Irrigation and Drainage - 2011 - Darshana - SIMULATION AND OPTIMIZATION FOR IRRIGATION AND CROP PLANNING

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 11

IRRIGATION AND DRAINAGE

Irrig. and Drain. 61: 178–188 (2012)


Published online 9 May 2011 in Wiley Online Library (wileyonlinelibrary.com) DOI: 10.1002/ird.633

SIMULATION AND OPTIMIZATION FOR IRRIGATION AND CROP PLANNING†


DARSHANA1, ASHISH PANDEY1*, MANFRED OSTROWSKI2 AND R. P. PANDEY3
1
Department of Water Resources Development and Management, IIT Roorkee, Roorkee, India
2
T.U. Darmstadt, Germany
3
National Institute of Hydrology Roorkee, Roorkee, India

ABSTRACT
In this study, simulation and optimization models were assembled for the optimization of irrigation systems and their
operation. The simulation model CROPWAT was used for estimation of the crop water requirement, time and depth. The
evolutionary algorithm (GANetXL) was used for the optimal planning of cropping pattern, maximization of net benefits and
minimization of irrigation water requirements for the study area of Holeta catchment, Ethiopia. The study area encompasses
three command areas, i.e. farm A, farm B and Tsedey State Farm, and five different type of crops, i.e. potato, tomato, apple,
peach and winter wheat. The simulation results of the CROPWAT model illustrated that crop water requirement for apple was
highest (993 mm), followed by peach (908 mm), tomato (470 mm), potato (443 mm) and wheat (294 mm). The study reveals
that fruit crops have more crop water requirements than cereals. The results of the GANetXL show that when the cropped area
and water allocated was varied between extreme values, 23% of water can be saved. The total benefit from the study area can
be enhanced by USD 34 ha−¹ and can be helpful in improving the economic conditions of the farmers. Copyright # 2011 John
Wiley & Sons, Ltd.
key words: CROPWAT model; irrigation; optimization; simulation; genetic algorithm

Received 24 November 2009; Revised 22 February 2011; Accepted 23 February 2011

RÉSUMÉ
Dans cette étude, des modèles de simulation d’optimisation ont été assemblés pour optimiser les systèmes d’irrigation et leur
conduite. Le modèle de simulation CROPWAT a été utilisé pour l’estimation quantitative et temporelle des besoins en eau des
cultures. L’algorithme évolutionnaire (GANetXL) a été utilisé pour la planification optimale de la répartition des cultures, la
maximisation des bénéfices nets et la minimisation des besoins en eau d’irrigation pour la zone d’étude du bassin versant de
Holeta, en Ethiopie. La zone d’étude englobe trois mailles hydraulique, à savoir les fermes A et B et la ferme d’Etat de Tsedey;
cinq cultures différentes ont été considérées: pommes de terre, tomate, pomme, pêche et blé d’hiver. Les résultats des
simulations du modèle CROPWAT ont montré que les besoins en eau des cultures les plus élevés étaient ceux de la pomme
(989 mm), puis ceux de la pêche (912 mm), de la tomate (469 mm), des pommes de terre (443 mm) et du blé (293 mm).
L’étude confirme ainsi que les besoins en eau de l’arboriculture sont supérieurs à ceux de la céréaliculture. Les résultats de
l’GANetXL montrent qu’en faisant varier la superficie cultivée et l’eau allouée entre deux valeurs extrêmes, il existe un
optimum qui permet d’économiser 23% de l’eau d’irrigation et d’améliorer la marge nette de la zone d’études de USD 34 ha−¹,
améliorant ainsi la condition économique des agriculteurs. Copyright # 2011 John Wiley & Sons, Ltd.
mots clés: modèle CROPWAT; irrigation; optimisation; simulation; algorithme évolutionnaire

INTRODUCTION
Water allocation and its management for agricultural
purpose is a complex issue depending on social, environ-
* Correspondence to: Ashish Pandey, Department of Water Resources mental and political factors. It has generally been
Development and Management, IIT Roorkee, Roorkee‐247 667, India,
Uttarakhand, E‐mail: ashish_nerist@yahoo.co.in, ashisfwt@iitr.ernet.in performing far below its potential, particularly in drought‐

Simulation et optimisation de la planification de l'irrigation et des cultures. prone areas, making the level of agricultural production and

Copyright # 2011 John Wiley & Sons, Ltd.


