Crop Modeling with

Special Reference to
Wheat Crop

1
 Modeling …
Getting information about how something will behave
without actually testing it in real life using models

“A model is a simplified representation of a

system or a process”
 Based on the assumption that any system or a
process can be expressed in a mathematical
equation or a set of equations
eg., Yield= Biomass X HI
Crop Simulation: Mimicking the real world situation or
Modeling Growing the crop on the computer 2
Why modeling in agriculture..
Population Decision making
increases, Climate change and policy
pressure on Climatic variability planning
resources

Sustainability Estimation of impact Forecasting of

Precision farming on crop yields and crop yields and
Adaptation strategies yield gaps

•Require deep understanding of the system

•Require bulk of information (spatial and temporal)
→Costly, Cumbersome and sometimes unreliable

Modeling
(Low cost and rapid method)
3
 Systems Approach
• Agricultural system is a complex system that
includes reactions within the plant and Model
interactions of plant with biotic, abiotic and
management factors.
• For an agronomist, crop is a system with leaf,
stem, root and processes as elements which

Complexity
interact strongly.
 Systems approach for crop and soil studies
• Crop modeling is a significant part of
Systems approach
• Universal models do not exist, like in fields
of physics and engineering.
• Different models for different crops, Real world system
production levels, sites and subsystems. 4
 Model Approach

The model approach is selected on the basis:

• Purpose of model application 5
• Experimental data available for developing and/or testing the model
Descriptive/Explanatory

6
 Crop simulation model
• CSM integrates the scientific
knowledge from different
disciplines and studying the
CSM in general calculate
system from several points of
or predict crop growth and
view viz.,
yield as a function of
• Physical, physiological,
chemical, biochemical, bio- -Genetics
technological, agronomical, -Weather
entomological, pathological, -Soil conditions
economic benefit angles, etc -Crop management
• CSM are designed to mimic
the system at short intervals
7
 INPUTS

Soil condition Weather data

Model

Crop management Genetics

Simulation

Growth Development
Yield
8
OUTPUTS
 Some examples
Software Detail
DSSAT Framework of crop simulation models including
modules of CERES, CROPGRO and CROPSIM
APSIM Modeling framework for a range of crops
CERES Series of crop simulation models
CropSyst Wheat & other crops
ORYZA1 Rice, water
WOFOST Wheat & maize, Water and nutrient
EPIC Erosion Productivity Impact Calculator
GWM General weed model in row crops
GOSSYM Cotton
CROPGRO Soybean and other legumes
9
DSSAT Modular V4.6- Cropping System Model
(Developed by IBSNAT in 1989)

Jones et al (2003)
European J Agron 18: 235-65 10
APSIM modeling framework
(Developed by APSRU, Australia, 1991)

Keating et al (2003) 11
European J Agron 18: 267-88
 Steps of modeling
Model understanding

Minimum data set (Inputs)

Outputs

Model calibration
(Model parameterization)

Model validation & evaluation

Model application
(Can be used to perform ‘What-if’ 12
experiments)
Flowchart of daily biomass growth
calculations in CropSyst model

Stockle et al (2003) 13
European J Agron 18: 289-307
Minimum data set for model operation
– Data to develop model
• Information about growth of the plant as affected by relevant
environmental and physiological variables collected in:
– in controlled environment (to evaluate the effect of single
environmental factors)
– in field experiment (to evaluate the response of plant to combined
environmental factors)
– Data to validate model
• Independent experimental data on crop phenology, growth and
yield
– Data to run model
• Environmental data on weather and soil
• Management data (Crop, cultivar, planting date, plant density,
amount and timing of input applications, etc.) 14
 Model Calibration
• Fine-tuning/modification/minor adjustments of
some model parameters so that data simulated by
the model comply with the observed data

• More reliable data→ further experimentation than

extensive modification of model parameters to
achieve an acceptable fit which can cause changes
in model structure

15
 Model validation Types of Model validation
 Confirmation that calibrated -Apparent validation
model is close to reality
using same data directly
(Simulated vs Observed data
not used in calibration stage)
(Optimistic output)
 Determines whether the -Internal validation
model adequately fulfils its using same data
initial objectives (Reproducibility test)
 Determines the suitability for -External validation
an intended purpose using different data
 Testing the prediction (Transportability in space
capability/accuracy of the and time)
model in new and different
environments
16
 Model Evaluation
Prediction capability of the model can be tested by
Sensitivity analysis: Examines the response of outputs to changes
in input data and model parameters
Statistical indicators
-Intercept and slope of regression of observed vs. predicted
variables
-Nash-Sutcliffe modeling efficiency (ME): >0.75
-Root mean square error (RMSE):
-Normalized RMSE: <10 %
-Index of agreement (d):
-Coefficient of determination (R2): >0.75
Best fit: R2 =1, RMSE=0 and d=1
17
Performance of CERES-Wheat model
for various data sets

