using a KNN model.

Case Studies

Fig. 2 shows the location of sending and receiving basins.Namely, one sending basin in Khuzestan �in the western part ofthe country� with two receiving basins including Rafsanjan plainin Kerman and Zayandeh-Rud River basin in Isfahan �in the cen-tral part of the country� are considered as case studies.

One of the water transfer projects is from Solegan in theKaroon River to the Rafsanjan plain �here is called receivingBasin 1�. Rafsanjan is located in the central part of Iran and hasan area of 12,421 km2 located between 54°, 52� and 56°, 34�longitudes and 29°, 51� and 31°, 31� latitudes. This region isclassified as an arid area. Rafsanjan has hot summers and drywinters. Average annual rainfall is about 90 mm. The major ob-jective of this water transfer project is to supply water demand tothe Rafsanjan agriculture plain for production of pistachio, anexclusive and expensive product �over $7/kg in a local marketfor dried pistachio�. The Water Transfer project from Solegan toRafsanjan is designed for supplying an average of 250 MCM peryear.

Start

Generating the initial populationConsists of two discharges in each month (Solegan & KohrangIII)

Constraints of the model Consist of:1- Physical constraint (tunnel)

2- Continuity equation in transferring reservoir3- Demand supply

Determination of the fitness function of each chromosomeObjective function: Maximize benefits minus costs

Benefits:� Increase of agricultural production (receiving basin-Solegan project)� Decrease in pumping cost (receiving basin-Solegan project)� Benefit of water release of Zayandeh-Rud Dam (receiving basin-

KohrangIII project)Costs:

� Decrease in agricultural production (sending basin)� Decrease in hydropower energy generation (sending basin)� Increase of dredge cost (sending basin)� The capital and operation and maintenance costs of the water transfer

project (sending river)� The treatment cost of the water quality variable with respect to water

quality standards (sending river)

Use water qualitysimulation andthen train an

Artificial NeuralNetwork Model

Comparing the results of the objectivefunction in this iteration with the last

iteration, if it is better

Store and Replace the last result

Apply Genetic algorithm

Selection (Tournament)Crossover

Coding of chromosomeMutation

Encoding of chromosome

Is this the finaliteration?

Store the result of the model

End

No

Yes

Yes

No

Fig. 1. Flowchart of the GA optimization model for interbasin watertransfer projects

JOURNAL OF IRRIGATION AND DRAINAGE ENGINEERING © ASCE / FEBRUARY 2010 / 91

Second water transfer project is the water transfer project fromKoohrang III Tunnel to the Zayandeh-Rud Reservoir �here iscalled receiving Basin 2�. The objective of this water transferproject is supplying water demand to the Zayandeh-Rud Reser-voir. The reservoir inflow includes natural river inflow and inflowfrom the first and the second Koohrang Tunnels, with an annualaverage of 1,600 MCM. The inflow to the reservoir will increasewith the construction of the third Koohrang Tunnel called Koo-hrang III in this study. This water transfer project is designed fortransferring an average of 250 MCM per year �Karamouz et al.2007�.

Simulation Model of the Karoon River„Sending Basin…

Lower part of the Karoon River, the largest river in Iran, withmore than 450 km of length is subject to major impacts fromwater transfer projects. This part of the river supplies the waterdemands of 16 cities, major industrial and agroindustrial estab-lishments �water demand of about 1 billion m3�, and about700,000 ha �1 ha=10,000 m2� of agricultural lands. In the lowerpart, Karoon River water pollution due to increasing water with-drawal and wastewater discharge to this river has already endan-gered the aquatic life of the river. The affected study area �fromGotvand Dam to Darkhoein Station� and different components ofwater resources including the pollution sources are shown in Fig.3 �Karamouz 2004�. This part of the river is considered to evalu-ate how the water transfer projects are altering the river waterquality.

Water Quality Simulation Model of the Karoon River

The basic equation of water quality simulation models developedin this study is based on a one-dimensional advection-dispersionmass transport equation, which is numerically integrated overspace and time for each water quality constituent, using Qual2Ksoftware, developed by the U.S. EPA �2004�.