SIMULATION AND OPTIMIZATION FOR IRRIGATION AND CROP PLANNING 179

irrigation benefit economically inadequate (Raju and irrigate and how much to irrigate. In the second phase the
Kumar, 1999). Efficient irrigation systems would play a relationship between yield and total water applied (rain +
key role in achieving the secured food supply of the world’s irrigation) was developed as per the criterion suggested by
population. Between 60 and 70% of all inland water Doorenbos and Kassam (1979). In the third phase the results
resources are actually used for irrigation purposes, with from the first and second phases were used for the
efficiencies ranging from 30 to 70%. It is obvious that optimization of irrigation system structure, such as type of
available water resources, although renewable, have an cropping pattern using multi‐objective optimization using
upper limit. It is being realized that the available water an evolutionary algorithm (GANetXL).
resources may decrease in coming years due to their
excessive exploitation. The feasible solutions to this
problem appear to be scientific improvement in irrigation
STUDY AREA
technology, including development of optimum irrigation
systems, i.e. engineering and agricultural infrastructure, and The Holeta catchment lies in the upper course of the Awash
optimum operation of such systems. Computer‐based basin, Ethiopia. It is a small catchment with an area of 119.0
simulation and optimization have a significant role to play km2. The loamy soil of the study area is the main cause of
in supporting management decisions in an environment erosion in the rainy season. The average altitude of the study
where the interacting processes are complex and vary area varies between 2400 and 3200 metres above mean sea
spatially and temporally, and where scheme managers use a level. The environment is seasonally humid with long‐term
rule of thumb to predict the effects of their decisions on a (1976–2005) average annual rainfall of 1055 mm, of which
scheme’s performance (Kuo et al., 2001). 85% is received between the months of June and September,
Optimization, simulation and simulation–optimization and mean maximum and minimum temperatures of 22.2
approaches are usually used for irrigation planning and and 6.1 °C, respectively. Wind speed in the Holeta catchment
management. The simulation approach include models of varies from 175 to 425 km per day. The mean relative
soil water balance (Liu et al., 1998; George et al., 2000; humidity varies from 84% in August to 55% in February.
Mateos et al., 2002), soil water dynamics (Singh and Singh, The town of Holeta is only major settlement in the
1997; Shang et al., 2004) and crop simulation (Rinaldi, catchment of the Holeta River, which is the capital of the
2001; Bergez et al., 2002). These simulation models were Wolmera‐Genet area and 45 km from Addis Ababa.
able to give a more detailed description of evapotranspira- The Holeta river tributary passes Holeta town about 1 km
tion and crop growth. For the optimization approach, to the east. It serves as the water supply for people and
irrigation was scheduled with dynamic programming (Rao cattle outside of the central town, which is supplied by
et al., 1988; Naadimuthu et al., 1999), nonlinear program- wells. The catchment has fertile soils (loam) and a high
ming (Ghahraman and Sepaskhah, 1997), etc., to obtain potential for water resource development. However,
maximum crop yield or profit. development in the valley is limited to a few hundred
The traditional optimizing irrigation planning model hectares of irrigation and a small milling plant due to a
attempts to obtain optimum values, thus satisfying the very high seasonal variation of water availability. Farmers
objective function and constraints. Moreover, traditional in this area depend exclusively on rain‐fed agriculture and
optimization models in irrigation planning have been of most crops are grown in the main rainy season.
extensive interest (Matanga and Marino, 1979; Paudyal and There are three different farms in the Holeta catchment,
Gupta, 1990; Raman et al., 1992; Singh et al., 1999). To i.e. farm A, farm B and Tsedey State Farm, and location of
maximize the gross benefit of a yield, subject to total water the study area is presented in Figure 1. Irrigation sector A is a
supply, Matanga and Marino (1979) developed an area‐ smallholding of about 300 ha (approximately 600 family
allocation model. Thereafter, to resolve the complex heads). The Holeta River is the main source of surface water
problem of irrigation management within a large heteroge- in the study area. Water is diverted from the Holeta River to
neous basin, Paudyal and Gupta (1990) applied a multi‐ the farms, passing through a 3 km contour canal having a
level optimization technique. Kuo and Liu (2003) developed maximum capacity of 150 l s−¹. There is no facility to store
a simulation and optimization model and applied it to an rain water in the rainy season for use in the dry season. The
irrigated area in Delta, Utah, to optimize the economic average monthly runoff (1975–1991) of the Holeta River
benefit, simulate the water demand, and search the related varies from 0.68 million cubic metres (MCM) in the month
crop area percentage with specified water supply and planted of January to 20.4 MCM in the month of August. The farms
area constraints. are poorly managed and constructed without any proper
In this study, the problem was divided into three phases. water control structure. Therefore, irrigation efficiency of
In the first phase the CROPWAT model was used for farms A and B is very low (30%). Water is distributed to
optimization of irrigation system operation, such as when to every farmer on a rotational basis. A village committee is

Copyright # 2011 John Wiley & Sons, Ltd. Irrig. and Drain. 61: 178–188 (2012)
180 D. DARSHANA ET AL.

