You are on page 1of 9

Genetic diversity in chestnuts of Kashmir valley

Pak. J. Agri. Sci., Vol. 51(3), 639-647; 2014
ISSN (Print) 0552-9034, ISSN (Online) 2076-0906
http://www.pakjas.com.pk

MODELING GROWTH, DEVELOPMENT AND SEED-COTTON YIELD FOR
VARYING NITROGEN INCREMENTS AND PLANTING DATES USING DSSAT
Aftab Wajid1,*, Ashfaq Ahmad1, Manzoor Hussain2, Muhammad Habib ur Rahman1,
Tasneem Khaliq1, Muhammad Mubeen1, Fahd Rasul1, Usman Bashir1, Muhammad Awais1,
Javed Iqbal3 Syeda Refat Sultana4 and Gerrit Hoogenboom5
1

Agro-Climatology Lab., Department of Agronomy, University of Agriculture, Faisalabad, Pakistan; 2Plant Breeding
& Genetics Division, Nuclear Institute for Agriculture and Biology (NIAB) Faisalabad, Pakistan; 3Cotton Research
5
Institute, Sahiwal, Pakistan; 4University College of Agriculture, University of Sargodha; AG Weathernet,
Washington State University, USA
*
Corresponding author's e-mail: aftabwajid@hotmail.com

CSM-CROPGRO-Cotton Model under DSSAT V 4.0.2.0 has been extensively tested and validated in many studies, mainly
in United States. The objective of this study was to test and validate this model in three cotton growing regions of Pakistan
(Faisalabad, Multan and Sahiwal) for dynamic simulation of development, growth and seed cotton yield of four cotton
cultivars (CIM-496, CIM-506, NIAB-111 and SLH-284) at varying nitrogen increments (50, 100, 150 and 200 kg ha-1) sown
at different timings (20 May and 10 June). The model was first calibrated with data (phenology, biomass, LAI, and yield
components) collected during 2009 at all locations against the best performing treatment May 15 sowing, cv. CIM-496 and
200 kg N ha-1 (D1V1N4) in field trials. The model was then tested with data recorded against remaining thirty-one treatments
for all locations. Similarly, the data of year 2010 was used for validation. The simulated values of crop phenology (days to
anthesis and maturity) by the model were reliable with the recorded data, with root mean square error (RMSE) less than 2
days during both years. Although RMSE values for LAI approached higher than 1 in many of the treatments, these values for
total dry matter and seed-cotton yield were reasonably good (367 to 497 kg ha-1and 122 to 227 kg ha-1, respectively). There is
a dire need to assess impact of climate variation on seed cotton yield under various climatic regions of Pakistan to ensure
fiber quality and yield in future.
Keywords: Gossypium hirsutum L., CSM-CROPGRO-Cotton, simulation, crop growth, phenology
INTRODUCTION
Cotton is the most important cash crop and plays a vital role
in the economy of Pakistan. It accounts for 7.8% of value
added in agriculture and 1.6% of GDP (Government of
Pakistan, 2012). The cotton plant due to its narrow range of
ecological adaptability is very much influenced by the
climatic conditions and various agronomic factors, such as
sowing time, genotypes and nitrogen fertilization. A great
number of studies have been conducted on the effect of these
factors on growth and yield of seed-cotton in Pakistan
(Soomro et al., 2001; Akhtar et al., 2002; Ali et al., 2004;
Arshad et al., 2007; Jawdad et al., 2012; Nazar et al., 2012;
Shabbir et al., 2012). However, to study the interaction of all
these management strategies in different agro-ecological
conditions needs long and costly field experiments.
Crop growth models can assist in the synthesis of research
understandings about the interaction of genetics,
physiological and the environmental interaction across
disciplines and organization of data and are important tools
for agronomic management strategy evaluation (Jones et al.,
2003; Hoogenboom et al., 2004; Boote et al., 2010, Wajid et

al., 2013; Mubeen et al., 2013). During the last decade, these
models have been used extensively in agriculture to simulate
crop responses to different abiotic factors. Recently many
mechanistic simulations have been reported on both the
development and yield of cotton crop from sowing to
maturity in response to nonspecific site environment (Hima
et al., 2004; Kakni et al., 2005). The cotton boll maturation
period module in these models took solar radiation and N
nutrition factors into account in addition to temperature and
variety maturity profile. Crop growth models take
parameters in consideration like cultivar characteristics,
minimum temperature, maximum temperature, solar
radiation and crop management’s factors (Mahamood et al.,
2003; Hoogenboom et al., 2011). Li et al. (2009) used the
new semi empirical model to simulate the cotton leaf and
boll nitrogen concentration and also worked on direct
indicator of nitrogen fertilizer effect on growth, development
and cotton seed. The Cropping System Model (CSM)CROPGRO-Cotton model is part of the suite of crop
simulation models that encompass the Decision Support
System for Agro technology Transfer (DSSAT) (Jones et al.,
2003; Hoogenboom et al., 2004). The model simulates