Based on the available data from the existing monitoring sys-tems, total dissolve solid �TDS� or electric conductivity �EC�,biochemical oxygen demand �BOD�, and dissolve oxygen �DO�are the selected water quality indices. Using the available data for12 years �1992–2003� from the existing monitoring system, the

monthly water/wastewater quality and quantity data of the firstten years was used for calibration and the past two years forvalidation.

The results of the simulation model show that BOD and DOconcentrations do not violate the water quality standards along theriver. This is because of the carrying capacity of the river forhandling BOD. TDS concentration increases along the river dueto domestic, agricultural and industrial wastewater discharges.Therefore, only TDS variation along the river is simulated toincorporate in the optimization model. The Qual2k simulationmodel is used taking into account the available data for 33 yearsalong the river for generating more data to train and test the ANNmodel.

Estimation of EC „TDS… Variable Using the ANN Model

In this study, an ANN model is developed to estimate TDS vari-able at the water quality control point �Ahvaz City�. Since theinput-output relation is easy to formulate in the neural networkmodels, it can be considered as a simple expression of input-output computations in any descriptive or optimization models.

The ANN model has been used for the examination of differ-ent types of multilayer perception �MLP� architectures. The in-puts are the index of month �shows the physical and climateconditions�, discharge, and TDS in headwater �Gotvand Dam�.From 396 produced data points, 300 data points �about 75% ofavailable data� have been used for training and the remaining dataare used for the testing of the ANN model. Eq. �1� estimates TDSvariable at Ahvaz City using the ANN model

TDS2 = Purline�w2 � �tan sig�w1 � �m,Q,TDS1�� + b1� + b2�

�1�

where m=index of the month; Q=discharge, TDS1 and TDS2are water quality characteristics in headwater and control pointin downstream, respectively. The best ANN model, a MLP, in-cludes three layers with 3, 5, and 1 neurons respectively, andminimum error in the testing period �RMSE=0.10899�. The char-acteristics of the ANN model including weights and biases ineach layer �w1, w2, b1, and b2� are presented in Table 1. ThisANN model is placed in the optimization model to simulate theTDS variation.

RafsanjanDaranjir

Solegan

Persian Gulf

Koohrang Karoon

ZayandehroodDam

Oman sea

Persian Gulf

Caspian Sea

Fig. 2. Iran’s major watersheds and the close-up of the three basins, the sending basin, and two receiving basins


Structure of the Optimization Model

The main objective of the proposed model is to maximize thedifference between the associated benefit and cost

Maximize Z = ��k=1

3

�t=1

12

�y=1

23

Benefitk,t,y − �m=1

5

�t=1

12

�y=1

23

Costm,t,y�2�

Subject to

Benefit1,t,y = �t

�y

�V2t,y � CPDr � Pcr�

t = 1, . . . ,12, y = 1, . . . ,T �3�

Benefit2,t,y = �t

�y �V2t,y � Hb � Hr�

� � 0.102� Pp�

t = 1, . . . ,12, y = 1, . . . ,T �4�

Fig. 3. Different components of the water resources including the pollution sources system in the sending region, Khuzestan Province �Karamouz2004�


