Agricultural Water Management 36 Ž1998.

1–21

Two-level optimization model for Lower Indus

Basin
N.K. Garg ) , Abbas Ali
Dept. of CiÕil Engineering, I.I.T., New Delhi-16, India
Accepted 27 November 1997

Abstract

The present study aims at the development of a model to schedule the sowing dates of the
crops in such a manner that the peak water requirements of different crops are more uniformly
distributed over different months and hence, more area can be irrigated for a given canal and tube
well capacities. For this purpose, a two-level optimization model for optimal use of surface and
groundwater has been developed. In the first level, the model gives optimal cropping patterns and
monthly water withdrawals from canal and tube well for a given set of sowing dates to maximize
the net economic returns. At second level, the sowing dates are varied within the allowable limits
and the optimized sowing dates are obtained using an integer programming model. The sowing
dates at the first level are then taken as the sowing dates obtained from the second level. The
process is repeated until it converges. The sensitivity analysis for some of the parameters is also
done. The model is applied to Dadu canal command of Lower Indus Basin. The results show an
overall increase of 40% in the crop intensities and 38% in the benefits over the existing ones by
using the two-level optimization model. q 1998 Elsevier Science B.V.

Keywords: Conjunctive use; Optimal cropping pattern; Irrigation management

1. Introduction

The Indus Basin irrigation system is one of the largest of its kind in the world. The
water demand of the system far exceeds its related water resources. It is, therefore, very
important to optimally manage and use the scarce water resources. Considerable work
has been done on the Indus Basin System ŽHarza Engineering International, 1963;
Revelle, 1964; Irrigation and Agricultural Consultants Association, 1966; Tipton and

)
Corresponding author.

0378-3774r98r$19.00 q 1998 Elsevier Science B.V. All rights reserved.

PII S 0 3 7 8 - 3 7 7 4 Ž 9 7 . 0 0 0 5 7 - 7
2 N.K. Garg, A. Ali r Agricultural Water Management 36 (1998) 1–21

Kalmbach, 1967; Lieftinck et al., 1968; Chaudhry et al., 1974; O’Mara and Duloy,
1984.. All these studies have recommended the conjunctive use of surface and ground-
water in the Indus Basin to enhance cropping intensities.
Several studies have dealt with the development of models for conjunctive use of
surface and groundwater. Nieswand and Granstron Ž1971. have developed a set of
chance constrained linear programming models for the conjunctive use of surface and
groundwater. Smith Ž1973. has developed a linear programming model for irrigation
planning in countries with little experience of major water control efforts. Wanyoung
and Haimes Ž1974. have developed a systems analysis approach for optimal conjunctive
use of surface and groundwater. They have chosen a two-dimensional asymmetrical grid
network model to represent the aquifer system. Haimes and Dreizin Ž1977. have
developed a methodology for solving the problem of large scale groundwater systems by
decomposing the large system into interacting submodels. Rydzewski and Rashid Ž1981.
have developed an approach for optimal allocation of surface and groundwater resources
to three agricultural areas in the Jordan Valley. Khepar and Chaturvedi Ž1982. have
developed the decision models for optimal groundwater management alternatives in
conjunction with optimal cropping pattern using fixed yield and alternative levels of
water use approaches based on water production functions. Bredefoeft and Young
Ž1983. have studied the extent of groundwater development to ensure water availability
against periods of low stream flows. They have used a simulation model which couples
hydrology of the conjunctive stream aquifer system to an economic model. Paudyal and
Gupta Ž1990. have used the multilevel optimization techniques to solve the problem of
irrigation management in the large heterogeneous basin. They have decomposed the
large system into interconnected subsystems to reduce the dimension of the problem and
solved it through a multilevel iterative process.
All these studies have addressed the various problems associated with the conjunctive
use but none have studied the effects of adjustments in peak water requirements which
can have significant impact on the cropping pattern.
In this study, a two-level optimization model has been developed to obtain the
optimal cropping pattern considering the impact of variation in sowing dates on the peak
water requirements. At the first level, a linear programming model is developed which
optimizes the net return from the crops and depicts the optimal cropping pattern and
monthly water withdrawals from canal and tube well for a given set of sowing dates. At
the second level, an integer programming model has been developed for optimizing the
sowing dates using a linearized relationship between crop coefficient and percentage
growing season. Both the objective functions are optimized independently at first and
second levels, respectively.

2. Study area

The Indus Basin in Pakistan contains one of the largest irrigated area in the world.
The available area in the basin is about 20 million ha. The basin can be subdivided into
upper and lower Indus Basin. The lower Indus Basin as shown in Fig. 1 consists of
about 6 million ha area irrigated by three barrages viz. Gudu, Sukker and Kotri.
 N.K. Garg, A. Ali r Agricultural Water Management 36 (1998) 1–21 3

Fig. 1. The Study area with Lower Indus Canal System.