Wajid, Ahmad, Hussain, Rahman, Khaliq, Mubeen, Rasul, Bashir, Awais, Iqbal, Sultana & Hoogenboom
growth, development, and yield of cotton for different
weather and soil conditions and management practices (Ortiz
et al., 2009). CROPGRO (DSSAT) is one of the first
packages that modified weather simulation generators/or
introduced a package to evaluate the performance of models
for climate change situations (Murthy, 2004).
CROPGRO-Cotton is a newly developed crop model and
has many parameters (Pathak et al., 2009). In Pakistan,
studies on the use of crop models in cotton for evaluation
and validation have not been reported. The present study
was, therefore, conducted with the objectives to evaluate
CROPGRO-Cotton model for growth, development and seed
cotton yield, and to validate model under various climatic
conditions with independent set of data.
MATERIALS AND METHODS
Two field experiments at three locations Post Graduate
Agricultural Research Station (PARS) Faisalabad (31.26°N,
73.06°E), Cotton Research Station (CRS) Sahiwal (31.58°N,
72.20°E) and Central Cotton Research Institute (CCRI)
Multan (30.12°N, 71.26°E) were conducted to predict the
phenology, growth and development of cotton (Gossypium
hirsutum L.) productivity in Punjab, during 2009 and 2010.
The experiment was conducted in split-split plot design in
both seasons at three locations. There were two sowing dates
D1=20 May and D2=10 June in main plots, four cultivars
Viz. V1=CIM-496, V2=CIM-506, V3=NIAB-111and
V4=SLH-284 in sub plots and four nitrogen levels N1=50 kg
ha-1, N2=100 kg ha-1 (farmers’ practice, control), N3=150 kg
ha-1, N4 =200 kg ha-1 in sub-sub plots. Each experiment
consisted of three replications with a net plot size of 3x10m.
Crop husbandry: The crop husbandry operations during
both the seasons were kept normal according to the
recommendations of Agriculture Department except
varieties, sowing dates and nitrogen levels which were
applied according to the treatments under study. The crop
was sown on due dates i.e. on 20th, May and 10th June
uniformly at 75 cm apart, using bed-furrow method with
seed rate 25 kg ha-1. Thinning was completed after crop
emergence to maintain uniform plant-to-plant distance of 30
cm. Nitrogen was applied in the form of urea in two splits,
50 kg ha-1 at sowing, in first split to all experimental units
and remaining in second split according to treatments at
flower initiation before15th August in both seasons. All other
agronomic practices such as irrigation, weeding, plant
protection measures and earthing up etc. were kept normal
and uniform for all the treatments.
Crop growth modeling: Field data collected from the
experiments during 2009 and 2010 growing seasons was
used for calibration and validation of CROPGRO-Cotton

model, respectively. Standard meteorological, soil, plant
characteristic and crop management data were obtained for
each site and used as input data for the model. Decision
Support System for Agro technology Transfer (DSSAT) was
used for estimation of crop genetic coefficients using
sensitivity analysis selecting the best treatment
simultaneously at three sites. The model was run using
experimental data of year 2009 for calibration and for
genetic co-efficient calculation but the validity of the model
was assessed by using the independent set of data recorded
during year 2010 with same set of crop genetic coefficients.
Model calibration and evaluation: Calibration is a process
of adjusting some model parameters to our own conditions.
It is also necessary for getting genetic co-efficient for new
cultivars used in modeling study. So the model was
calibrated with data (that included phenology, biomass, LAI,
and yield components) collected during 2009 at all locations
against treatment May 15 sowing, cv. CIM-496 and 200 kg
N ha-1 (D1V1N4) that performed best in field trials. Cultivar
coefficients successively started from CSDL (critical short
day length) and PPSEN slope of the relative response to
development to photoperiod with time to PODOUR, the time
required for cultivar to reach final pod load under optimal
conditions (Photo thermal days). Fifteen coefficients control
the phenology, growth and seed cotton yield (Hoogenboom
et al., 1994) in CROPGRO-Cotton model. To select the most
suitable set of coefficients, an iterative approach was used.
Calculated coefficients for four cotton cultivars and their
detailed descriptions are given in Table 1. To check the
accuracy of the model simulations it was run with data
recorded against remaining thirty-one treatments for all
locations. The data on phenology, development and growth
for year 2010 was used for validation of CROPGRO-Cotton
model. During all this process available observed data on
crop phenology (flowering and maturity date), crop growth
(leaf area index and total dry matter production) and seed
cotton yield was compared with simulated values using same
genetic coefficients. Simulation performance was evaluated
by calculating different statistical indices like root mean
square error (RMSE) (Wallach and Goffinet, 1989) and
mean percentage difference (MPD) across all locations. For
individual treatments error (%) between simulated and
observed values were calculated. These measurements were
calculated as under.