Figure 1. Map of study area

established to coordinate farm activities and to negotiate evapotranspiration (Smith, 1992). The Penman–Monteith
with other entities about their share of water from the Holeta approach is a combination of both aerodynamic and
River. Cereals are produced by farmers under rain‐fed radiation terms as reported by FAO Paper No. 56, and the
agriculture from June to November for subsistence. Potatoes general equation is as follows:
and tomatoes are the dominant horticultural crops irrigated
0:408ΔðRn −GÞ þ γ Tþ273
900
u2 ðes −ea Þ
by the farmers, although several other crops are also ETo ¼ (1)
cultivated. Planting of potatoes and tomatoes takes place in Δ þ γð1 þ 0:34u2 Þ
the first week of January and February, respectively. Tsedey where ETo is the reference crop evapotranspiration
State Farm is run by a state‐owned enterprise for (mm day−1); Rn is the net radiation at the crop surface
horticultural development. It grows deciduous fruit trees, (MJ m−2 day−1); G is the soil heat flux (MJ m−2 day−1); T
mainly apples on 100 ha and peaches on 50 ha. The farm has is the mean daily air temperature at 2 m height (°C); u2 is
a well‐constructed intake and water control structure the wind speed at 2 m height (m s−1); es is the saturation
downstream of the mill and it is well managed. The vapour pressure (kPa); ea is the actual vapour pressure
irrigation efficiency of Tsedey State Farm is 50%. Irrigation (kPa) ; (es – ea) is the saturation vapour pressure deficit
sector B is the same as sector A except that the area brought (kPa); Δ is the slope vapour pressure curve (kPa °C); γ is
under irrigation is 100 ha (approximately 200 families). the psychometric constant (kPa °C) and 900 is the con-
Shortage of water is more pronounced in this farm than version factor for daily basis calculation.
elsewhere in the upstream area. There are periods when the For the calculation of effective precipitation, the USDA
available water is fully diverted by the upstream users, Soil Conservation method was used (Smith, 1991). This
leaving no water in the river for farm B users. method assumes that crops can use almost 60–80% of
precipitation up to 250 mm per month. Over 250 mm per
METHODOLOGY month, the crops benefit from only 10% of the total
precipitation. In other words as precipitation increases, its
CROPWAT model efficiency decreases. The empirical equation is as follows:
The CROPWAT model for Windows was used for the
estimation of crop water requirements. It is a program that 125−0:2Ptot
Peff ¼ Ptot  for Ptot < 250 mm (2)
uses the Penman–Monteith method to calculate crop 125

Copyright # 2011 John Wiley & Sons, Ltd. Irrig. and Drain. 61: 178–188 (2012)
SIMULATION AND OPTIMIZATION FOR IRRIGATION AND CROP PLANNING 181

Peff ¼ 125 þ 0:1 Ptot for Ptot > 250 mm (3) GIt ¼ ∑ci Pc  Ac  Yc (5)
where Peff is the effective rainfall (mm) and Ptot is the total
rainfall (mm). CPt ¼ ∑ci Cp  Ac þ ∑ci Cwt  Uc (6)

where Pc is crop price (US$ t−¹), Ac is the crop area for crop c
Input data requirement of the CROPWAT model
(ha), Yc is yield for crop c (t ha−¹), Cp = costs of production per
The CROPWAT model uses climatic, crop and soil data unit area (US$ ha−¹), Cwt = cost of irrigation water (m3), Uc is
for the calculation of the crop water requirement. The irrigation release (106 m3) for crop c. Table I shows the crop
meteorological data include: (1) maximum and minimum prices and cost of production per unit area.
temperature, (2) wind speed, (3) sunshine hours, (4)
relativity humidity and (5) rainfall. The climatic data of Objective Function 2 (OF2): Minimization of
the study area are taken from the CLIMWAT model. In this irrigation water requirement (MIR).
study, for estimation of the crop water requirement, five
crops, namely potato, tomato, apple, peach and winter OF2 ¼ MIR ¼ ∑ INðc; mÞ  Ac (7)
ðc; mÞ
wheat, are considered. The CROPWAT model contains
default data for potato, tomato and winter wheat. The crop where IN(c, m) is the irrigation water need for crop c in
parameters of apple and peach are not available in the month m (106 m3 ha−¹).
CROPWAT model. The crop parameters of apple and peach
can be created through the CROPWAT model interface. Constraints. The objective function is subject to the
The cropping pattern consists of planting date, crop following constraints (Equations 8 and 9).
parameters (including Kc value, stage days, root depth,
depletion fraction) and the area planted (0–100% of the total Area availability. The sum of all crop area is equal or
area). In this study, it was taken that apple, peach, potato, less then total farm area.
tomato and winter wheat crops were planted on 20 October,
28 February, 2 January, 5 February and 20 November ∑ Ac ≤ ∑ TA (8)
c
respectively. It was assumed that the crops were planted at
the same time. The CROPWAT model requires information where TA is the total area and Ac is the area of crop c.
on soil type such as: total available moisture, readily
available moisture and initial available moisture. The Water demand to water availability. Total irrigation
CROPWAT model contains default data for loamy soil. water needed in the irrigation area should not exceed total
water available for irrigation area:
MATHEMATICAL MODEL FORMULATION ∑ INðc;mÞ  Ac ≤ ∑ WA (9)
ðc;mÞ
The objective of the model was maximization of net benefits
(Equation 4) and minimization of irrigation water require- where WA is the total water available.
ments (Equation 7). These two competing objectives of the Decision variables are the amount of land in the irrigated
system are expressed as follows. area for growing crop c (ha), denoted as Ac. It is assumed
that the soil is homogeneous and irrigation efficiency is
Objective function 30% for farms A and B and 50% for Tsedey State Farm. It
is important to adjust crop patterns on an economic basis
Objective Function 1 (OF1): Maximization of net
and for optimal use of resources.
benefits (MNB). The maximization of net returns is
commonly aspired to by every decision maker in the
planning process. However, farmers are more biased towards Relationship between yield and total water applied
such objectives and always prefer crop planning which can Yaron and Bresler (1983) reported that dated response
provide them with greater financial returns. The net returns functions can be incorporated into optimization models
under different crops from the command areas of Holeta farm which provide guides regarding the timing and quantities of
A, farm B and Tsedey State Farm are to be maximized. These irrigation water applied under given conditions. Many
are obtained by deducting the cost of irrigation water, authors have proposed semi‐empirical water production
fertilizer and labour (CP) from the gross economic benefits functions that relate crop yield to the amount of evapo-
(GI) of crops and can be expressed as transpiration. It has been pointed out that an important
factor in irrigation water management is the timing of soil
OF1 ¼ MNB ¼∑nt¼1 ðGIt −CPt Þ (4) moisture stress imposed on a crop in relation to its

