Professional Documents
Culture Documents
Babel - 2005 A Model or Optimal Water Allocation
Babel - 2005 A Model or Optimal Water Allocation
Babel - 2005 A Model or Optimal Water Allocation
DOI: 10.1007/s11269-005-3282-4
C Springer 2005
Key words: multi-objective decision-making, net economic return, optimization, reservoir operation,
simultaneous compromise constraint technique, water allocation, weighting technique
1. Introduction
Water is necessary for life and socioeconomic development of a country. Rapid pop-
ulation rise along with improved standards of living, urbanization and industrial
growth have led to increased demand, competition and conflicts among different
water use sectors. While traditional water demands for agriculture, domestic and
industrial sectors remain critical, other in-situ demands, such as for fishing, naviga-
tion, tourism, wetlands and environmental needs are getting increased importance.
At the same time quality degradation, physical destruction of surface water bod-
ies, higher cost of infrastructure, and temporal and spatial variation of water have
further limited the accessibility to the freshwater. As a consequence, there is an
ever-increasing pressure on water resources that are available for various uses,
which generally leads to an important issue of water allocation all over the world.
As long as water demand remains small compared to its availability, all users
can coexist without conflicts and as such the problem of water allocation does not
694 M. S. BABEL ET AL.
arise. But this is not the case; with the increase in demand the conflicts between
users intensify and increase in frequency and its impacts on water resources be-
come noticeable. Thus, to avoid the present as well as future conflicts between the
competing demands, researchers and scientists have given increased emphasis in
developing tools and techniques for improved management of the water resources.
The development and application of mathematical models, which provide a good
insight into the intricacies of various problems related to proper water allocation,
is therefore necessary.
Water allocation has received considerable attention in the recent past by the
scientific community. Bielsa and Duarte (2001) developed an economic model for
allocating water between two competing sectors, namely irrigation and hydropower
in North Eastern Spain. Similarly, Reca et al. (2001) developed an economic opti-
mization model for water resources planning in deficit irrigation systems. A linear
programming (LP) optimization model for analyzing inter-seasonal allocations of
irrigation water was introduced by Salman et al. (2001) whereas Wardlaw and Bhak-
tikul (2001) also used an optimization approach based on genetic algorithms (GAs)
for real time allocation of irrigation water supplies. Optimization models were also
developed for the Aral Sea Basin, which accounted for the multi-objective nature
of water planning as well as the uncertain nature of water availability in the basin
(McKinney and Cai, 2002). Quba´a et al. (2002) optimized the allocation pattern
for available water resources within and among the competing sectors in order to
achieve the highest economic return from water use using LP. Aquarius, a model-
ing system for river basin water allocation, was developed by Diaz et al. (2000),
which is also based on the economic efficiency operational criterion requiring the
reallocation of stream flows until the net marginal return in all water uses is equal.
Few studies on water allocation in ecological perspectives were also carried out.
Reid and Brooks (2000) analyzed the effects of environmental water allocations
(EWAs) in the Murray-Darling basin, Australia and Doupe and Pettit (2002) in
the Ord River, Western Australia. Monitoring and scientifically rigorous adaptive
management practices were applied for the success of EWAs. Voogt et al. (2000)
carried out a study on water allocation between wetlands and irrigated agriculture
in Gediz Basin, Turkey.
Rights- and market-based approaches in water allocation have also been ap-
plied. Rosegrant et al. (1995) suggested that markets in tradable water rights could
be effectively implemented with appropriate design of water laws, institutions and
regulations. Becker (1995) explored the implications of the transformation of the
system of water allocation to agriculture in Israel from one in which allocation
was practiced without any permission to trade with water rights. Huffaker et al.
(2000) studied the role of prior appropriation in allocating water resources in West-
ern U.S.A. The link between policy and basin hydrology for water allocation was
studied by Green and Hamilton (2000), where a simulation model was developed
to allocate water among three user sectors based on flow and shortage rights, con-
sumptive rights and irrigation efficiencies.