The study area, commanded by the Dadu Canal off taking from right bank of Sukker
barrage, consists of 210,000 ha Cultureable Command Area ŽCCA. and lies between
latitude 298N and longitude 678E ŽFig. 1.. The area is situated in a hot arid zone. The
climate is characterized by large seasonal fluctuations in temperature, sparse rain fall,
with an average annual rainfall of 91 mm, falling mainly in July and August and high
potential evapotranspiration.
The major sources of water include canal water, groundwater and hill torrents.
However, generally very little flow reaches the study area from hill torrents and
4 N.K. Garg, A. Ali r Agricultural Water Management 36 (1998) 1–21

possibilities of its exploring is limited considering the anticipated cost of development.

Most of the area is underlain by a large alluvial sand aquifer and the depth of ground
water ranges between 30 m and 60 m. The canal water is supplied through Dadu canal
having capacity of 89.00 m3rs. Rice is the dominant crop in Kharif Žfrom May to
October., while the other main crops are cotton, sorghum and oilseed. In Rabi Žfrom
November to April., wheat is the principal crop along with mainly gram, mustard and
fodder. Sugarcane is growing all year round. The existing cropping intensities are 28%
for Kharif and 50% for rabi. The surface drainage facilities are provided under North
Dadu Surface Drainage Project.
The irrigated area is level flood plain and consists of mainly two soil groups viz.
Indus Alluvial and Piedmont Alluvial. The Indus Alluvial consists of mainly fine sandy
loam and occasional silt clay while Piedmont Alluvial comprises of deep dense clays
and stratified medium to fine textured materials. The Indus Alluvial covers approxi-
mately 75% of the area and the rest is covered by the Piedmont Alluvial.

3. Formulation of optimization model

The two-level optimization model as shown in Fig. 2 has been developed to obtain
the optimal cropping pattern and the sowing dates. The cropping areas are obtained at
first level using a linear programming model for a given set of sowing dates and are fed
at second level into an integer programming model to obtain the optimized sowing dates.
The non-linear relationship between crop coefficient and percentage growing season has
been approximated as the sum of linearized segments. The relationship between the crop
coefficient and the sowing date for a particular crop and month is obtained using
regression analysis. The initial sowing dates at first level is then replaced by new dates

Fig. 2. A schematic representation of the proposed model.

 N.K. Garg, A. Ali r Agricultural Water Management 36 (1998) 1–21 5

obtained at second level and new cropping pattern is obtained. The process is repeated
until it converges. From the irrigation project reports prepared for the Lower Indus Basin
ŽHarza Engineering International et al., 1991., it is found that the actual operation and
maintenance cost Žnot the subsidized cost. of canal water is Rs105 for every ha m
ŽUS$1.00 s Rs.40. while for the tube well water for an average lift this cost is
Rs950rha m. As the canal water is always cheaper Žincluding conveyance losses. than
the tube well water therefore the optimization model will use the groundwater only if the
surface water is exhausted. Hence, the coupling of groundwater hydraulics with surface
water is not considered in the optimization model. However, the interaction between
surface water and groundwater is considered in the model by imposing a groundwater
balance constraint on the groundwater withdrawals on annual basis. The optimum policy
for pumping the optimized volume of groundwater obtained from the proposed model
has been studied separately.
Further, the non-linear relationship between the pumping cost and the pumping head
is also not considered in this study because of the fact that groundwater will be utilized
only if surface water is exhausted. Therefore, variation in the pumping cost with the
change in pumping head will not effect the optimal pattern and may result in minor
variations in the net benefits only. However the non-linearity has been taken into
account while formulating the groundwater withdrawal model Žwork will be reported
elsewhere..

4. First level linear programming model

The linear programming model for the first level is conceptually similar to that
developed by Smith Ž1970., and used by Lakshminarayana and Rajagopalan Ž1977.. Ten
crops—rice, cotton, sorghum, oil seed and fodder in Kharif and wheat, gram, mustard
and fodder in Rabi along with the perennial sugarcane have been considered. These
crops are the major crops growing in Rabi and Kharif seasons and are similar to the
existing pattern. The objective function comprises of four components: return from
crops, cost of operation and maintenance of canal, tube well and surface drainage
respectively. No change in the existing farm practices, i.e., use of fertilizers, seeds, etc.
has been assumed in the model. Therefore, the non-water related costs are considered in
the crop production cost while calculating the return from crops. The model is applied
on monthly basis and therefore, the variables in the model are: monthly withdrawals
from the canal and tube well for irrigation, monthly pumpage from tube well to drainage
and areas under different crops on different categories of soil. The objective function can
be mathematically expressed as under.
NC TS
Maximize Z s Ý Ý A i j )YPH i j )VPQ i
is1 js1