Copyright # 2011 John Wiley & Sons, Ltd. Irrig. and Drain. 61: 178–188 (2012)
182 D. DARSHANA ET AL.

Table I. Crop prices and cost of production

Potato Tomato Wheat Apple Peach

Production cost (US$ ha−¹) 152 364 606 1 460 1 460


Crop price (US$ t −¹) 30 45 40 91 91

Source: Woreda Ministry of Agricultural (2004).

physiological stage of growth. Different water production that encloses the current members but uses members on the
functions which take into account the developmental stage same front. The selection operator used is crowded
of a crop can be found in the literature (Vaux and Pruitt, tournament selection (Deb and Goldberg, 1989), where both
1983). One of the most currently used is the multiplicative children and parents in the same generation compete with
form of the Stewart formula proposed in the FAO each other. The members having a better rank are selected
methodology (Doorenbos and Kassam, 1979; Juan et al., automatically. If any two members have the same rank, the
1996). one having the greater crowding distance is selected. The
A relationship between crop yield and water applied was process of selection, recombination and mutation processes is
developed using the procedure as suggested by Doorenbos iterated until there is a sufficient number of generations. The
and Kassam (1979). The equation is as follows: detail process of GANetXL is explained in the manual of
    GANetXL 2006 (http://centres.exeter.ac.uk/cws/downloads/
Ya ETa
1− ¼ Ky 1− (10) cat_view/25‐software/42‐ganetxl).
Y max ETmax Prior to the random search process, the user must specify
four control factors: (1) number of generations, (2) chromo-
where Ky is the crop yield reduction factor, Ya, ETa is the
some size within one population, (3) probability of crossover
crop actual yield and evapotranspiration, respectively and
and (4) probability of mutation. The parameters of the NSGA‐
Ymax, ETmax are the maximum crop yield and potential
II optimization algorithm were set taking into account
evapotranspiration, respectively.
suggestions in the literature for specifying these parameters
(Deb, 2001). Multi‐point crossover and the Simple by Gene
Multi‐objective evolutionary algorithms (MOEA) mutator were chosen. A multi‐point crossover operator was
In this study, GANetXL (Evolutionary Optimization for used to achieve greater diversity in the generated chromo-
Microsoft Excel) was used for optimization of benefits and somes. The best value of the probability of crossover and
irrigation water requirements. It is a genetic algorithm add‐ mutation is considered to be equal to 0.8 and 0.01 respectively,
on for Microsoft Excel supporting single and multi‐ which is also obtained through a trial‐and‐error process.
objective optimization. The source code has been taken Goldberg (1989) revealed that satisfactory GA performance
from http://cws/ganetxl and then the problem was formu- requires high crossover and low mutation probabilities, as
lated in this code. GANetXL supports only one algorithm well as a moderate population. The total net profit is maxi-
(NSGA II) for multi‐objective problems and three (gener- mized while total irrigation water is minimized. The algorithm
ational, generational elitist, steady state) for a single is run for a maximum generation of 3000 and population size
objective problem. In this study, Non Sorting Genetic of 150. A random seed value of 1 was assumed.
Algorithm II (NSGA‐II) proposed by Deb et al. (2000) is
used to identify feasible solutions with trade‐offs between
RESULTS AND DISCUSSION
competing objectives (i.e. minimize irrigation requirements
and maximize net benefits). In this study, the CROPWAT model was used for the
NSGA‐II is a multi‐objective optimization algorithm estimation of reference crop evapotranspiration, crop water
based on non‐dominated sorting. In this algorithm the entire requirement and net irrigation requirement for different
population is sorted according to their non‐dominance over crops, namely apple, peach, potato, tomato and winter
the other solutions. Based on the number of dominated wheat. In this case the calculated crop water requirement is
solutions, a rank is assigned to a particular member of the a function of area. From the results it was found that crop
population. A front is comprised of members having the same water requirement increased from initial stage to crop
rank. The front consisting of rank 1 members is the current development stage (Table II). The peak amount of water is
best estimate of the Pareto front. To preserve diversity among required during the growing and crop development stages.
a population of the same generation, the crowding‐distance As seen from Table III, the reference crop evapotrans-
approach is used. The crowding distance is the largest cuboid piration (ETo) value varies greatly among these crops. The