Timsina and Humphreys (2006), Asia & 18

Australia
 Calibration and validation of DSSAT-CSM-CERES-
Wheat model for grain yield and biomass of rainfed wheat

Statistics Calibration Validation

(2003-04 to 2005-06) (2006-07 to 2009-10)
Yield Biomass Yield Biomass
Root mean square error 0.10 0.12 0.09 0.17
(RMSE)
Nash-Sutcliffe modeling 0.87 0.97 0.57 0.91
efficiency (ME)
Index of agreement (d) 0.95 0.99 0.87 0.98
Coefficient of 0.99 0.99 0.84 0.95
determination (R2)

Vashisht et al (2013), Ballowal Saunkhri

Reg Environ Change 13: 989-98 19
 Evaluation of CERES-Wheat model in predicting phenological
dates, grain yield and biomass production

Year Anthesis (DAP) Maturity Grain yield (t/ha) Biomass (t/ha)

(DAP)
P red Obs P red Obs Pred Obs Pred Obs
99-2000 104 106 137 140 6.1 5.4 15.2 14.6
2000-01 106 103 128 125 5.8 5.3 15 14.7
03-04 112 107 130 128 4.9 4.3 13 13.3
04-05 100 102 133 136 5.2 4.7 13.5 12.9
NRMSE 3 2 11.8 3.4
(%)
d-Index 0.72 0.91 0.71 0.92
R2 0.60 0.96
Andarzian 0.97
et al (2014), Iran 0.86
J Saudi Soc Agric Sci doi: 10.1016/[Link].2014.04.004
20
Validation of DSSAT-CSM-CERES-Wheat model
for phenology, growth and yield of wheat

Data sets N SDObs SDsim Abs Normal d-index

RMSE RMSE
Anthesis (day) 22 5.5 4.1 5.3 7.0 0.70
Maturity (day) 28 5.2 3.7 4.5 3.4 0.73
Grain yield (kg/ha) 32 1166 972 617 15.1 0.92
Straw yield (kg/ha) 24 1958 1014 2962 39.9 0.57
Biomass yield 24 2915 1920 3091 23.1 0.72
(kg/ha)
HI (%) 24 4.3 2.6 13.5 35.1 0.34

Timsina et al (2008), Punjab

Agric Water Mgt 95: 1099-1110 21
 Validation of CropSyst model

RMSE=0.09 RMSE=0.05
RMSE=0.24

Jalota et al (2011), Punjab

Expl Agric 47: 609-28 22
 Absolute per cent deviation between observed
and predicted grain yield with CERES-wheat
and CropSyst models during validation

Irrigation 4-Irrigations 3-Irrigations 2-Irrigations

regimes CERES CropSyst CERES CropSyst CERES CropSyst
-Wheat -Wheat -Wheat
N-levels

N0 9.8 7.54 61.10 15.09 66.45 2.29

N60 0.62 1.07 17.93 9.84 11.15 9.84
N90 0.71 0.20 16.32 1.82 12.9 0.52
N120 6.19 4.38 10.32 7.56 2.72 13.79
N150 0.79 3.05 1.38 4.30 0.38 11.24

Singh et al (2008), New Delhi

23
Agric Water Mgt 95: 776-86
 Model Validation Skill Scores for Leaf area index and
grain yield of wheat
Genotype APSIM DSSAT
R2 RMSE NRMSE R2 RMSE NRMSE
Leaf area index
Tatara 0.99 0.34 0.02 0.56 1.40 0.39
NARC 0.97 0.38 0.02 0.51 1.11 0.29
Sehar 0.94 0.48 0.05 0.48 1.30 0.59
SKD-1 0.93 0.51 0.06 0.59 1.30 0.55
F-Sarhad 0.91 0.65 0.08 0.51 1.63 0.91
Grain yield (t/ha)
Tatara 0.95 0.26 0.00 0.62 0.73 0.03
NARC 0.92 0.48 0.02 0.56 0.70 0.03
Sehar 0.83 0.48 0.02 0.51 0.70 0.03
SKD-1 0.87 0.49 0.02 0.54 0.68 0.03
F-Sarhad 0.84 0.49 0.02 0.58 0.44 0.03
Ahmed and Hassan (2011), Pakistan 24
19th Intl Cong on Modelling and Simulation, Perth, Australia, 12–16 Dec 2011, pp 1174-80
Comparison of observed and APSIM simulated
LAI, biomass, yield and soil water content