12
y Ý C sw swt q C gw Ž gwt q pd t . q C dw dwt Ž 1.
ts1
6 N.K. Garg, A. Ali r Agricultural Water Management 36 (1998) 1–21

where Z s objective function; i s index for the crops; j s index for the type of soil;
t s index for the time period Žmonth.; NC s total number of crops; TS s total types of
soil; A i j s Irrigated area of ith crop under jth type of soil Žha.; YPH i j s yield per ha
Žquintalsrha, 1 quintalss 100 kg. of ith crop under jth type of soil; VPQ i s value per
quintal ŽRsrquintal. of ith crop; C sw , C gw , C dw s operation and maintenance cost of
canal, tube well and drainage respectively ŽRsrha m.; swt s flow diverted to canal for
irrigation in t th month Žha m.; gwt s pumping from tube well for irrigation in t th month
Žha m.; pd t s pumpage from tube well to drainage in t th month Žha m.; dwt s water
drained by surface drainage in t th month Žha m..dwt can be expressed as:
dwt s swt Ž RO1 q a 1 RO 2 . q gwt RO 2 q pd t
NC TS
q RFt Ž RO 2 y RO 3 . ) Ý CFi t Ý A i j q RFt q RO3 )CCA
is1 js1

where RO1 s fraction of water delivered to canal lost as surface runoff; RO 2 s fraction
of water delivered to irrigated area lost as surface runoff; RO 3 s fraction of available
water Žrainfall. to non-irrigated area lost as surface runoff; RFt s rainfall in the t th
month Žmt.; CFi t s crop factor for ith crop in t th month, equal to 1 if crop i is available
in t th month otherwise, it is 0; CCA s cultureable command area Žha.; a 1 s fraction of
canal water delivered to irrigated area. a 1 can be expressed as
a 1 s 1 y RO1 y GR 1 y ET1
in which GR 1 s fraction of water delivered to canal, lost as groundwater recharge;
ET1 s fraction of water delivered to canal, lost as non-beneficial evapotranspiration.
Limited heterogeneity has been considered in the model by introducing variable soil
types. Further, it is to be noted that the operation cost of canal and drainage system has
been taken as uniform for the entire study area while the tube well operation cost is
calculated by taking the average pumping lift in the well.
The objective function is bounded by several constraints, as follows.
Ž1. Monthly canal diversion must not exceed water available at the river.
swt F RWt Ž 2.
in which RWt s water available at river in t th month.
Ž2. Monthly water requirement of crops must be met.
NC TS NC TS
Ý Ý A i j )WR t i j y a 2 Ž a 1 swt q gwt . q Ý CFi t Ý A i j ) a 2 RFt F 0.0 Ž 3.
is1 js1 is1 js1

where WR t i j s water requirement of ith crop under jth type of soil in t th month;
a 2 s fraction of water delivered to the irrigated area, available for consumptive use. a 2
can be expressed as
a 2 s 1 y RO 2 y GR 2 y ET2
where GR 2 s fraction of water delivered to irrigated area lost as groundwater recharge;
ET2 s fraction of water delivered to irrigated area lost as non-beneficial evapotranspira-
tion.
 N.K. Garg, A. Ali r Agricultural Water Management 36 (1998) 1–21 7

It may be noted that a factor CFi t has been introduced in the above constraint while
calculating the contribution of the rainfall. It is necessary since the rainfall falling on
uncropped areas in a particular month is not available for the consumptive use. This fact
was not considered by Lakshminarayana and Rajagopalan Ž1977. when calculating the
rainfall contribution.
Ž3a. Crop area in each month under a particular type of soil must not exceed the
maximum area available under that type of soil.
In the model formulated by Lakshminarayana and Rajagopalan Ž1977., the constraint
on total cropped area was imposed on annual basis which restricts the maximum annual
intensity of irrigation to 100%. This constraint has been modified in this model and
imposed on a monthly basis. This will allow the intensity of irrigation to go up to 200%
Ž100% in each rabi and kharif seasons. and will ensure that the overlapping of crop
areas in both the seasons is averted.
NC
Ý CFi t A i j F CCA j Ž j s 1,2, . . . TS . Ž 4a .
is1

where CCA j s cultureable command area under jth type of soil Žha..
Ž3b. A minimum area may be required to be irrigated in a year under a particular type
of soil.
NC
Ý A i j G u j CCA j Ž j s 1,2, . . . TS. Ž 4b.
is1

where u j s fraction of cultureable command area under jth type of soil.