Benefit3,t,y = �t

�y

�F�R2t,y�-F�R1t,y�� y = 1, . . . ,T �5�

F�R� = �− 1.88 � R2 + 2603.5 � R

− 0.759 � R2 + 419.5 � R t = 10,11,12,1,2,3

t = 4, . . . ,9

�6�

R2t,y = It,y + I3t,y + St,y − St+1,y �7�

R1t,y = It,y + St,y − St+1,y �8�

Cost1,t,y = �t

�y

��V1t,y + V2t,y� � CPDk � Pck�

t = 1, . . . ,12, y = 1, . . . ,T �9�

Cost2,t,y = �t

�y

��G�Q − �X1t,y + X2t,y�� − G�Q��

t = 1, . . . ,12, y = 1, . . . ,T �10�

G�Q� = − 7 � 10−4 � Q2 − 1.355 � Q + 745.9 �11�

Cost3,t,y = �t

�t

�L�Q� − L�Q − �X1t,y + X2t,y��

t = 1, . . . ,12, y = 1, . . . ,T �12�

L�X� = 0.0788 � X2 − 20.703 � X + 4,000 �13�

Cost4,t,y = �t

�y

�Phct,y� t = 1, . . . ,12, y = 1, . . . ,T �14�

Cost5,t,y = � ��Cpt,y − Cst,y� � Pt if Cpt,y � Cst,y

0 otherwise�

t = 1, . . . ,12, y = 1, . . . ,T �15�

Cpt,y = ANN�Q,C,X1t,y,X2t,y� �16�

V1t,y = It,y + St,y − St+1,y t = 1, . . . ,12, y = 1, . . . ,10 �17�

V2t,y = Ist,y + Sst,y − Sst+1,y t = 1, . . . ,12, y = 1, . . . ,10 �18�

X1 mint,y � DKt,y � X1t,y � X1 maxt,y � DKt,y

t = 1, . . . ,12, y = 1, . . . ,10 �19�

X2 mint,y � DSt,y � X2t,y � X2 maxt,y � DSt,y

t = 1, . . . ,12, y = 1, . . . ,1 �20�

where t=index of the month; y=index of the year; T=number ofoptimization years; m=index of the cost; k=index of the benefit;i=interest rate; Benefit1,t,y =benefit of the agriculture productionsin month t of year y; Benefit2,t,y =benefit of the decreasing pump-ing costs in Rafsanjan plain in month t of year y; Benefit3,t,y

=benefit of the water releasing of the Zayandeh-Rud Dam inmonth t of year y; Cost1,t,y =cost of the decreasing agricultureproduction in the Karoon basin in month t of year y; Cost2,t,y

=cost of the increasing dredge actives in month t of year y;Cost3,t,y =cost of the hydropower generation reduction in month tof year y; Cost4,t,y =capital and OM costs in month t of year y;Cost5,t,y =cost of the water treatment in critical points in month tof year y; V1t,y ,V2t,y =optimal transferred water volume fromKoohrang III to Zayandeh-Rud and from Solegan to Rafsanjan inmonth t of year y, respectively �MCM�; X1t,y ,X2t,y =optimaltransferred discharge from Koohrang III to Zayandeh-Rud andfrom Solegan to Rafsanjan in month t of year y, respectively�cms�; CPDr ,CPDk=crop per drop in Rafsanjan and plains�Kg /m3�; Pcr ,Pck=average price of crops in Rafsanjan andKhuzestan plains �$/Kg�; Hb=depth of water table in RafsanjanPlain �m�; Hr=total duration of pumping from groundwater in themonth �hr�; Pp=price of electricity needed for water pumping�$/KWh�; �=pumping efficiency �%�; F�R�=benefit function ofwater released from Zayandeh-Rud Dam in month t of year y;R1t,y ,R2t,y =release function from Zayandeh-Rud Dam with andwithout considering inflow from Koohrang III Tunnel in montht of year y, respectively; It,y =inflow to Zayandeh-Rud Reservoirin month t of year y �MCM�; I3t,y , Ist,y =inflow to Koohrang IIIand Solegan Dams in month t of year y, respectively �MCM�;St,y ,Sst,y =storages of Zayandeh-Rud and Solegan Dams at thebeginning of month t of year y �MCM�; Q=headwater dischargein Karoon River �cms�; G�Q�=cost function of dredging; L�X�=benefit function of the hydropower energy generation; Phct,y

=cost of instruction and maintenance interbasin water transferprojects in month t of year y; Cst,y =standard values of waterquality indices in the river in month t of year y; Cpt,y =values ofwater quality indices in the control point in month t of year y;ANN=simulation model of water quality indices in control points�ANN model�; Pt=water treatment cost due to water quality ex-ceeded the standard value; C=values of water quality indices inthe headwater; X1 mint,y ,X1 maxt,y =minimum and maximumtransferred discharge from Koohrang III in month t of year y,respectively �CMS�; X2 mint,y ,X2 maxt,y =minimum transferreddischarge from Solegan in month t of year y respectively �CMS�;and DKt,y ,DSt,y =percentage of monthly water allocation fromKoohrang III and Solegan considering monthly demand in montht of year y, respectively. Eqs. �3�–�16� are classified and explained

Table 1. Characteristics of Developed ANN Model

w1 w2 b1 b2

Input 1 Input 2 Input 3 Output

0.0527 �2.862 0.4839 1.1844 �2.61 0.2265

�1.2519 1.7258 �0.4438 0.1956 1.3199

2.5331 0.3893 0.6885 0.2642 �0.1005

2.0273 �0.3184 0.4429 �0.1923 1.4914

�0.1577 0.5079 �2.3933 �0.358 �2.1082


in the cost-benefit analysis based for the receiving basins �1 and2� and sending basin as follows.