640

Simulating cotton yield using CROPGRO model
Table 1. Calculated Genetic coefficients for four cotton cultivars
ECO #

PA0001
PA0002
PA0003
PA0004

VRNAME CSDL PPSEN EMFL
1
2
3
CIM - 496
23
0.01
49.1
CIM - 506
23
0.01
49.1
NAIB-111
23
0.01
49.0
SLH - 284
23
0.01
40.0

FLSH
4
18.0
18.0
18.1
18.2

FLSD
5
18.0
18.0
18.1
18.2

SD- FL-LF LFMAX SLAVR SIZLF XRFT WTPSD SFDUR SDPDV PODUR
PM
6
7
8
9
10
11
12
13
14
15
59.7 99.99
3.99
188
265
0.850
0.183
41.9
27
3.9
60.9 90.00
4.00
188
266
0.850
0.189
38.0
27
3.8
60.0 99.99
4.09
188
287
0.840
0.188
39.0
27
3.9
66.6 110.99
3.99
189
290
0.780
0.180
36.0
27
3.8

CSDL= Critical Short Day Length below which reproductive development progresses with no day length effect (for short day plants)
PPSEN = Slope of the relative response of development to photoperiod with time (positive for short day plants) (1/hour)
EM-FL = Time between plant emergence and flower appearance (R1) (photo thermal days)
FL-SH = Time between first flower and first pod (R3) (photo thermal days)
FL-SD = Time between first flower and first seed (R5) (photo thermal days)
SD-PM = Time between first seed (R5) and physiological maturity (R7) (photo thermal days)
FL-LF = Time between first flower (R1) and end of leaf expansion
LFMAX = Maximum leaf photosynthesis rate at 30 C, 360 vpm CO 2, and high light (mg CO2/m2-s)
SLAVR = Specific leaf area of cultivar under standard growth conditions (cm2/g)
SIZLF= Maximum size of full leaf (three leaflets) (cm2)
XFRT = Maximum fraction of daily growth that is partitioned to seed + shell
WTPSD = Maximum weight per seed (g)
SFDUR = Seed filling duration for pod cohort at standard growth conditions (photothermal days)
SDPDV = Average seed per pod under standard growing conditions (#/pod)
PODUR = Time required for cultivar to reach final pod load under optimal conditions (photothermal days)

Where Pi and Oi are the predicted and observed values for
studied treatments, respectively and n is the number of
observations. Linear regression analysis between predicted
and observed seed cotton yield and total dry matter at
harvest was done to evaluate the validity of model at various
experimental sites. Model performance improves as R2 value
approaches to one while RMSE, MPD and error proceed to
zero.
RESULTS
The CROPGRO-Cotton model performed well in simulating
crop phenology (days to flowering and maturity), crop
growth (LAI and TDM) and final crop yield (seed-cotton
yield) for three sites based on the estimation of the cultivar
coefficients. The model performed equally well with the
same set of Genetic coefficients for crop phenology, crop
growth and seed cotton yield. The corresponding simulation
results are described as below.
Days to anthesis: The pooled data of three sites presented in
Table 2 shows that model simulated days to anthesis one day
higher in both planting dates. At low fertilizer rate (50 kg N
ha-1) days taken to flowering were delayed than crop getting
nitrogen @200 N kg ha-1. Similarly the model showed one
day higher in anthesis than observed in all the cultivars.
According to model simulations crop reached flowering
stage 62-66 days after sowing in all treatments during 2009

and the observed days ranged from 61-65 days which
indicated that model was fit and worked well under our
environmental conditions. Efficacy of the CROPGROCotton model shows that error % for individual treatments
ranged from 1.54 to 1.64 for observed and simulated days to
flowering. Figure 1 shows fitness of the data between
observed and simulated data during year 2009. Similarly,
validation of model during year 2010 shows that crop
reached at anthesis stage in 60-67 days after sowing. The
observed values ranged from 59-66, closer to simulated
which depicts the usefulness of model with independent set
of data. Coefficient of determination (R2), goodness of the
model had high value of 0.96, simulated and observed values
was closer to 1.1 line, showed that model simulated anthesis
dates very well as it was displayed in the field (Fig. 1).
Days to maturity: The data presented in Table 3 reveals that
model simulated 2 days more for days to maturity than the
observed ones in both the plantings. The May sown crop
matured in 173 days whereas June sown matured in 151
days according to model simulations. The observations for
May and June sown crop for maturity were 171 and 149
days, respectively in year 2009. The two nitrogen levels
showed simulation of one day higher. Similarly all the
cultivars except NIAB-111 (difference of two days) showed
a difference of one day in simulated and observed. Root
mean square error ranged between 0.98-1.54 with MPD of
0.86 during this year.