Ž4a. The existing crop production must be met.
In the absence of reliable data for the minimum requirements and to avoid a sudden
change in the existing equilibrium, it is assumed that the minimum crop production must
be at least equal to the existing level.
TS
Ý A i j )YPH i j G bi TCP Ž i s 1,2, . . . NC. Ž 5a.
js1

where bi s fraction of total crop production allotted to ith crop for the existing
production; TCP s total existing crop production Žquintals.
Ž4b. Maximum production of a crop must not exceed the allowable limit.
TS
Ý A i j )YPH i j F Yi TCP Ž i s 1,2, . . . . . . NC. Ž 5b.
js1

where g i s fraction of total crop production allotted to ith crop for the maximum
production.
Ž5. The monthly quantity of surface water, groundwater and surface drainage must
not exceed capacities of canal, tube well and surface drainage system, respectively.
swt
n tsw sw F CC Ž 6.
Nt )h sw
8 N.K. Garg, A. Ali r Agricultural Water Management 36 (1998) 1–21

gwt q pd t
n tgw F TWC Ž 7.
Ntgw )h gw
dwt
n tdw F DC Ž 8.
Nt )h dw
dw

where n tsw , n tgw , n tdw s ratio of peak to average demand of canal, tube well and drainage
systems in t th month respectively; Nt sw , Ntgw , Ntdw s operational days of canal, tube
well and drainage systems, respectively in t th month; h sw , h gw , h dw s efficiency of
canal, tube well and drainage systems, respectively; CC, TWC, DC s daily maximum
capacities of canal, tube well and drainage systems Žha m., respectively.
Ž6. Total water withdrawal from aquifer in a year must not exceed the mining rate.
12
Ý Ž pd t q gwt . y GR 1 )swt y GR 2 Ž a 1 swt q gwt .
ts1

NC TS NC TS

is1 js1
ž
yRFt )GR 2 ) Ý CFi t Ý A i j y RFt )GR 3 CC A y Ý CFi t Ý A i j
is1 js1
/
qAETq ARF q ASTSq ASD y ASFS
y AAR F AMR Ž 9.
where AET s annual evapotranspiration from aquifer Žha m.; ARF s annual aquifer
return flow Žha m.; ASTSs annual seepage to streams and lakes Žha m.; ASD s annual
spring discharge Žha m.; ASFSs annual seepage from streams and lakes Žha m.;
AAR s annual recharge from adjoining aquifer Žha m.; AMRs allowable mining rate.
In the preceding groundwater balance equation, the determination of many terms is
quite difficult because of non-availability of relevant data in the study area. We have
assumed that total withdrawal from the system through pumping can not exceed the
groundwater recharge available by seepage from the canal, irrigated area and rainfall in
the study area. The groundwater balance equation thus becomes:
12
Ý Ž pd t q gwt . y GR 1 )swt y GR 2 Ž a 1 swt q gwt .
ts1
NC TS
yRFt )GR 2 ) Ý CFi t Ý A i j
is1 js1

NC TS

ž
yRFt )GR 3 CCA y Ý CFi t Ý A i j
is1 js1
/ F 0.00 Ž 10 .

5. Second level optimization model

At the second level the authors propose an integer programming model to obtain the
optimized sowing dates. The peak demand of different crops can coincide in a particular
 N.K. Garg, A. Ali r Agricultural Water Management 36 (1998) 1–21 9

month exhausting the available water supply and limiting the total crop production. The
peak period of different crops is a function of the sowing dates of those crops.
Therefore, a model is proposed as Second Level Optimization to minimize the peak
water requirement by adjusting the sowing dates. The variation in the sowing date is
restricted within reasonable bounds by bringing forward or postponing from the normal
sowing date.
In order to calculate the crop water requirement, any of the four methods described
by Doorenbos and Pruitt Ž1977. can be used to determine reference crop evapotranspira-
tion Et . For converting Et values into crop water requirement, suitable crop coefficients
k should be evolved for different crops, soils and climatic conditions and also for
different stages of growth of the same crop. As far as possible, k values should be
verified from the direct methods available to determine the crop water requirements.
Therefore, monthly crop water requirement, in general, can be expressed as
NC TS
Monthly WR t i j s Ý Ý A i j Et k t i j Ž 11 .
is1 js1

where Et s reference crop evapotranspiration in t th month; k t i j s crop coefficient of the

ith crop in t th month under jth type of soil.
The crop coefficient Ž k . is a non-linear function of percentage growing season. The
relationship between the crop coefficient and percentage growing season for a particular
month can be approximated as a linear one. The percentage growing season is a function
of sowing date and therefore, the crop coefficient for a particular month depends upon
the sowing date of a particular crop. The relationship between crop coefficient and
sowing date for a particular month, soil and crop is obtained through regression analysis.
This relationship can be expressed as
kti j s mj xi q cj Ž 12 .
where x i s sowing date for the ith crop in the region; m j and c j are regression
constants for jth type of soil.
Using Eqs. Ž12. and Ž11., the monthly water requirement for a particular month in
terms of sowing date will become:
NC TS
Monthly WR t i j s Ý Ý A i j Et Ž m j x i q c j . Ž 13 .
is1 js1