Copyright # 2011 John Wiley & Sons, Ltd. Irrig. and Drain. 61: 178–188 (2012)
SIMULATION AND OPTIMIZATION FOR IRRIGATION AND CROP PLANNING 183

Table II. Estimation of crop water requirement (mm month−1)

Month Apple Peach Potato Tomato Wheat

January 69 90 58 – –
February 83 62 109 71 –
March 96 60 133 95 –
April 102 60 115 121 –
May 95 73 28 112 –
June 87 58 – 71 18
July 83 65 – – 43
August 85 76 – – 96
September 85 83 ‐ – 100
October 79 90 – – 37
November 64 95 – – –
December 65 96 – – –
Total 993 908 443 470 294

Table III. Actual cropping pattern (ha) and estimated crop water requirement of the study area

Crop name Area (ha) ET0 CWR Effective IWR


(mm) (mm) rainfall (mm)
Farm A Farm B Tsedey State Farm (mm)

Apple – – 100 1 236 993 514 475


Peach – – 50 1 236 908 271 641
Potato 150 50 – 490 443 100 343
Tomato 150 50 – 515 470 172 297
Wheat 150 50 – 785 294 206 86

perennial apple and peach crops indicated the highest value reaches up to 90 mm, and beyond 90 mm the relationship
of reference crop evapotranspiration (1236 mm) and potato between potato yields and applied water is curvilinear.
has the lowest (490 mm). Tomato and wheat gave reference Equation (12) presents the water–yield relationship for
crop evapotranspiration values of 515 and 785 mm the tomato crop. It can be seen from Figure 3 and Equation
respectively. The maximum/potential evapotranspiration or (12) that the tomato yield increases linearly with increase of
crop water requirement (ETcrop or ETm) values for the crops applied water up to 74 mm, and beyond 74 mm the relation
followed a similar pattern to those of reference evaporation. between tomato yields and applied water is nonlinear:
Apple recorded the maximum evapotranspiration value of

993 mm, followed by peach (908 mm), tomato (470 mm), Y ¼ − 0:0004x 2 þ 0:177x − 2:09 R 2 ¼ 0:71 (12)
potato (443 mm) and wheat (294 mm). As seen from
Table II, the results for apple and peach indicated similar In the case of winter wheat the following relationship
amounts of reference evapotranspiration but different (Equation 13) was obtained. The relationship presented in
amount of crop water requirement. This may be due to Figure 4 indicates that the drop in wheat yield increases
the higher value of the crop coefficient for apple. Using crop linearly with increase of applied water up to 72 mm, and
yield data for the period from 1975 to 1991 for the study beyond 72 mm the relation between tomato yields and
area, water–yield relationships for potato, tomato, winter applied water is also nonlinear.
wheat, apple and peach were developed and are presented in

Figures 2–6. Y ¼ − 0:0003x2 þ 0:115x þ 6:68 R2 ¼ 0:78 (13)
The following regression equation represents the water–
yield relationship for the potato crop: A similar relationship could be obtained for yield of apple
and water used as shown in Equation (14) and Figure 5. It

Y ¼ − 0:0005x2 þ 0:178x þ 2:76 R2 ¼ 0:58 (11) can be seen from Figure 5 that the apple yield increase is
linear with increase of applied water up to 108 mm, and
It can be seen from Figure 2 and Equation (11) that the beyond that value the relation between apple yields and
potato yield increases linearly when application of water applied water is nonlinear.

Copyright # 2011 John Wiley & Sons, Ltd. Irrig. and Drain. 61: 178–188 (2012)
184 D. DARSHANA ET AL.