A MODEL FOR OPTIMAL ALLOCATION OF WATER TO COMPETING DEMANDS 695
2. Model Development
2.1. CONCEPTUAL FRAMEWORK
In the study the actual demand of a sector (i) is referred to as the normal demand
(Dni ). While allocating water the normal demand may or may not be satisfied.
The bare minimum amount of water required by a sector, that must be satisfied, is
defined as the minimum demand (Dmi ).
The developed model, named as integrated water allocation model (IWAM),
consists of three modules: reservoir operation module (ROM), economic analysis
module (EAM) and water allocation module (WAM) and runs on monthly time step.
ROM is based on the standard reservoir operation algorithm and requires reservoir
characteristics, inflows and operating rules as an input and the output of the module
is the amount of available water (AW), depending on the normal demand, water
availability and technical feasibility, for allocating to different demand sectors.
EAM calculates the net economic return (NER) from one cubic meter of water used
by sectors. The WAM allocates water among various competing sectors depending
on the user-defined objective(s); either one of the two single objectives that are the
696 M. S. BABEL ET AL.
maximization of satisfaction (OF1 ) and the maximization of NER (OF2 ) or these two
objectives taken together (OF12 ). The model considers six water demand sectors,
namely agriculture, domestic, industry, hydropower, recreation and environment.
The conceptual framework of the IWAM is depicted in Figure 1. Depending
on the AW and the sectoral demand, the model allocates water among various
A MODEL FOR OPTIMAL ALLOCATION OF WATER TO COMPETING DEMANDS 697
competing users. If the AW is more than the total normal demand, the water is
supplied according to the Dni . When the AW is less than the total minimum demand
of the sectors, the allocated supply is either equity-based or stressed or priority-
based.
Equity objectives are particularly concerned with the fairness of allocation
among economically desperate groups, and may or may not be consistent with
efficiency objectives. In stress-based allocation the deficiency in water is equally
shared among all the sectors concerned. Priority-based supply may be according to
the choice of the user (user-defined) or based on the doctrines in practice; namely
riparian doctrine and prior-appropriation doctrine. The user-defined priority is suit-
able for the local prevailing allocation practices (if any) and can be either single
priority or multiple priorities allocation. In case of multiple priorities the user has
the privilege to assign the parts of Dmi to be allocated to particular sectors with
high priority and low priority. If the AW is in between the total Dmi and the total
Dni , the allocation is performed using optimization techniques.
The basic algorithm used in the model for optimizing the water allocation to different
sectors is linear programming. Two techniques included in the model to turn the
two objectives into a single objective function are weighting technique (WT) and
simultaneous compromise constraint (SICCON) technique. These techniques are
described below.
Weighting Technique: This technique assigns various weights (dimensionless
relative preference values) to the objective functions according to their degree of
importance. All the objective functions are grouped into a single objective function
and hence the multi-objective decision making problem virtually turns into a sin-
gle objective decision-making problem. Ranking technique has been employed to
translate the human feelings into the quantitative term of weights.
SICCON Technique: This technique is based on a “compromise-constraint ap-
proach” (Tabucanon, 1976). To find a compromise solution between any two objec-
tives, a compromise constraint, which forces the objectives to be an equal weighted
difference from the individual optimal solution, is added to the problem. A single
objective problem with the weighted sum of the original objective functions, subjec-
tive to the compromise constraint plus the original ones, is solved. The compromise
constraints for all possible combinations (by two) of objectives are incorporated,
each of which has additional two deviational variables representing positive and
negative deviations from the supposed-to-be-zero values of the left-hand side of
the standard form of the compromise constraints. The equivalent objective func-
tion then becomes maximization of two parts: summation of all weighted objective
functions minus the summation of all deviational variables.