The objective function for the second level will then become the minimization of the
monthly water requirement, subject to the constraint that the sowing date can only be
varied within the allowable limits and can be formulated as the following integer
programming model:
NC TS
Minimize WR t i j s Ý Ý A i j Et Ž m j x i q C j .
is1 js1

Subject to PR i F x i F PTi Ž i s 1,2,3, . . . ,NC., where PR i s maximum bringing forward

limit of the ith crop; PTi s maximum postponing limit of the ith crop.
10 N.K. Garg, A. Ali r Agricultural Water Management 36 (1998) 1–21

Additional constraints can be imposed in the model considering the available

resources. For example, it may be desirable to keep a minimum difference in the sowing
dates of different crops of a particular cropping season. It is to be noted that the bringing
forward and postponing of the sowing of the crops are restricted within a limit such that
the variations do not effect the crop yield. From the studies conducted on the Indus basin
ŽSir M. MacDonald and Partners et al., 1988., it is revealed that the sowing date of a
crop can vary with in 1 month without effecting the yield. Otherwise, the crop yield
must also be changed at first level according to the optimized sowing dates obtained
from the second level in successive iterations. The optimal cropping areas obtained from
first level are taken as the crop areas for the second level objective function.
By choosing the sowing dates corresponding to the maximum of these optimized
monthly minimum values will ensure that the water demand for different crops for the
peak month will be minimum. Therefore, the dates corresponding to the months having

Fig. 3. The flow chart for the proposed two-level optimization model.
 N.K. Garg, A. Ali r Agricultural Water Management 36 (1998) 1–21 11

Fig. 4. The relationship between crop coefficient and percentage growing season for sugarcane.

maximum values of second level objective function for Kharif and Rabi crops respec-
tively, are selected for the input in the model at first level. The process is repeated until
it converges. The approach has been schematically presented in flow chart ŽFig. 3.. The
first level linear programming model is solved using simplex method and second level
integer programming model is solved using cutting plane technique ŽBilly and Gillett,
1976..
To elaborate Eq. Ž12., an example calculation, showing the relationship between the
crop coefficient and sowing dates for sugarcane crop, is presented here.
The non-linear relationship between crop coefficient Ž k . and percentage growing
season for sugarcane is shown in Fig. 4. To obtain regression constants Žm & c., the
sowing dates are varied within the allowable bringing forward and postponing limits and
the crop coefficients for various sowing dates are obtained from the curve ŽFig. 4. after
calculating the percentage growing season on monthly basis. The values of monthly crop
coefficient corresponding to various sowing dates are given in Table 1.
The sowing dates also include the sowing month. It would require an adjustment for
the month also if the sowing dates are varied within the allowable limits which may
change the sowing month. To adjust for the month, integer variable x i is varied between
maximum bringing forward and postponing limits, that is 0–30 Žthe data from the area
ŽSir M. MacDonald and Partners et al., 1988., indicates the bringing forward and
postponing limit to be 15 days. and the normal sowing date has been assumed at x i
equals to 15. For example, the sowing date of sugarcane is March 1, therefore x i s 0
would mean February 14, x i s 30 means March 16 and x i s 15 means normal sowing
date.
Once the crop coefficients corresponding to the different sowing dates are obtained,
the linear regression analysis is performed to obtain the regression constants Ž m and c .
for each month. The values of regression constants for sugarcane for each month of its
growth period are given in Table 1.

6. Model application
The two-level optimization model, formulated in previous section, has been applied
to the Dadu canal command of the Lower Indus Basin. The main data for this command
 12
N.K. Garg, A. Ali r Agricultural Water Management 36 (1998) 1–21
Table 1
The values of crop coefficients and regression coefficients for sugarcane
Sowing date x Monthly crop coefficient Ž k .
March April May June July August Sept. Oct. Nov. Dec. Jan.
Feb. 14 0 0.58 0.68 0.77 0.86 0.95 1.00 0.96 0.87 0.78 0.69 0.54
Feb. 21 7 0.56 0.66 0.74 0.84 0.93 1.00 0.98 0.89 0.80 0.71 0.57
March 1 15 0.54 0.63 0.72 0.81 0.90 1.00 1.00 0.92 0.82 0.74 0.59
March 9 23 0.53 0.61 0.70 0.79 0.88 0.97 1.00 0.94 0.85 0.76 0.64
March 16 30 0.52 0.59 0.68 0.77 0.86 0.95 1.00 0.96 0.87 0.78 0.68
Monthly regression coefficient Ž m. y0.00223 y0.00302 y0.00288 y0.00302 y0.00302 y0.00171 0.001314 0.00302 0.00302 0.00302 0.0046
Monthly regression constant Ž c . 0.5814 0.6794 0.7653 0.8594 0.9494 1.0096 0.9682 0.8705 0.7786 0.6906 0.5349
 N.K. Garg, A. Ali r Agricultural Water Management 36 (1998) 1–21 13