Figure 5. Relationship between apple yield (t ha−1) and total water


−1
Figure 2. Relationship between potato yield (t ha ) and total water use (mm)
applied (mm)

Figure 3. Relationship between tomato yield(t ha−1) and total water


applied (mm) Figure 6. Relationship between peach yield (t ha−1) and total water
applied (mm)

optimum water application at different stages of crop


development. However, excess application of irrigation water
beyond the optimum value may lead to reduction of crop yield.
The irrigation water requirement was highest for peach
(641 mm) and lowest for wheat (86 mm). The irrigation
water requirement for apple, potato and tomato crops was
475, 343 and 297 mm respectively. This shows that there is
more demand for soil moisture by peach and apple than by
the other three crops which had comparable irrigation water
requirements. Apple has a higher crop water requirement
Figure 4. Relationship between wheat yield (t ha−1) and total water than peach but a lower irrigation water requirement. This is
applied (mm)
due to the occurrence of more effective rainfall during the
peak water requirement period of peach.
 The GANetXL algorithm for determining the optimal
Y ¼ − 0:0004x 2 þ 0:205x − 0:38 R 2 ¼ 0:80 (14) cropping pattern which can maximize the benefits and
In the case of the peach crop the relationship is linear minimize the irrigation water requirement was used in this
with increase of applied water up to 131 mm, and beyond study. The total cultivated area was kept constant. The areas
131 mm the relation is nonlinear. The water–peach crop of the different crops were changed to discover the optimal
yield relationships are shown in Figure 6 and Equation (15): cropping pattern. But the limit of maximum and minimum
crop area was fixed. The actual cropping pattern in the study

Y ¼ − 0:0006x2 þ 0:283x− 8:16 R 2 ¼ 0:71 (15) area of the Holeta catchment is presented in Table III. The
MOGA (multi‐objective genetic algorithm) was applied
The above water– yield relationship for various crops with a population of 150 chromosomes and 3000 genera-
confirms that the optimum crop yield can be obtained with tions, a multi‐point crossover with probability of 0.8 and

Copyright # 2011 John Wiley & Sons, Ltd. Irrig. and Drain. 61: 178–188 (2012)
SIMULATION AND OPTIMIZATION FOR IRRIGATION AND CROP PLANNING 185

simple By Gene mutation with probability of 0.01. Figure 7 0.39 × 106 m3 and after optimization was 1.92 × 106,
shows the optimal cropping area of different crops which 7.24 × 106 and 0.77 × 106 m3 respectively for potato, tomato
can maximize net benefits and minimize the irrigation and winter wheat for farm A. It was found that with the
requirement for farm A, farm B and Tsedey State Farm. It optimal cropped area, total benefit increased by an amount
was found that the area increased for those crops having a of US$140 000 more than the actual case, and the optimum
low irrigation demand and decreased for those crops with a irrigation water requirement is less than the actual case
high irrigation demand. The results of the optimization from the whole of farm A. It was concluded that after
problem are discussed below. optimization the increase in net benefits was US$69 ha−¹,
and the decrease in irrigation requirement was about 25%
Farm A from the whole of farm A.
Figure 8 shows the Pareto optimal set for the crop
Farm B
planning model of two objective functions (maximizing
total net benefits and minimizing irrigation water) when the The Pareto optimal front for the crop planning model of
crop area is allowed to change between minimum and two objective functions (maximizing benefits and minimiz-
maximum. It was found that the solution is well distributed ing irrigation water requirement) is shown in Figure 9 when
along the Pareto front. Table IV shows the optimized the cropping area is allowed to change between minimum and
benefits and irrigation water requirement corresponding to maximum. It was found that the solution is well distributed
the optimal area of potato, tomato and winter wheat of farm along the Pareto front. Table IV shows the optimized benefits
A. It can be seen from Table IV that before optimization and irrigation water requirement corresponding to the
the total benefits for potato, tomato and winter wheat were optimal area of potato, tomato and winter wheat of farm B.
US$86 500, US$153 000 and US$64 100 respectively, and It can be seen from Table IV that before optimization the
after optimization were US$32 300, US$250 000 and US total benefits for potato, tomato and winter wheat were
$161 000 respectively. The irrigation water requirement US$28 900, US$51 100 and US$20 200 respectively and
before optimization was 5.14 × 10 6 , 4.45 × 10 6 and after optimization were US$5 781, US$90 100 and US$39 600

Figure 7. Actual and optimal area of different crops in the study area

Figure 8. Pareto optimal set for crop planning model when maximizing total benefits and minimizing irrigation water from farm A

Copyright # 2011 John Wiley & Sons, Ltd. Irrig. and Drain. 61: 178–188 (2012)
186 D. DARSHANA ET AL.