The principle of the compromise constraint approach used in the SICCON tech-
nique is illustrated in Figure 2 and is explained as follows:
698 M. S. BABEL ET AL.
If fh (x) and fl (x) are two objective functions expressed in same unit whose
individual maxima are respectively z h∗ and zl∗ , and if both functions decrease in
numerical value simultaneously (in other words both move inside the feasible re-
gion), then the equation of the locus of their point (or surface) of intersection has
the following form:
where, θh , θl are the rates at which the respective functions are relaxing, { f h (x) −
z h∗ }, { fl (x) − zl∗ } are the extents at which objective functions fh (x) and fl (x) are
relaxing. Since the rate of relaxation is inversely proportional to the weight, θh ∝
1/wh and θl ∝ 1/wl with wh and wl being the weights given to objectives fh (x) and
fl (x), respectively, Equation (1) becomes
In Figure 2, as the point of intersection (O) of the objectives (fh (x) and fl (x))
does not fall within the feasible region (S), the two objectives have to move inside S
A MODEL FOR OPTIMAL ALLOCATION OF WATER TO COMPETING DEMANDS 699
until such a point where their common is feasible. In other words, the point where
‘O’ first touches the feasible region (at Z∗∗ ) is the maximum feasible compromise.
This maximum feasible compromise has the label of being the “optimal” solu-
tion, which is hereinafter called the compromise solution. Due to its compromising
nature, Equation (2) is referred to as the “compromise constraint” to be added to
other constraints, thus forcing the two objective functions to pass through Z∗∗ .
Objective Functions
As stated earlier, the IWAM considers two single objective functions (OF1 and
OF2 ) and a multi-objective function (OF12 ), which are described as below:
1 n
Sai
OF 1 = (3)
n i=1 Dni
(ii) To maximize the net economic return: This objective function is defined as the
ratio of the total economic value (as represented by the summation of the water
allocated multiplied by the NER of each of the sectors) to the maximum achievable
total economic value. The objective function is represented as
6
Sa ∗ NERi
OF 2 = i=1 i
(4)
AW ∗ NERmax
(iii) To maximize the satisfaction as well as the NER of various demand sectors: This
is defined for SICCON technique as the sum of the two single objective functions
multiplied by the respective weights given to them less the sum of deviational
variables and is given as
− +
OF 12 = w1 ∗ OF 1 + w2 ∗ OF 2 − [σ12 + σ12 ] (5)
700 M. S. BABEL ET AL.
Constraints
(i) Water availability constraint:
n
Sai ≤ AW (6)
i=1
(ii) Water demand constraint:
Dni ≥ Sai ≥ Dm i (7)
(iii) Water supply constraint:
n n
Sai ≤ Dn i (8)
i=1 i=1
(iv) Non-negativity constraints:
Sai ≥ 0, Dni ≥ 0, Dm i ≥ 0 (9)
In addition to the above constraints, which are common to all the three objective
functions, an additional compromise constraint developed between OF1 and
OF2 is introduced for solving the multi-objective function OF12 by SICCON
technique. This is presented as below:
(v) Compromise Constraint:
1 6
Sai 6
i=1 Sai ∗ NERi
w1 ∗ − OF 1 − w2 ∗ − OF 2
n i=1 Dni AW ∗ NERmax
− +
+ (σ12 − σ12 )=0 (10)
However, this compromise constraint is not required for solving OF12 by WT.
In the IWAM the NER from the six sectors is calculated by EAM. This is an input
for OF2 and OF12 . The equations for the calculation of NER are provided in Annex
I and the computation process is briefly described below.
multiplied by the ratio of monthly water use to the total annual water use. When
this is divided by the total monthly water use, the NER per unit volume of water is
obtained.
Domestic Sector: The NER per unit volume of water from domestic sector is esti-
mated using the inverse demand function for water (Ringler, 2001). This is calcu-
lated as water use benefit minus water supply cost on monthly basis.
Industry Sector: The monthly NER from industry sector is also estimated using
the inverse demand function for water. This is also calculated as water use benefit
minus water supply cost on a monthly basis.
Recreation Sector: The model developed in this study may take into account vari-
ous types of recreational activities, and sites such as swimming pool, water sport,
golfing, lake resort, park, boating, fishing, picnic areas and tourist places. The NER
from recreation sector is calculated on a monthly basis.
Environment Sector: The NER from the use of water for environmental preservation
and protection is difficult to estimate. In the absence of well-established approach,
the NER from environmental sector is taken as the average of all other sectors for
the illustration.