Fig. 5. The process of convergence.

has been obtained from Harza Engineering International et al. Ž1991. and Sir M.
MacDonald and Partners et al. Ž1988. and some important data are given in Appendix A.
Monthly crop water requirement has been calculated using A Guide for Estimating
Irrigation Water Requirements, 1984, which has extensively used data from Doorenbos
and Pruitt Ž1977.. The required data for calculating the crop water requirement are taken
from the work of Harza Engineering International et al. Ž1991. and Sir M. MacDonald
and Partners et al. Ž1988.. These calculated crop water requirement values are verified
from the available field values.
The existing pattern shows 30% losses in conveyance system and 32% losses in the
field. The same figures have been taken in this study.

7. Optimum cropping pattern and monthly water releases

The two types of crop area variation patterns have been considered. In the variation
pattern-1, the sowing date for a particular crop is kept the same throughout the command
area and the normal sowing dates are optimized using the second level optimization
model while in the variation pattern-2, the crop area of a particular crop is divided into
different zones and the sowing dates are optimized for each of such cropping zones

Fig. 6. The optimal cropping pattern.

14 N.K. Garg, A. Ali r Agricultural Water Management 36 (1998) 1–21

Fig. 7. Water releases for crop area variation pattern-1.

Fig. 8. Water releases for crop area variation pattern-2.

allowing the flexibility in sowing dates of the same crop in the same command in
different zones. The crops considered for distributed shifting includes sugarcane, rice,
cotton, oilseed, wheat and fodder which are the major crops of pattern-1 and the crop
areas are divided into four zones. The model required a maximum number of two
iterations to converge. The process of convergence is shown in Fig. 5.
The optimal cropping pattern and water releases obtained from two-level optimization
model for both the crop area variation patterns are presented in Figs. 6–8.
The results show a considerable improvement over the existing cropping pattern as
shown in Fig. 9. There is an improvement of 24.4% in crop intensities and benefits with
normal sowing dates. There is a further increase of 15.6% in cropping intensities and
13.7% in benefits after optimizing the sowing dates at second level optimization model

Fig. 9. The comparison between existing and proposed cropping patterns.

 N.K. Garg, A. Ali r Agricultural Water Management 36 (1998) 1–21 15

Fig. 10. The crop water requirement for crop area variation pattern-1.

with an overall improvement of 40% over the existing cropping intensities and 38% over
the benefits. A further improvement of 5.4% in crop intensities and 4.2% in benefits is
obtained by using variation pattern-2. It may be noted that there is a total increase of
45.3% in cropping intensities with 42.2% increase in benefits as compared to the
existing ones.

8. Variation of water requirement with sowing dates

Fig. 10 shows the crop water requirements for each month obtained for crop area
variation pattern-1 and Fig. 11 shows these results for crop area variation pattern-2.
Comparing the crop water requirements for normal sowing dates ŽFig. 12. with the
pattern obtained from the model for both the crop area variation patterns, it can be seen
that the total monthly water requirements of all the crops are now more uniformly

Fig. 11. The crop water requirement for crop area variation pattern-2.
16 N.K. Garg, A. Ali r Agricultural Water Management 36 (1998) 1–21

Fig. 12. The crop water requirement for normal sowing dates.

distributed. The optimized distribution of peak water demands of different crops now
makes it possible to cultivate more area within the existing water availabilities, canal
and tube well capacities.

9. Effect of increasing tube well capacities

The existing tube well capacities are exhausted during peak demand period while the
slackness is found in the groundwater balance constraint and it is behaving as inequality
constraint. Therefore the annual recharge is more than the annual withdrawals from the
groundwater and more benefits can be achieved by increasing the tube well capacities.
Fig. 13 shows the effect of increasing the tube well capacities on benefits. The results
show a maximum increase of 56.4% in crop intensities with 60.8% in benefits as
compared to the existing ones and the tube well capacities can be increased up to a

Fig. 13. The effect of increasing tube well capacities on optimal benefits.
 N.K. Garg, A. Ali r Agricultural Water Management 36 (1998) 1–21 17

Fig. 14. The optimal tube well capacities corresponding to seepage losses.

maximum limit of 76% more than the existing ones. The increase in tube well capacities
can be achieved within the overall framework of the existing infrastructure.