Table IV. Actual and optimized benefits and irrigation water for different crops

Farm Crop Benefits before Benefits after Irrigation demand before Irrigation demand after
name optimization (US$) optimization (US$) optimization (106 m3) optimization (106 m3)

Farm A Potato 86 500 32 300 5.14 1.92


Tomato 153 000 250 000 4.45 7.24
Wheat 64 100 161 000 0.39 0.77
Farm B Potato 28 900 5 780 1.71 0.34
Tomato 51 100 90 100 1.48 2.61
Wheat 20 200 39 600 0.13 0.25
Tsedey State Apple 131 000 190 000 2.85 3.99
Farm Peach 61 600 12 300 1.89 0.38

Figure 9. Pareto optimal set for crop planning model when maximizing total benefits and minimizing irrigation water from farm B

respectively. The irrigation water requirement before opti- net benefits and minimizing irrigation water) when the crop
mization was 1.71 × 106, 1.48 × 106 and 0.13 × 106 m3 and area is allowed to change between minimum and maximum.
after optimization was 0.34 × 10 6 , 2.61 × 10 6 and It was found that the solution is well distributed along the
0.25 × 106 m3 respectively for potato, tomato and winter Pareto front. Table IV shows the optimized benefits and
wheat for farm B. It was found that with the optimal cropping irrigation water requirement corresponding to the optimal
area, the increase in total benefit was US$35 300 more than area of apple and peach. It can be seen from Table IV that
the actual case, and the optimum irrigation water requirement before optimization the total benefits were US$131 000 and
is less than the actual case from farm B. It was concluded that US$61 600 for apple and peach and after optimization were
after optimization the increase in net benefits was US$23 ha−¹ US$190 000 and US$12 300. The irrigation water
and the decrease in irrigation requirement was about 26% requirement before optimization was 2.85 × 10 6 and
from the whole of farm B. 1.89 × 106 m3 and after optimization was 3.99 × 106 and
0.38 × 106 m3 for apple and peach for the Tsedey State Farm.
Tsedey State Farm It was found that with the optimal cropped area, total benefit
increased by an amount of US$9 700 more than the actual
Figure 10 shows the Pareto optimal set for the crop case, and optimum irrigation water requirement is less than
planning model of two objective functions (maximizing total
the actual case from the whole of Tsedey State Farm. It was
concluded that after optimization the increase in net benefits
was US$65 ha−¹ and the decrease in irrigation requirement
was about 8% from the whole of Tsedey State Farm.
It can be seen from Figure 11 that with the optimal
cropped area, total benefit increased by an amount of US
$185 000 than that of actual from the entire study area.
Further, Figure 12 reveals that the optimum irrigation water
requirement is less than the actual case. Thus, it was found
that the optimum distribution of the cropped area and
optimum irrigation application may increase net benefits in
Figure 10. Pareto optimal set for crop planning model when maximizing the order of US$34 ha − ¹ and reduce the irrigation
total benefits and minimizing irrigation water from Tsedey State Farm requirement by 23%.

Copyright # 2011 John Wiley & Sons, Ltd. Irrig. and Drain. 61: 178–188 (2012)
SIMULATION AND OPTIMIZATION FOR IRRIGATION AND CROP PLANNING 187

REFERENCES
900
Objective Function 1 781
Benefits (103US$)