Data required for ROM includes the monthly inflows to the reservoir, reservoir
physical characteristics (elevation-area-volume relationships) and operating rules,
rainfall and evaporation, percolation, installed capacity, tail water characteristics,
turbine efficiency and monthly water demand by the six sectors, etc. EAM requires
data related to estimation of NER for the six sectors as described earlier. Normal and
minimum water demands by the six sectors, AW, specified priority of allocation,
NER (if OF2 or OF12 is considered) and weights to be given to OF1 and OF2 (if
OF12 is considered) are the input data for WAM.
3. IWAM Application
The applicability of the IWAM is demonstrated through an example with three cases
that analyze the effect of decision-maker defined parameters on water allocation as
follows:
702 M. S. BABEL ET AL.
the government policies. For example, for a developed country the weightage given
to OF1 may be less as compared to OF2 , whereas for a developing or underdeveloped
country, the weightage given to OF1 may be more than that of OF2 .
Varying the weights from 0 to 1.0 for OF1 and OF2 and for the zero minimum
water demand by each of the sectors, the model is run using the weighting and
SICCON techniques. The results of water allocation to different sectors for OF12
are represented in Tables I and II for the two techniques.
With the increased importance given to OF1 , the value of OF12 gradually de-
creases, whereas if NER (OF2 ) is given more weightage, the value of OF12 grad-
ually increases. In case of WT (Table I), when one of the objectives is assigned
zero weight, the model optimizes water allocation for the other objective function.
However, this is not the case for the SICCON technique (Table II). With the multi-
objective function virtually turning into a single objective function, the compromise
constraint, which forces the objectives to be an equal weighted difference from the
individual optimal solutions, cancels itself and the basic compromising nature of
SICCON technique ceases.
Tables I and II show that the number of sectors allocated with full water demand
in WT is either equal or more compared to the SICCON technique. However, the
number of sectors receiving water with SICCON technique is more than WT. This
is also reflected by the higher value of the objective function with WT as compared
to the SICCON technique.
With higher significance to OF1 in WT, the sector with the least normal demand
is satisfied first and then the sector with next higher normal demand to maximize the
overall objective function. In contrast, when the significance of OF2 is increased the
demand sector having highest NER is fulfilled first followed by the sector with next
lower NER in both the techniques. It is noted that the agriculture sector is allocated
no water in this case (Tables I and II) because of the highest normal demand as well
as lowest NER among the sectors considered.
Table III presents the total economic benefit for Case 1 with two optimization
techniques. In case of WT, the highest economic return of US$ 8,133,377 is achieved
Table II. Water allocation in February for Case 1 by SICCON technique for OF12
Allocated water with different weights
Table III. Total economic benefit with various weights to two objectives for
Case 1
with three different combinations of preference values (w1 = 0.0 and w2 = 1.0;
w1 = 0.2 and w2 = 0.8; w1 = 0.4 and w2 = 0.6). This is because, in all these
combinations of weights, in this case the allocation is limited to the two (domestic
and industrial) sectors only. The lowest return of US$ 2,113,843, as expected, is
achieved with w1 = 1.0 and w2 = 0.0. The highest economic return of US$
7,667,660 with w1 = 0.2 and w2 = 0.8 and the lowest return of US$ 5,920,761
with w1 = 0.8 and w2 = 0.2 is obtained with SICCON technique.
normal demand. With zero minimum demand defined for all sectors, the water is
allocated to hydropower, recreation and environment sectors, as their requirements
are lower than the other three sectors. With the minimum demand taken as 20%
of the normal demand for all sectors, the model satisfies the minimum demand
first and then the remaining water is supplied to the sector with the lowest normal
demand and then to the next higher demand in order to maximize the objective
function. Therefore, the value of OF1 decreases with the increase in the minimum
water demand. Figures in bold face in Table IV show that the water is supplied to
these sectors with 100% satisfaction.
The results of water allocation for OF2 with varying minimum demands by
the six sectors are presented in Table V. As expected, when the objective is to
maximize the net economic return, the water is supplied to the sectors with higher
NER, however, with the constraint of satisfying the assigned minimum demand for
each of the sectors. The industry sector gets the highest priority followed by the
domestic sector in fulfilling the normal demand. Again, the value of OF2 decreases
with the increase in the minimum water demand.