641

Wajid, Ahmad, Hussain, Rahman, Khaliq, Mubeen, Rasul, Bashir, Awais, Iqbal, Sultana & Hoogenboom
Model evaluation for second year also remained satisfactory
for early and late planting in which prediction remained 2
days and 1 day more in early vs late plantings. The early
sown crop matured in 175 days vs late sown with 149 days
according to model simulations with corresponding observed
values of 173 days and 148 days in 2010. Cultivar CIM-496
and NIAB-111 gave same trend for days taken to maturity
between observed and simulated (161 vs. 163) and other two
cultivars CIM-506 and SLH-284 also took same number of
days to maturity (160 vs. 162). Mean percent difference
during 2010 was 1.16 showing acceptable limit. Simulated

and observed maturity dates were very close to 1.1 line for
both the years having higher values of R2 (0.99) showed the
goodness of the model during evaluation and validation
(Fig. 2).
Leaf area index (LAI): During 2009, model prediction for
earlier sown crop was higher (3.82) than late sown with
canopy development of 3.31. Model predicted higher (4.42)
LAI @ 200 kg N ha-1 than at low rate of 50 kg N ha-1 (3.18).
As regards cultivars, SLH-284 produced higher leaf area
index over rest of the cultivars. RMSE ranged from 0.83 to
1.50 during this year (Table 4).

Table 3. Comparison between simulated and observed days to maturity at different nitrogen levels, sowing dates
and cotton cultivars during year 2009 and 2010
Treatment
Calibration (2009)
Validation (2010)
Obs.
Sim.
Error (%)
RMSE
Obs.
Sim.
Error (%)
RMSE
May Sown
171
173
1.17
1.32
173
175
1.16
1.31
June Sown
149
151
1.34
1.14
148
149
0.68
1.33
N 50 kg ha-1
163
164
0.61
1.21
162
164
1.23
1.42
N 200 kg ha-1
157
158
0.64
1.54
158
160
1.27
1.34
CIM-496
160
161
0.63
1.07
161
163
1.24
1.17
CIM-506
161
162
0.62
0.98
160
162
1.25
1.09
NIAB-111
160
162
1.25
1.21
161
163
1.24
1.19
SLH-284
159
160
0.63
1.42
160
162
1.25
1.24
MPD
0.86
1.16
68
Days to anthesis (2009)

Days to anthesis (2010)

Observed

66

64

62

60

58
58

60

62

64

66

68
58

60

62

64

66

68

Simulated
Simulated
Figure 1. Relationship between simulated and observed days to anthesis for cotton cultivars during growing
season of 2009 and 2010

642

Simulating cotton yield using CROPGRO model
Table 4. Comparison between simulated and observed leaf area index at different nitrogen levels, sowing dates and
cotton cultivars during year 2009 and 2010
Treatment
Calibration (2009)
Validation (2010)
Obs.
Sim.
Error (%)
RMSE
Obs.
Sim.
Error (%)
RMSE
3.81
3.82
0.17
1.22
3.41
3.88
13.88
0.57
May Sown
3.17
3.31
4.63
0.95
3.03
3.18
4.96
1.20
June Sown
3.09
3.18
2.91
1.26
2.68
3.10
15.80
0.93
N 50 kg ha-1
4.34
4.42
1.69
1.13
4.07
4.18
2.62
1.10
N 200 kg ha-1
3.45
3.47
0.68
1.49
2.88
3.20
11.00
0.91
CIM-496
3.45
3.43
-0.68
1.18
3.09
3.25
4.96
0.68
CIM-506
3.52
3.54
0.66
1.50
3.23
3.43
6.09
1.05
NIAB-111
3.56
3.66
2.81
0.83
3.31
3.41
3.12
1.28
SLH-284
1.61
7.80
MPD

180
Days to Maturity (2009)

Days to Maturity (2010)