10. Effects of seepage losses

Fig. 14 shows the optimal tube well capacities corresponding to the seepage losses
from the conveyance systems and the irrigated fields. It can be seen that the existing
tube well capacities are more than the required optimal capacities for the seepage losses
up to 10%. The optimal tube well capacities will increase to 57% more as compared to
the existing ones corresponding to 20% seepage losses and will attain a maximum value
of 76% more than the existing ones for the 47% seepage losses.
Fig. 15 shows the effect of seepage losses on the benefits assuming both the optimal
and existing tube well capacities. It can be seen that the benefits obtained from both the
capacities are almost same for the seepage losses up to 20%. Beyond 20% losses, the
benefits are more with optimal capacities as compared to the existing ones. Further, the
benefits are maximum corresponding to 20% seepage losses and the seepage losses can
be reduced up to this limit. As the difference in benefits obtained from optimal and
existing capacities for 20% losses is insignificant, the existing tube well capacities will

Fig. 15. The optimal benefits corresponding to the existing and optimal tube well capacities.
18 N.K. Garg, A. Ali r Agricultural Water Management 36 (1998) 1–21

be sufficient up to this limit. The seepage losses can be reduced either by lining the
conveyance systems or by improving the farm water management techniques.

11. Conclusion
A two-level optimization model is developed in this study and is applied to Dadu
canal command of the lower Indus Basin. The results show an overall increase of 40%
in the crop intensities and 38% in benefits over the existing ones. This increase is
obtained mainly by optimizing the crop sowing dates and without making any change in
the infrastructure. An additional 5.4% increase in area and 4.2% increase in benefits will
be obtained by just dividing the crop area in four zones having different sowing dates
which can also be managed within the existing infrastructure. An increase of 56.4% in
crop intensities from existing ones can be obtained by increasing the tube well capacities
to 76%. The irrigation efficiencies taken in this study are the same as the existing ones.
Further improvement in the cropping intensities can be obtained by improving the
existing efficiencies. The two-level optimization model is quite general and can be
applied to other irrigation management systems.

Acknowledgements
The writers wish to express their sincere thanks to Prof. M.C. Chaturvedi for his
valuable comments and suggestions.

Appendix A
The important data used in the present study are given below:

Canal capacity 89 m3rs

Tube well Capacity 19.84 m3rs
Drainage capacity 25.51 m3rs
Canal efficiency 80%
Ratio of peak to average demand 1.1
Runoff losses from conveyance system 10%
Recharge losses from conveyance system 15%
Evaporation losses from conveyance system 5%
Runoff losses from irrigation field 10%
Recharge losses from irrigation field 12%
Evaporation losses from irrigation field 10%
Canal water cost Rs105rha m
Tube well water cost Rs950rha m
Drainage water cost Rs60rha m
Maximum preponing limit for sowing dates 15 days
Maximum postponing limit for sowing dates 15 days
 N.K. Garg, A. Ali r Agricultural Water Management 36 (1998) 1–21 19

Appendix B. Notation

The following symbols are used in this paper:

A Irrigated area under a crop

AAR Annual aquifer recharge
AET Annual evaporation from water table
AMR Annual mining rate
ARF Annual return flow
ASD Annual spring discharge
ASFS Annual seepage from streams and lakes
ASTS Annual seepage to streams and lakes
c Regression constant
C dw Operation cost of drainage system
C gw Operation cost of tube well
C sw Operation cost of canal
CC Maximum capacity of canal
CCA Cultureable Command Area
CF Crop factor
DC Maximum capacity of drainage system
dw Water drained by surface drainage system
E Reference crop evapotranspiration
ET1 , ET2 Fraction factors for non-beneficial evapotranspiration from canal
and irrigated area respectively
GR 1 , GR 2 , GR 3 Fraction factors for groundwater recharge from canal, irrigated
area and non-irrigated area respectively
gw Pumping from tube well for irrigation
i Index for crops
j Index for types of soil
k Crop coefficient for a particular month
m Regression coefficient
N Number of days of operation
NC Number of crops
pd Pumpage from tube well to drainage
PR Maximum preponing limit in terms of days
PT Maximum postponing limit in terms of days
RF Rainfall
RO1 , RO 2 , RO 3 Fraction factors for surface runoff loss from canal, irrigated area
and non-irrigated area respectively
Rs Pakistani rupees
RW Water available at river
sw Flow diverted to canal
t Index for month
TCA Total cropped area
TCP Total existing crop production
20 N.K. Garg, A. Ali r Agricultural Water Management 36 (1998) 1–21