Table VI presents the results for OF12 using the SICCON technique for various
minimum demands with equal weights (w1 = w2 = 0.5) given to the two single
objectives. The SICCON technique maximizes the OF12 with a compromise be-
tween the two individual objectives and hence the value of OF12 (Table VI) lies in
between OF1 (Table IV) and OF2 (Table V). When the minimum demand is zero,
the water is supplied to domestic, industry, hydropower and environment sectors in
such a way as to maximize the overall objective function.
Comparing the three objective functions for the zero minimum demand
(Tables IV–VI), the number of sectors supplied with water is 3 for OF1 , 2 for
OF2 and 4 for OF12 . The increase in number of sectors allocated with water for
OF12 is because of the compromising nature of the optimization technique. The
economic benefit in monetary terms is US$ 2,113,843 for OF1 , US$ 8,133,377 for
OF2 and US$ 6,966,834 for OF12 . Similar calculations are done for other minimum
Table VI. Water allocation in February for Case 2 by SICCON technique for OF12
demands and the results are presented in Table VII. The total economic benefit with
compromise solution, as expected, is in between the economic benefits obtained
with the two single objectives.
which has the second highest NER. As a result of limited water availability and low
economic returns, agriculture, hydropower, recreation and environment sectors are
allocated no water. For OF12 by WT, the industry and hydropower sectors are
allocated full normal demand whereas the domestic sector gets 24.7% of its normal
demand and no water is allocated to other sectors. In case of the SICCON technique,
in addition to the full satisfaction of the industry and hydropower sectors, domestic
and environment sectors are partially satisfied with 19.7% and 9.2%, respectively of
their normal water demands and the agriculture and recreation sectors get no water.
As discussed earlier, the SICCON technique tries to compromise the solution and
therefore satisfies more number of sectors.
The econometric analysis for this case also depicts quite similar results. The least
total economic return of US$2,113,843 is obtained for OF1 whereas the maximum
total economic return of about US$8,133,377 is obtained for OF2 . For the multi-
objective analysis, the weighting and SICCON techniques give the total economic
returns of US$7,111,895 and US$7,029,003 respectively.
among economic sectors including equitable allocation, water quality aspects in re-
lation to its use, sharing of benefit among the competing sectors, etc. The stochastic
nature of model variables, water availability and demand and other aspects will also
be taken into consideration in further development of the model.
Annex I
The equations for the calculation of net economic return (NER) from the six sectors
are given below.
Agriculture Sector:
N Pagr = A × B (11)
Where,
1
A = n
cp=1 M W (m, cp)
n
(A × Y × P)agr − A ∗ {F + M + L + O}agr − W Sagr ∗ 12
m=1 M W (m, cp)
B= 12
cp=1 m=1 M W (m, cp)
∗ {M W (m, cp)}12
m=1
Domestic Sector:
1 w(dom) α 1
N Pdom = w0 (dom) ∗ p0 ∗ ∗ + 0.743 −
1+α w0 (dom) 1+α
−w(dom) ∗ ws c(dom) (12)
710 M. S. BABEL ET AL.
Where,
NPdom = Net profit from domestic sector (US$/m3 )
w0 (dom) = Maximum normal monthly withdrawal for domestic sector (m3 )
w(dom) = Actual water withdrawal for domestic sector (m3 )
p0 = Value of water in domestic sector at full use (US$/m3 )
e = Price elasticity of demand in domestic sector
α = 1/e
ws c(dom) = Water supply cost for domestic sector (US$/m3 )
Industry Sector:
1 w(ind) α 1
NPind = w0 (ind) ∗ p0 ∗ ∗ + 0.743 −
1+α w0 (ind) 1+α
−w(ind) ∗ ws c(ind) (13)
Where,
Power Sector:
Where,
Recreation Sector:
1
n n
n
NPrec = ∗ Bavi − Q avi ∗ W Sprice − O Mavi Q avi (15)
n i=1 i =1 i =1
A MODEL FOR OPTIMAL ALLOCATION OF WATER TO COMPETING DEMANDS 711
Where,
References
Becker, N., 1995, ‘Value of moving from central planning to a market system: Lessons from the Israeli
water sector’, Agric. Econ. 12(1), 11–21.