Solegan to Rafsanjan Water Transfer Planin the Rafsanjan Region „Receiving Basin 1…

Eq. �3�: The benefit gained from water transfer to Rafsanjan isdue to increased agricultural products obtained from the crop perdrop coefficient �CPD� multiplied by transferred volume of waterand price of products.

Eq. �4�: the benefit from the reduction in pumping cost multi-plication of transferred volume of water, pumping duration�hours�, and average depth of ground water level and division bypumping efficiency. Eq. �14�: the total the annual cost of Soleganto Rafsanjan water transfer project that includes initial invest-ment, OM costs during the operating period. The annual andmonthly costs of the project for a period of 30 years are $30.3 and$2.23 million considering 10% interest rate and the initial invest-ment of $285.4 million.

Koohrang III to Zayandeh-Rud Water Transfer Planin Zayandeh-Rud Basin „Receiving Basin 2…

Eqs. �5�–�8�: The benefit of water transfer to Zayandeh-Rud Damobtained from the difference between water values before andafter the project. Since Zayandeh-Rud basin is affected by severdrought during 1990s, the benefit of water transferring is esti-mated based on drought damages reduction. The benefit-releasefunctions are obtained using Eq. �6� for the first �fall and winter�and the second �spring and summer� seasons, respectively�Araghinejad 2005�.

The annual and monthly costs of the project for a period of 30years are $10.8 and $0.903 million considering 10% interest rateand the initial investment of $102.5 million.

Karoon Basin „Sending Basin…

Eqs. �12� and �13�: the cost of decreasing in hydropower genera-tion in dams located along the Karoon River. These values areestimated based on the hydropower generation function �Rabeiet al. 2004� through the production of less hydropower generationdue to water transferring based on the price of electricity.

Eqs. �10� and �11�: the dredging cost is estimated for beforeand after the transfer project using the proposed cost function byZahiri and Korestani �2004�. The base flow at the control point�Ahvaz City� is considered as 250 cm �cubic meter per second�.Eq. �9� presents the reduction in the benefit of agricultural pro-duction considering of the amount of water transferred from thesending basin �Karoon basin�.

Eqs. �15� and �16�: the removal cost of the excess totaldissolved solids �TDS� from associated standard value after

water transfer. The TDS variation in control point is estimatedusing the ANN model. The standard value for TDS is consideredas 1,200 mg/lit. The removal of TDS is carried out using evapo-ration ponds that is suitable for the climate condition in the studyarea.

Other constraints of water transfer model are as follows:Eqs. �17� and �18�: the continuity equations of Koohrang III’sand Solegan Dams considering their minimum and maximumstorage volumes and river discharge during the planning horizon.Eqs. �19� and �20�: the allocation range of transferred discharge ina given month is calculated by multiplying the monthly demandpercentage during the planning horizon by the minimum andmaximum figures of the transferred water.

The proposed model is solved using the GA method. In theGA setting, the structure of decision variables as genes of a chro-mosome along the Karoon River is shown in Fig. 4. In this study,there are two decision variables in each month, which are thetransfer flows from Koohrang III and Solegan Tunnels. The opti-mization period is 23 years �1981–2003�; therefore, each chromo-some has 23�12�2=552 genes. After trial and error, the bestparameters of the GA model, crossover and mutation probabilitiesare considered 0.8, 0.01, respectively, with 100 chromosomes ineach generation.

Results

The results of the optimization model are presented in quantityand quality issues in following sections. The effects of the Sole-gan water transfer project in restoration of the Rafsanjan aquiferare evaluated. Finally, water transfer operating rules for real timeoperation are developed using the KNN model.