750 Bergez JE, Deumier JM, Lacroix B, Leroy P, Wallach D. 2002. Improving
596
600 irrigation schedules by using a biophysical and a decisional model.
450 European Journal of Agronomy 16: 123–135.
Deb K. 2001. Multi‐Objective Optimization using Evolutionary Algorithms.
300
John Wiley and Sons: Chichester, UK.
150
Deb K, Goldberg DE. 1989. An investigation of niche and species for-
0 mation in genetic function optimization. In Proceedings of the 3rd ICGA.
Actual Optimal
Morgan Kaufmann. George Mason University: United States; 42–50.
Deb K, Agrawal, S, Pratap A, Meyarivan T. 2000. A Fast and Elitist
Figure 11. Total benefits from the study area Genetic Algorithm NSGAII. Technical Report 2000001. Indian Institute
of Technology, Kanpur, India.
Doorenbos J, Kassam AH. 1979. Yield response to water. Irrigation and
Drainage Paper No. 33. FAO: Rome, Italy; 193 pp.
George BA, Shende SA, Raghuwanshi NS. 2000. Development and testing of an
irrigation scheduling model. Agricultural Water Management 46: 121–136.
Ghahraman B, Sepaskhah AR. 1997. Use of a water deficit sensitivity
index for partial irrigation scheduling of wheat and barley. Irrigation
Science 18: 11–16.
Goldberg DE. 1989. Genetic algorithms. In Search, Optimization and
Machine Learning. Addison‐Wesley; New York.
Juan JA, Tarjuelo JM, Valiente M, and Garcia P. 1996. Model for optimal
cropping patterns within the farm based on crop water production
function and irrigation uniformity. I. Development of a decision model.
Agricultural Water Management 31: 115–143.
Figure 12. Total irrigation requirement Kuo SF, Liu CW. 2003. Simulation and optimization model for irrigation
planning and management. Hydrological Processes 17: 3141–3159.
Kuo SF, Liu CW, Merkley GP. 2001. Application of the simulated
CONCLUSIONS annealing method to agricultural water Resource management. Journal
of Agricultural and Engineering Research 80 (1): 109–124.
The following conclusions have been drawn from the study. Liu Y, Teixeira JL, Zhang HJ, Pereira LS. 1998. Model validation and crop
The simulated results from the CROPWAT model show that coefficients for irrigation scheduling in the North China plain.
crop water requirement for apple was highest (993 mm), Agricultural Water Management 36: 233–246.
followed by peach (908 mm), tomato (443 mm), potato Matanga GB, Marino MA. 1979. Irrigation planning: 1. Cropping pattern.
Water Resources Research 15: 672–678.
(470 mm) and wheat (294 mm). The water–yield relation-
Mateos L, Lopez‐Cortijio I, Sagardoy JA. 2002. SIMIS: the FAO decision
ships for potato, tomato, winter wheat, apple and peach support system for irrigation scheme management. Agricultural Water
were developed. The simulated crop water requirement Management 56: 193–206.
from the CROPWAT model and the developed water–yield Naadimuthu G, Raju KS, Lee ES. 1999. A heuristic dynamic optimization
relationship were used to optimize the cropping pattern of algorithm for irrigation scheduling. Mathematical and Computer
Modelling 30: 169–183.
different crops in the study area using the GANetXL
Paudyal GN, Gupta AD. 1990. Irrigation planning by multilevel
optimization algorithm. Analysis of the results from the optimization. Journal of Irrigation and Drainage Engineering, ASCE
GANetXL algorithm showed approximately 23% of water 116: 273–291.
saving when the crop areas were allowed to change between Raju KS, Kumar DN. 1999. Multicriterion decision making in irrigation
minimum and maximum. The net benefit from the area was planning. Agricultural Systems 62: 117–129.
US$34 ha −¹. The results of this study might be helpful in Raman H, Mohan S, Rangacharya NCV. 1992. Decision support for crop
planning during droughts. Journal of Irrigation and Drainage
achieving optimal benefits from the studied farms in Engineering, ASCE 118: 229–241.
particular and from farms all over the globe. Rao NH, Sarma PBS, Chander S. 1988. Irrigation scheduling under a
limited water supply. Agricultural Water Management 15: 165–175.
Rinaldi M. 2001. Application of EPIC model for irrigation scheduling of
ACKNOWLEDGEMENTS sunflower in Southern Italy. Agricultural Water Management 49: 185–196.
Shang SH, Li XC, Mao XM, Lei ZD. 2004. Simulation of water dynamics
The financial support provided by the DAAD (Deutscher and irrigation scheduling for winter wheat and maize in seasonal frost
Akademischer Austausch Dienst) to conduct a part of the areas. Agricultural Water Management 68: 117–133.
research work in the TU Darmstadt, Germany, is gratefully Singh R, Singh J. 1997. Irrigation planning in wheat (Triticum aestivum)
acknowledged. The authors would also like to thank the under deep water table conditions through simulation modeling.
Agricultural Water Management 33: 19–29.
anonymous referees for contributing insightful remarks and Singh R, Refsgaard JC, Y de L. 1999. Application of irrigation
useful suggestions, which led to a substantially improved optimization system (IOS) to a major irrigation project in India.
manuscript. Irrigation and Drainage Systems 13(3): 229–248.

Copyright # 2011 John Wiley & Sons, Ltd. Irrig. and Drain. 61: 178–188 (2012)
188 D. DARSHANA ET AL.

Smith M. 1991. CROPWAT: Manual and Guidelines. FAO of UN: Rome, Yaron D, Bresler E. 1983. Economics analysis of on farm irrigation using
Italy. response functions of crops. In Advances in Irrigation, vol. 2, Hillel D
Smith M. 1992. CROPWAT: a Computer Program for Irrigation Planning (ed.). Academic Press: New York; 223–255.
and Management. Food and Agriculture Organization of the United http://centres.exeter.ac.uk/cws/downloads/cat_view/25‐software/42‐ganetxl
Nations: Rome, Italy. http://centres.exeter.ac.uk/cws/technology/20‐geneticalgorithms/76‐ganetxl
Vaux HJ, Pruitt WO. 1983. Crop water production functions. In Advances http://www.fao.org/nr/water/infores_databases_climwat.html
in Irrigation, vol. 2, Hillel D (ed.). Academic Press: New York; 61–79. www.kimberly.uidaho.edu/ref‐et/fao56.pdf

Copyright # 2011 John Wiley & Sons, Ltd. Irrig. and Drain. 61: 178–188 (2012)

You might also like