Observed

170

160

150

150

160

170

180

Simulated

150

160

170

180

Simulated

Figure 2. Relationship between simulated and observed days to maturity for cotton cultivars during growing
season of 2009 and 2010
When CROPGRO-Cotton model was validated with 2010
data, percentage error in observed and simulated LAI was
higher than 2009 in some of the treatments. For instance, the
model showed 13.88, 15.80 and 11 % errors in planting date
May sown, N rate of 50 kg ha-1 and cultivar CIM-496;
however, the % error for rest of the treatments ranged from
2.62 to 6.09. Similarly MPD in this year was higher (7.80)
compared to 2009 (1.61). RMSE during 2010 ranged from
0.57 to 1.28 for different treatments. Relationship between
simulated and observed LAI for all studied treatments
demonstrated in Fig.3, indicating strong association.
Total dry matter (TDM kg ha-1): The data in Table 5
indicated a significant and positive correlation among
simulated and observed TDM during both the season at all
three sites. Pooled date for three sites showed that model

slightly over estimated (0.21%) TDM in early sown crop
than late in which model estimation was 3.39% more than
observed data. Percent difference increased to 6.97% at
higher fertilizer rates (200 kg N ha-1) than lower 50 kg N ha-1
with percent error of 2.14. Root mean square difference for
cultivars ranged from 377 to 417 kg ha-1 between observed
and simulated data of total dry matter accumulation during
2009. Model validation with independent set of data from
three sites in second year was also good with MPD of 3.96%
only. RMSE values were ranging from 367 kg ha-1 to 449 kg
ha-1 which showed that model was robust in validation under
various climatic conditions. Coefficient of regression (R2)
between observed and simulated data for the pooled data at
three sites was estimated at 0.87 and 0.89 for 2009 and 2010,
respectively (Fig. 4).

643

Wajid, Ahmad, Hussain, Rahman, Khaliq, Mubeen, Rasul, Bashir, Awais, Iqbal, Sultana & Hoogenboom
Table 5. Comparison between simulated and observed total dry matter (kg ha -1) at different nitrogen levels, sowing
dates and cotton cultivars during year 2009 and 2010
Treatment
Calibration (2009)
Validation (2010)
Obs.
Sim.
Error (%)
RMSE
Obs.
Sim.
Error (%)
RMSE
11650
11675
0.21
496.67
10871
11594
6.65
449.00
May Sown
9871
10206
3.39
391.00
9451
9873
4.47
405.33
June Sown
8957
9149
2.14
393.00
8855
9000
1.64
367.33
N 50 kg ha-1
11858
12684
6.97
479.67
11757
11895
1.17
411.67
N 200 kg ha-1
10168
10710
5.33
417.00
10085
10332
2.45
398.67
CIM-496
9833
10145
3.17
377.33
9816
10558
7.56
398.33
CIM-506
10535
10843
2.92
405.67
10555
10995
4.17
396.67
NIAB-111
10647
10864
2.04
407.00
10620
11002
3.60
401.33
SLH-284
3.27
3.96
MPD
4.5
Leaf Area Index (2009)

Leaf Area Index (2010)

Observed

4.0

3.5

3.0

2.5
2.5

3.0

3.5

4.0

4.5
2.5

3.0

3.5

Simulated

4.0

4.5

Simulated

Figure 3. Relationship between simulated and observed leaf area index (LAI) for cotton cultivars during growing
season of 2009 and 2010
13000
Total Dry Matter (2009)

Total Dry Matter (2010)

Observed

12000

11000

10000

9000

8000
8000

9000

10000

11000

12000

13000
8000

Simulated

9000

10000

11000

12000

13000

Simulated

Figure 4. Relationship between simulated and observed total dry matter (Kg ha -1) for cotton cultivars during
growing season of 2009 and 2010
Seed cotton yield: The data presented in Table 6 shows that
the May sown crop produced seed cotton yield of 2260 kg
ha-1 whereas June sown produced 1765 kg ha-1 according to

model simulations; the observations for May and June sown
crop were 2162 and 1681 kg ha-1, respectively, during cotton
growing season 2009. The two nitrogen levels showed error

644

Simulating cotton yield using CROPGRO model
Table 6. Comparison between simulated and observed seed cotton
sowing dates and cotton cultivars during year 2009 and 2010
Treatment
Calibration (2009)
Obs.
Sim.
Error (%)
RMSE
2162
2260
4.53
216.33
May Sown
1681
1765
5.00
144.67
June Sown
1513
1561
3.17
134.67
N 50 kg ha-1
2196
2313
5.33
227.00
N 200 kg ha-1
1756
1877
6.89
180.33
CIM-496
1854
1951
5.23
185.00
CIM-506
1899
2033
7.06
198.00
NIAB-111
1920
2020
5.21
205.00
SLH-284
5.30
MPD

yield (kg ha-1) at different nitrogen levels,
Validation (2010)
Sim.
Error (%)
1760
4.02
1502
5.03
1421
3.72
1771
6.43
1548
1.64
1654
4.55
1658
4.54
1634
5.08
4.38