24..1

Number of genes in each year

Length of time of optimization model (552 genes)

Transferring discharge in firstmonth (Solegan)

Transferring discharge in firstmonth (Koohrang)

Genes in each month

Fig. 4. Structure of the GA chromosome used in optimization model

%###

$(##

$###

(##

#

(##

$###

# %## '## )## +## $### $%## $'##� � � � � � � �

��

��

� �

� �

�

��

�

�

�

�

�

�

��

�

�

�

�

�

�

�

�

��

%##

$(#

$##

(#

#

(#

$##

# %## '## )## +## $### $%## $'## teration

Objectivefunctionvalue(MillionDollars)

Fig. 5. Values of objective function of optimization model in cascadeiteration


Water Quantity Assessments

The optimal monthly discharge is determined using the optimiza-tion model subject to the model constraints. Fig. 5 shows thevariation of fitness value in different iterations of the GA model.As it can be seen, the fitness value has an increasing slope, up to600 iterates, and after that the objective function levels off. It hassome rapid jump until the 300th iterations and after that it con-verges toward a minimum of $57.4 million.

Fig. 6 shows the optimal monthly discharges of transferredwater from the Koohrang III and Solegan Tunnels in 276 months�23 years�. These figures show that in the last three years ofoptimization period, the water demands are not supplied com-pletely due to hydrological drought in the case study area.

Fig. 7 shows the annual benefits and costs of the water transferprojects for different components of objective function and Table2 shows percentage of these values during the planning horizon.The maximum benefit is related to agricultural production in theRafsanjan plain �receiving Basin 1� which is due to the high valueof crop products. The minimum benefit is related to the decreasein the pumping cost. The maximum costs are due to decreasing

agricultural production in the Khuzestan plain �sending basin� andenvironmental costs resulted from changing water quality in theKaroon River. As shown in Table 2, the present value of the netbenefit over 23 year time horizon is about 130 million dollars.Therefore, the cost per cubic meter of water is about $0.38.

In order to evaluate the effects of agricultural price and CPDcoefficient on the net benefit, sensitivity analysis has been done.The variations of the objective function versus agricultural priceand CPD variations are shown in Fig. 8. As shown in this figure,if the market price falls below 96% of the selected price in thesending and receiving basins, the transfer projects will not beeconomical. These projects are economically justified if the agri-cultural CPD coefficient �kg of production per cubic meter ofwater use� falls below 0.58. Currently the average productionlevel �CPD� is 0.7 kg /m3 in Iran �0.6 is assumed for the basinsin this paper� and the international norm is above 1 kg /m3.Therefore, the project seems to remain economically feasible ifthe rate goes up and market fluctuations will not affect its fate,significantly.

0

2

4

6

8

10

12

14

16

18

0 50 100 150 200 250 Month

watertransfering(cms)

Koohrang Solegan

Fig. 6. Monthly optimal discharges during the optimization period for Koohrang III and Solegan projects

0

20

40

60

80

100

120

140

160

180

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23

Year

Benefit&Cost(MillionDollars)

Agricultural Benefit (Rafsanjan region) Pumping Benefit

Total benefit in zayanderoud basin Agricultural Cost (Khuzestan region)Dredge Cost Hydropower Cost

Treatment Cost Physical cost

Fig. 7. Values of yearly benefits and costs in the optimization period of two projects


Evaluation of Water Quality Results

For assessing the effects of the transfer project on the water qual-ity of the Karoon River, the monthly simulation model is used.This model analyzed TDS, DO, and BOD considering maximumwithdrawal of transferred water from the river. The results showthat water quality remains in an acceptable range �TDS�1,200 mg / lit, BOD�10 mg / lit� and the water transfer impactson water quality variation are not significant. However the TDSvariation could be affected by the future conditions of the riverbranches in upstream and the wastewater releases along the river.Therefore, the water treatment cost can be taken into accountwhen the river water quality is critical. The comparison of waterquality conditions both before and after the transfer projectsimplementation shows the concentration of TDS increases by 0.7mg/lit per unit volume �1 m3� at the control point in Ahvaz.

Impact on Groundwater of Receiving Basin 1

Aquifer of the receiving Basin 1 is the source of water supply fordifferent domestic, agricultural �97% of water demands� and in-dustrial sectors. Due to over withdrawal of water from this aqui-fer, it has been classified as unauthorized aquifers for withdrawaland development. There are 1,381 wells �1,308-deep and 73-semideep� and 153 aqueducts in the region. The storage coeffi-cient of aquifer is estimated as 0.05 and the Thiessen area of theaquifer is estimated as 4 ,107.91 km2. The level of water subsid-ence is calculated using Eq. �21� as follows:

�h =�V

S � A�21�

where A=Thiessen area; �h=drawdown of groundwater level;�V=change in aquifer volume; and S=storage coefficient.