Bielsa, J. and Duarte, R., 2001, ‘An economic model for water allocation in north eastern Spain’,
Water Res. Dev. 17(3), 397–410.
Diaz, G. E., Brown, T. C., and Sveinsson, O., 2000, Aquarius: A Modeling System for River Basin
Water Allocation, General Technical Report RM-GTR-299-Revised, U.S. Department of Agri-
culture, Forest Services, Rocky Mountain Forest and Range Experiment Station, Fort Collins,
CO. Retrieved from http://unfccc.int/program/mis/meth/aquarius.pdf on 12 August 2004.
Doupe, R. G. and Pettit, N. E., 2002, ‘Ecological perspectives on regulation and water alloca-
tion for the Ord River, Western Australia’, River Res. Appl. 18(3), 307–320. Retrieved from
http://www3.interscience.wiley.com/cgi-bin/abstract/93520263/STRAT on 12 August 2004.
Green, G. P. and Hamilton, J. R., 2000, ‘Water Allocation, Transfers and Conservation: Links between
Policy and Hydrology’, Water Res. Dev. 16(2), 197–208.
Huffaker, R., Whittleysey, N., and Hamilton, J. R., 2000, ‘The role of prior appropriation in allocating
water resources into the 21st century’, Water Res. Dev. 16(2), 265–273.
McKinney, D. C. and Cai, X., 2002, ‘Multiobjective Optimization Model for Water Allocation in
the Aral Sea Basin: Stochastic Optimization’, Draft Working Paper, Environmental and Water
Resources Engineering Program, University of Texas at Austin, Austin, Texas.
Nayak, D. K., 2003, ‘Modelling Water Allocation in Competing Demand’, Master’s thesis, Water
Engineering and Management, Asian Institute of Technology, Bangkok, Thailand.
Quba a, R., El-Fadel, M., and Darwish, M. R., 2002, ‘Water pricing for multisectoral allocation: A
case study’, Water Res. Dev. 18(4), 523–544.
Reca, J., Roldan, J., Alcaide, M., and Camacho, E., 2001, ‘Optimization model for water allocation
in deficit irrigation system: I. Description of the model’, Agric. Water Manage. 48(2), 103–
116.
Reid, M. A. and Brooks, J. J., 2000, ‘Detecting Effects of Environmental Water Allocations
in Wetlands of the Murray-Darling Basin, Australia’, Regulated Rivers: Research and Man-
agement 16(5), 479–496. Retrieved from http://www3.interscience.wiley.com/cgi-bin/abstract/
73501955/START on12 August 2004.
Ringler, C., 2001, ‘Optimal Water Allocation in the Mekong River Basin’, ZEF-Discussion Papers
on Development Policy No. 38, Center for Development Research, University of Bonn, Bonn,
Germany, pp. 50.
Rosegrant, M. W., Schleyer, R. G., and Yadav, S. N., 1995, ‘Water policy for efficient agricultural
diversification: Market-based approaches’, Food Policy 20(3), 203–223.
Salman, A. Z., Al-Karablieh, E. K., and Fisher, F. M., 2001, ‘An inter-seasonal agricultural water
allocation system (SAWAS)’, Agric. Systems 68(3), 233–252.
712 M. S. BABEL ET AL.
Tabucanon, M. T., 1976, ‘Multiobjective Linear Programming for Decision Making in Industrial
Systems’, Ph.D. dissertation, Industrial Systems and Engineering, Asian Institute of Technology,
Bangkok, Thailand.
Voogt, K. D., Kite, G., Droogers, P., and Murray-Rust, H., 2000, ‘Modeling water allocation between
a wetland and irrigated agriculture in the Gediz Basin, Turkey’, Water Res. Dev. 16(4), 639–650.
Wardlaw, R. and Bhaktikul, K., 2001, ‘Application of generic algorithm for water allocation in an
irrigation system’, Irrigation and Drainage 50(2), 159–170.