Obs.
1692
1430
1370
1664
1523
1582
1586
1555

RMSE
178.67
127.67
122.00
178.00
163.00
169.67
168.33
167.67

2400
Seed cotton Yield (2009)

Seed cotton Yield (2010)

Observed

2200

2000

1800

1600

1400
1400

1600

1800

2000

1200
2400

2200

1400

1600

1800

2000

2200

2400

Simulated

Simulated

Figure 5. Relationship between simulated and observed seed cotton yield (kg ha -1) for cotton cultivars during
growing season of 2009 and 2010
%age range of 3.17 to 5.33. Similarly all the cultivars
showed %age difference of 5.21 to 7.06 which is an
acceptable limit.
Model evaluation for second year also remained satisfactory
for early (with RMSE of 179 kg ha-1) and late planting (with
RMSE of 128 kg ha-1). Increasing nitrogen rates increased
the RMSE value (122 to 178 kg ha-1). In case of cultivars,
the range of RMSE was very consistent for the four cultivars
(163 to 173 kg ha-1). Figure 4 illustrates the scatter of
simulated and observed seed cotton yields around the
regression line. There was strong and positive correlation
between observed and simulated data and common
regression accounted for 99 % variance in the data (Fig. 5).
DISCUSSION
Many scientists used Cropping System Model (CSM)CROPGRO-Cotton for the simulation of growth,
development, and yield of cotton for different weather and

soil conditions and management practices (Jones et al.,
2003; Hoogenboom et al., 2004; Ortiz et al., 2009; Pathak et
al., 2009). Similarly the model was found to have the ability
to evaluate the cotton performance for climate change
situations (Murthy, 2004). So it was need of the hour to
evaluate this model for quantification of management
options (planting dates and nitrogen rates for promising
cotton cultivars) under different climatic conditions in
Punjab (Faisalabad, Multan and Sahiwal). The selection of
two years (one year for calibration and the second for
evaluation) as has been used in some of the studies (Khaliq
et al., 2007; Mubeen et al., 2013; Wajid et al., 2013) proved
quite helpful in monitoring the model performance under
differentiated weather conditions of the three sites.
The calibration of model was done with data (that included
phenology, biomass, LAI, and yield components) collected
during 2009 at all locations against treatment May 15
sowing, cultivar CIM-496 and 200 kg N ha-1 that performed
best in field trials. Model simulations showed that crop

645

Wajid, Ahmad, Hussain, Rahman, Khaliq, Mubeen, Rasul, Bashir, Awais, Iqbal, Sultana & Hoogenboom
reached flowering stage 62-66 days after sowing in all
treatments and the observed days ranged from 61-65 days
which indicated that model was fit and worked well under
our environmental conditions. Similarly, validation of model
during year 2010 shows that crop reached at anthesis stage
in 60-67 days after sowing. RMSE values indicated that
model worked well for simulation under determined set of
cultivar coefficients.
In case of days to maturity, mean percent difference (MPD)
of 0.86 during calibration and MPD of 1.16 was found
during evaluation. The crop which was applied N @ 50 kg
ha-1 matured 2 days earlier over treatment with N application
of 200 kg ha-1. In general, error between simulated and
observed days to maturity was more at higher levels of
nitrogen application (200 kg ha-1) than lower level (50 kg ha1
). Li et al. (2009) worked on a simple process based cotton
model found that the simulated values of boll maturation
period were very consistent with the observed values, with
root mean square error (RMSE) lower than 3 days.
Model predicted higher (4.42) LAI @ 200 kg N ha -1 than at
low rate of 50 kg N ha-1 (3.18). Similarly when CROPGROCotton model was validated with 2010 data, %age error in
observed and simulated LAI was higher than 2009 in some
of the treatments. Similar results were found by Zamora et
al. (2009) who reported that LAI was underestimated in the
more shaded treatment rows 4 and 8. They further suggested
that additional research is needed to improve the ability of
model to simulate LAI under shading conditions.
The model overestimated TDM than the observed values
indicating that there is a potential of producing more TDM
under these sets of agro-ecological conditions. Percent
difference increased to 6.97 % at higher fertilizer rates (200
kg N ha-1) than lower 50 kg N ha-1 with percent error of 2.14.
Model validation with independent set of data from three
sites in second year was also good with MPD of 3.96%.
These results are in line with Ortiz et al. (2009) they
reported that CROPGRO-Cotton simulated biomass with
error range of 6% to 18.4%. Similarly Zamora et al. (2009)
found that the model provided a close agreement between
measured and simulated biomass both in 2001 and 2002 (R 2
= 0.95 and R2 = 0.92, respectively).
The model simulated seed cotton yield reasonably well with
error percentage of 3.17 to 7.06 during 2009 having RMSE
values 134.67 to 227.00 kg ha-1 in all the treatments of
sowing date, nitrogen levels and cultivars. Similarly during
2010, error %age in the prediction of seed cotton yield was
in the range of 1.64 to 6.43 with RMSE of 122.00 to 178.67
kg ha-1. The MPD was 5.30 and 4.38 during 2009 and 2010,
respectively. Cammarano et al., (2011) found that lint yield
was well simulated for all the treatments (Root Mean Square
Error 100 kg ha-1 for the 2008 growing season and 254 kg
ha-1 for the 2009 growing season).When the model was run
in different locations between south east Queensland and
northern New South Wales it accurately simulated cotton