The results show the water table depletion will be negligibleduring the planning horizon �23 years� in comparison with about11 m in the last 14 years. In the sending basin, the effect of water

transfer on groundwater resources is negligible because in thisregion the surface water is used primarily and the aquifers areshallow and not suitable for development.

Developing Water Transfer Operating Rules

In order to develop the operating rules for water transfer, a KNNmodel is used. The KNN model is a nonparametric estimation ofprobability densities and regression functions through weightedlocal average of the dependent variables. For more informationabout KNN application �see Karlsson and Yakowitz �1987�,Galeati �1990�, Kember and Flower �1993�, Todini �2000�, andAraghinejad et al. �2006��.

This model gives the K most similar patterns of situation ascompared to the result of the optimization model. To estimatetransferred discharges for the current time step, generated patternsby the optimization model are used. The independent variables�Xi� in the KNN model are the reservoir inflow in the previousmonth, the reservoir storage in the current month �Si�, and themonthly demand �Di� and the dependent variable is the volumeof transferred water discharges. The combination of indepen-dent variables is called “feature vector.” The distance of eachfeature vector at time �r� and time �t� is calculated based on thesquare root of difference between values of independent variablesas follows:

Dist = �W1 � �Ir − It�2 + W2 � �Sr − St�2 + W3 � �Dr − Dt�2

�22�

where Dist=distance between current and observed data, and Wi

=weight of independent variables optimized during the calibra-tion period in KNN model.

The best value of K and weights of different independentvariables for two interbasin water transfer projects are shown inTable 3. The equations for estimating the volume of the watertransfer are as follows:

Rr = �t=1

K �1

Dist

�t=1

K1

Dist

� � Rt �23�

where Rr ,Rt=monthly estimated �r� and optimal values �t� oftransferred water discharges. For the other basin �Koohrang IIITunnel�, the KNN model also replicates the optimization results.

Table 2. Present Values of the Benefits and Costs and Their Percentage in Objective Function �Million Dollars�: �a� Solegan; �b� Koohrang III; and �c�Karoon

Agriculturalbenefit�a�

Water transferbenefit

�b�

Agriculturalcost�c�

Dredgingcost�c�

Hydropowergeneration cost

�c�

Pollutionreduction cost

�c�

Capital +OMcosts�c�

Netbenefit

1,689.41 293.82 977.25 17.80 232.50 216.21 436.71 103

85.32% 14.68% 50.39% 0.92% 12.04% 13.35% 23.29%

Note: OM�operation and maintenance.

Table 3. Best Values of K and Weights of Different Inputs in KNNModel

Plan K W1 W2 W3

Solegan 8 0.1 0.4 0.5

Koohrang III 7 0.3 0.5 0.2

�3000

�2000

�1000

0

1000

2000

3000

4000

0.4 0.5 0.6 0.7 0.8 0.9 1 1.1 1.2

price ratio/ CPD

Netbenefit(MillionDollars)

Price ratio

CPD

Fig. 8. Sensitivity of model into variation of agricultural price andCPD


Summary and Conclusion

The impact of operation of two water transfer projects, Solegan toRafsanjan �receiving Basin 1� and Koohrang III �receiving Basin2� to Zayandeh-Rud Dam from the Karoon River are investigatedin this paper. The model is developed with an economic objectivefunction considering different components of interbasin watertransfer system and it is solved using GA. The water quality simu-lation for the sending basin using an ANN model is linked withthe optimization model. In order to determine the operating rules,a KNN model is developed to be used in real time operation. Themain benefit is related to the agriculture production in the Raf-sanjan plain for a high value crop �pistachio�. The maximum costsare for the loss of agricultural production in the Khuzestan plainand the environmental costs resulted when headwater quality inthe Karoon River is altered. The results show that if the sendingbasin receives $0.38 per cubic meter of water transfer, it couldaffect the loss of agricultural income and environmental costs.The main challenge of this paper was to develop a methodologyfor water transfer project assessment. The results show that thewater transfer projects could be economical for the case studies,but further investigation is needed to include a more comprehen-sive groundwater study as well as quantifying social impacts ofwater transfer projects.