TIA Total available area for irrigation

TS Total types of soil
TWC Maximum capacity of tube well
VPQ Value per quintal
WR Monthly water requirement of crop
x Sowing date of a crop
YPH Yield per hector
Z Objective function
a1 Fraction of canal water delivered to irrigated area
a2 Fraction of water delivered to the irrigated area, available for
consumptive use
bi Fraction of total crop production allotted to ith crop for the
existing production
gi Fraction of total crop production allotted to ith crop for the
maximum production
uj Fraction of cultureable command area under jth type of soil
h sw , h gw , h dw Efficiency of canal, tube well and drainage systems, respectively.
n tsw , n tgw , n tdw Ratio of peak to average demand of canal, tube well and drainage
systems in t th month, respectively

References
A Guide for Estimating Irrigation Water Requirements, 1984. Government of India, Ministry of Irrigation,
Water Management Division, New Delhi, India.
Billy, E., Gillett, 1976. Introduction to Operations Research. McGraw-Hill, New York.
Bredefoeft, J.D., Young, R.A., 1983. Conjunctive use of groundwater and surface water for irrigated
agriculture. Water Resour. Res. 19 Ž5., 1111–1121.
Chaudhry, M.T. et al., 1974. Optimal conjunctive use model for Indus Basin. J. Hydraulics Div. ASCE 100
ŽHY5., 667–687.
Doorenbos, J., Pruitt, W.O., 1977. Guidelines for predicting crop water requirements. Irrigation and Drainage
Paper No. 24. FAO, Rome.
Haimes, Y.Y., Dreizin, C., 1977. Management of groundwater and surface water via decomposition. Water
Resour. Res. 13 Ž1., 69–77.
Harza Engineering International—Water and Power Development Authority Pakistan, 1963. A program for
water and power development in Indus Basin in West Pakistan. An Appraisal of Resources and Potential
Development Report.
Harza Engineering International et al., 1991. Water sector investment planning study for Pakistan. Report
Water and Power Development Authority, 4 Vol., Lahore, Pakistan.
Irrigation and Agricultural Consultants Association, 1966. Program for the development of irrigation and
agriculture in West Pakistan, 23 Vol., Lahore, Pakistan.
Khepar, S.D., Chaturvedi, M.C., 1982. Optimal cropping and groundwater management. Water Resour. Bull.,
Am. Water Resour. Assn. 18 Ž4., 655–660.
Lakshminarayana, V., Rajagopalan, S.P., 1977. Optimal cropping pattern for basin in India. J. Irrigation
Drainage Div., ASCE 103 ŽIRI., 53–77.
Lieftinck, P. et al., 1968. Water and power resources of West Pakistan. A Study in Sector Planning, 3 Vol.,
Baltimore.
Nieswand, G.H., Granstron, M.L., 1971. A chance constrained approach to the conjunctive use of surface
waters and ground waters. Water Resour. Res. 7 Ž6., 1425–1436.
 N.K. Garg, A. Ali r Agricultural Water Management 36 (1998) 1–21 21

O’Mara, G.T., Duloy, J.H., 1984. Modelling efficient water allocation in a conjunctive regime—the Indus
Basin of Pakistan. Water Resour. Res. 20 Ž11., 1489–1498.
Paudyal, G.N., Gupta, A.D., 1990. Irrigation planning by multilevel optimization. J. Irrigation Drainage Div.,
ASCE 116 Ž2., 273–291.
Revelle, R., 1964. Report on land and water development in the Indus Plains. U.S. Govt. Printing Office,
Washington, DC.
Rydzewski, J.R., Rashid, A.H., 1981. Optimization of water resources for irrigation in East Jordan. Water
Resour. Bull., Am. Water Resour. Assn. 17 Ž3., 367–372.
Sir M. MacDonald and Partners et al., 1988. Lower Indus Region—Right Bank Master Plan. Report Water
and Power Development Authority, 11 Vol., Lahore, Pakistan.
Smith, D.V., 1970. Stochastic irrigation planning models. Report 3, East Pakistan Land and Water study,
Center for Population Studies, Harvard University, Cambridge, MA.
Smith, D.V., 1973. System analysis and irrigation planning. J. Irrigation Drainage Div., ASCE 99, 89–107,
No. IRI.
Tipton, Kalmbach, 1967. Regional plans, North Indus Plains. Report, Water and Power Development
Authority, Lahore, Pakistan.
Wanyoung, Yu., Haimes, Y.Y., 1974. Multilevel optimization for conjunctive use of groundwater and surface
water. Water Resour. Res. 10 Ž4., 625–636.