yield (y= 0.75x +218.2; r2= 0.79; RMSE= 395.3 kg ha-1).
Similarly Ortiz et al., (2009) reported that prediction by
CROPGRO-Cotton for seed cotton fell within a range of 11.2% to 2.7%. It is also evident from the yield data for
individual seasons that observed and simulated values of
sowing done on May gave the higher seed cotton yield than
late sowing; these results are in accordance with Ali et al.
(2004).
In conclusion, the performance of CSM-CROPGRO-Cotton
model under DSSAT V 4.0.2.0 was satisfactory for different
parameters of the cotton. Model predicted seed cotton yield
and total dry matter with acceptable root mean square error
and good agreement of d-statistic between observed and
simulated data. In future, there is a dire need to evaluate
model for the assessment of rising temperature, elevated
CO2 and changing rainfall pattern on the quality of seed
cotton yield.
REFERENCES
Akhtar, M., M.S. Cheema, M. Jamil, S.A. Shahid, and M.I.
Shahid. 2002. Response of cotton genotype to time of
sowing. Asian J. Plant Sci. 1: 538-539.
Ali, M., G. Mohy-ud-Din, M.A. Ali, S. Bashir, and L. Ali.
2004. Cotton yield as affected by different sowing dates.
Int. J. Agri. Biol. 6: 644-646.
Arshad, M., A. Wajid, M. Maqsood, K. Hussain, M. Aslam
and M. Ibrahim. 2007. Response of growth, yield and
quality of different cotton cultivars to sowing dates.
Pak. J. Agri. Sci. 44: 208-12.
Boote, K.J., J.W. Jones, G. Hoogenboom and J.W. White.
2010. The role of crop systems simulation in agriculture
and environment. Int. J. Agri. & Environ. Inform. Sys.
1: 41-54.
Cammarano, D., J. Payero, P. Grace, and B. Basso. 2011.
Evaluating the CROPGRO-Cotton Model for
Estimating Cotton Production. P.106-107.In the
Australian Sub-Tropics. Presented in: International
Annual Meetings. 16-19 Oct. 2011. American Society
of Agronomy-Crop Science Society of America- Soil
Science Society of America.
Government of Pakistan. 2012. Economic Survey of
Pakistan 2011-12. Finance and Economic Affairs
Division, Ministry of Finance, Islamabad Pakistan.
Hima, B.L., Y.C. Lu, A.G. Richardson, and K.R. Reddy.
2004. An economic analysis of climate change on
cotton production across the U.S. cotton belt. P. 106. In:
Beltwide Cotton Conference. 5-9 Jan. 2004. San
Antonio, Texas, USA.
Hoogenboom, G., J.W. Jones, P.W. Wilkens, C.H. Porter,
W.D. Batchelor, L.A. Hunt, K.J. Boote, U. Singh, O.
Uryasev, W.T. Bowen, A.J. Gijsman, A.S. Du Toit,
J.W. White, and G.Y. Tsuji. 2004. Decision Support