References

Araghinejad, S. �2005�. “Climate-based water resources planning andmanagement.” Ph.D. thesis, Amirkabir Univ., Tehran, Iran.

Araghinejad, S., Burn, D.H., and Karamouz, M. �2006�. “Long-leadprobabilistic forecasting of streamflow using ocean-atmospheric andhydrological predictors.” Water Resour. Res., 42, W03431.

Burn, D. H., and Yulianti, S. �2001�. “Waste-load allocation using geneticalgorithm.” J. Water Resour. Plann. Manage., 127�2�, 121–129.

Cai, X., McKinney, D. C., and Lasdon, L. S. �2001�. “Solving non-linear water management models using a combined genetic algorithmand linear programming approach.” Adv. Water Resour., 24, 667–676.

Draper, A., Jenkins, M., Kirby, K., Lund, J., and Howitt, R. �2003�.

“Economic-engineering optimization for California water manage-ment.” J. Water Resour. Plann. Manage., 129�3�, 155–164.

Feng, S., Li, L., Duan, Z., and Zhang, J. �2007�. “Assessing the impactsof South-to-North Water Transfer Project with decision support sys-tems.” Decision Support Sys., 42�4�, 1989–2003.

Galeati, G. �1990�. “A comparison of parametric and non-parametricmethods for runoff forecasting.” Hydrol. Sci. J., 35�1�, 79–94.

Gupta, J., and Zaag, P.V. D. �2008�. “Interbasin water transfers and inte-grated water resources management: Where engineering, science andpolitics interlock.” J. Phys. Chem. Earth., 33�1–2�, 28–40.

Karamouz, M. �2004�. Report of design of Karoon water quality moni-toring system and bid evaluation assistance, World Bank Report,Khuzestan Environmental Protection Office, Iran.

Karamouz, M., Mojahedi, A., and Ahmadi, A. �2007�. “Economic assess-ment of operational policies of inter-basin water transfer.” Water Re-sour. Res., 3�2�, 86–101 �in Persian�.

Karlsson, M., and Yakowitz, S. �1987�. “Nearest-neighbor methods fornonparametric rainfall-runoff forecasting.” Water Resour. Res., 23�7�,1300–1308.

Kember, G., and Flower, A. C. �1993�. “Forecasting river flow usingnonlinear dynamics.” Stochastic Hydrol. Hydraul., 7, 205–212.

Kerachian, R., and Karamouz, M. �2006�. “Optimal reservoir operationconsidering the water quality issues: A stochastic conflict resolutionapproach.” Water Resour. Res., 42, W12401.

Lund, J., and Israel, M. �1995�. “Optimization of transfer in urban watersupply planning.” J. Water Resour. Plann. Manage., 121�1�, 41–48.

Matete, M., and Hassan, R. �2005�. “An ecological economics frameworkfor assessing environmental flows: The case of inter-basin water trans-fers in Lesotho.” Glob. Planet. Change, 47, 193–200.

Rabei, F., Eslami, H., and Ghaderi, K. �2004�. “Evaluation of the effectsof the interbasin water transfer from the Karoon River on the reliabil-ity in the supplying the demand of hydropower energy.” Proc., WaterTransfer Seminar, Power and Water Univ. of Technology, Iran �inPersian�.

Todini, E. �2000�. “Real time flood forecasting operational experienceand recent advanced.” Flood issues in contemporary water manage-ment, J. Marsalek, et al., eds., Kluwer Academic, Dordrecht, TheNetherlands, 261–270.

Zahiri, A., and Korestani, S. �2004�. “The effects of interbasin watertransfer on the dredge activities in the Karoon River.” Proc., WaterTransfer Seminar, Power and Water Univ. of Technology, Iran �inPersian�.


Copyright of Journal of Irrigation & Drainage Engineering is the property of American Society of Civil

Engineers and its content may not be copied or emailed to multiple sites or posted to a listserv without the

copyright holder's express written permission. However, users may print, download, or email articles for

individual use.

Interbasin water transfer: economic water quality-based model

Documents