646

Simulating cotton yield using CROPGRO model
System for Agrotechnology Transfer. Ver. 4.0.
Honolulu, Hawaii: University of Hawaii.
Hoogenboom, G., J.W. Jones, P.W. Wilkens, W.D.
Batchelor, W.T. Bowen, L.A. Hunt, N.B. Pickering, U.
Singh, D.C. Godwin, B. Baer, K.J. Boote, J.T. Ritchie,
and J.W. White. 1994. Crop models. In: Tsuji, G.Y.,
Uehara, G., Balas, S. (Eds.), DSSAT Version 3. Univ.
of Hawaii, Honolulu, HI. 2: 95-244.
Hoogenboom, G., J.W. Jones, P.W. Wilkens, C.H. Porter,
K.J. Boote, L.A. Hunt, U. Singh, J.L. Lizaso, J.W.
White, O. Uryasev, F.S. Royce, R. Ogoshi, A.J.
Gijsman, and G.Y. Tsuji. 2011. Decision Support
System for Agro-technology Transfer (DSSAT) Version
4.5.X.X (Replace X.X with sub-version numbers in
your output files) [CD-ROM]. University of Hawaii,
Honolulu, pp.121-278.
Jawdad, D., M.A. Hilali, Z. Ayyoubi, R. Elias, R. Al-Rayan,
M.N. Al-Salti and B. Al-Safadi . 2012. Response of
cotton varieties to different environments: Flowering
behavior and fiber quality. Pak. J. Agri. Sci. 49:289298.
Jones, J. W., G. Hoogenboom, C.H. Porter, K.J. Boote, W.D.
Batchelor, A. Hunt, P.W. Wilkens, U. Singh, A.J.
Gijsman, and J.T. Ritchie. 2003. DSSAT Cropping
System Model. Eur. J. Agron. 18: 235-265.
Kakni, V.G., S. Koto, K.R. Reddy, P.V. Wallace, P.V.
Prasad, V. Ready and D. Zhar. 2005. Comparison of
pollen and physiological characters of cotton cultivars
as screening tools for high temperature tolerance. Ann.
Bot. 96: 59 - 67.
Khaliq, T., A. Ahmad, A. Hussain, A. Wajid and G.
Hoogenboom. 2007. Modeling nitrogen use efficiency
of maize cropping system in Pakistan. 4-8 Nov. 2007.
In: International Annual Meetings, American Society of
Agronomy-Crop Science Society of America-Soil
Science Society of America.
Li, W., Z. Zhou, Y. Meng, N. Xu and M. Fok. 2009.
Modeling boll maturation period, seed growth, protein,
and oil content of cotton (Gossypium hirsutum L.) in
China. Field Crops Res.112: 131-140.
Mahamood, R., M. Meo, D.R. Legates and M.L. Morrissey.
2003. The CERES-Rice model based estimates of
potential monsoon season rain-fed rice productivity in
Bangladesh. Prof. Geogr. 55: 259-273.
Mubeen, M., A. Ahmad, A. Wajid, T. Khaliq and A.
Bakhsh. 2013. Evaluating CSM-CERES-Maize model

for irrigation scheduling in semi-arid conditions of
Punjab, Pakistan. Int. J. Agric. Biol. 15: 1-10.
Murthy, V.R.K. 2004. Crop growth modeling and its
applications in agricultural meteorology. In: Sivakumar,
M.V.K. P.S. Roy, K. Harsen, S.K. Saha (eds.), Satellite
remote sensing and GIS applications in agricultural
meteorology workshop. 7-11 July. 2004, Dehra Dun,
India; p. 235-261.
Nazar, A., M. Iftikhar, B. Shahbaz and W. Ishaq. 2012.
Influence of irrigation water types and stress levels on
cotton fibre and yarn quality for different varieties. Pak.
J. Agri. Sci. 49:597-601.
Ortiz, B.V., G. Hoogenboom, G. Vellidis, K. Boote, R. F.
Davis and C. Perry. 2009. Adapting the CROPGROcotton model to simulate cotton biomass and yield under
southern root-knot nematode parasitism. Transactions of
the ASABE. 52: 2129-2140, 2009. American Society of
Agricultural and Biological Engineers, ISSN 21510032.
Pathak, T.B., J.W. Jones, C. Fraisse, D. Wright, G.
Hoogenboom and J. Judge. 2009. Uncertainty Analysis
of CROPGRO-Cotton Model. American Geophysical
Union, Fall Meeting; p.78, 2009.
Shabbir, A., M. Arshad, A. Bakhsh, M. Usman, A. Shakoor,
I. Ahmad and A. Ahmad. 2012. Apparent and real water
productivity for cotton-wheat zone of Punjab, Pakistan.
Pak. J. Agri. Sci. 49:357-363.
Soomro, A.R., R. Anjum, A.W. Soomro, A.M. Memmon
and S. Bano. 2001. Optimum sowing time for new
commercial cotton variety, Marvi (CRIS-5A). The
Pakistan Cottons 45: 25-8.
Wajid, A., M.H. Rahman, A. Ahmad, T. Khaliq, N.
Mahmood, F. Rasul, M.U. Bashir, M. Awais, J. Hussain
and G. Hoogeboom. 2013. Simulating the interactive
impact of nitrogen and promising cultivars on yield of
lentil (Lens culinaris) using CROPGRO-legume model.
Int. J. Agric. Biol. 15: 1331-1336.
Wallach, D. and B. Goffinet. 1989. Mean Square error of
prediction as a criterion for evaluating and comparing
system models. Ecol. Model. 44: 200-306.
Zamora, D.S., S. Jose, J.W. Jones and W.P. Cropper. 2009.
Modeling cotton production response to shading in a
pecan alley cropping system using CROPGRO.
Agroforestry Sys. 76: 423-435.

647