Chen et al (2010), China

25
Agric Water Mgt 97: 1175-84
Open DSSAT…

26
Create data files
Weather file

Soil
file

Genotype file
(Definition of
cultivars)
27
Location of the genotype file

28
Location of weather file

29
Program to generate weather data

30
Location of the input experiment file

31
Experiment file can be edited with a
text editor (Notepad)

32
Start simulation…

33
Running..

34
Select the Experiment..

35
View the results

36
 Retrieve output files for analysis

• C:/DSSAT/WHEAT/[Link]
• C:/DSSAT/WHEAT/[Link]
• C:/DSSAT/WHEAT/[Link]
• C:/DSSAT/WHEAT/[Link]
• C:/DSSAT/WHEAT/[Link]
• There are DOS text files
• Can be imported into Excel
 Analyse and present the results

37
 Areas of model application
 Research
• Research understanding
• Integration of knowledge across disciplines
• Improvement in experiment documentation and data organisation
• Genetic improvement
• Yield analysis
 Crop system management
• Cultural and input management
• Risk assessment and investment support
• Site specific farming
 Policy analysis
• Best management practices
• Yield forecasting
• Introduction of a new crop and cultivar
38
• Global climate change and crop production
Incorporating crop models into traditional
agronomic research
Question, Problem

Hypotheses

Experiments
Computer Experiments
Analysis

Conclusions, Decisions,
Recommendations

Combination of field experimentation and modeling is a powerful

approach that can give conclusive and meaningful results
39
40
 Simulated grain yield and water productivity
of rainfed wheat in relation to sowing dates
DOS 15 Oct 25 Oct 5 Nov 15 Nov 25 Nov 5 Dec 15 Dec Mean
Grain yield (kg/ha)
2001-10 2343 2318 2254 2215 2235 2212 2119 2242
2021-30 1574 2357 2814 2982 2426 1690 1301 2163
2031-40 1063 1335 1842 1832 1522 1495 1516 1515
2041-50 1003 1360 978 844 650 656 673 881
Mean 1496 1843 1972 1968 1708 1513 1402 -
Crop water productivity (kg grain/m3)
2001-10 1.19 1.17 1.15 1.14 1.15 1.15 1.13 1.15
2021-30 0.52 0.75 0.86 0.90 0.77 0.55 0.43 0.68
2031-40 0.36 0.46 0.59 0.57 0.52 0.51 0.51 0.50
2041-50 0.34 0.46 0.34 0.30 0.24 0.24 0.25 0.31
Mean 0.60 0.71 0.73 0.73 0.67 0.61 0.58 -
Vashisht et al (2013), Punjab 41
Reg Environ Change 13: 989-98
Simulated yield for different sowing
dates for different locations

Andarzian et al (2014), Iran 42

J Saudi Soc Agric Sci doi: 10.1016/[Link].2014.04.004
 Effect of sowing date and irrigation
management on simulated grain yield

Timsina et al (2008), Punjab

43
Agric Water Mgt 95: 1099-1110
 Simulated yield, duration of phenophases and WUE
and RE at different locations using CropSyst model
Time slice Yield F F-GF GF-M Total WUE Recovery
(kg/m3) Efficiency (%)

Amritsar
Present 57.4 q/ha 122 10 36 168 1.11 75.0
Mid-century 4.5 % ↑ 96 13 44 152 1.42 76.7
End-century 21 % ↓ 83 11 43 138 1.33 69.2
Ludhiana
Present 56.8 q/ha 115 10 37 162 1.22 69.6
Mid-century 7.8 % ↓ 95 13 45 153 1.35 72.4
End-century 15.4 % ↓ 82 12 45 139 1.46 63.6
Patiala
Present 61.8 q/ha 109 10 37 156 1.20 70.9
Mid-century 12.4 % ↓ 86 12 45 143 1.41 72.1
End-century 40.4 % ↓ 76 10 43 129 1.16 61.9

Jalota et al (2014), Punjab 44

Agric Syst 131: 77-86
Simulated yield, water and N-balance and –use
efficiency using CropSyst model
Water balance components
Time Yield Irr. RF ET E T D ∆SWS WUE
slice (q/ha) (mm) (mm) (mm) (mm) (mm) (mm) (mm) (kg/m3)
Present 56.4 424 103 445 94 350 69 13 1.27
Mid- 56.5 323 241 424 159 265 81 59 1.33
century
End- 49.3 200 209 318 151 167 94 -3 1.55
century
N-balance components
Time Appl Mineral Gas Leached Uptake Immobi Initial Final RE
slice ied ization loss lization (%)
Present 120 3 35 0 86 2 0 0 73.5
Mid- 120 2 28 2 89 1 4 5 76.9
century
End- 120 1 28 2 79 0 5 17 69.4
century
Jalota et al (2013), Punjab 45
Agric Water Mgt 116: 29-38
 Uses and Limitations of modeling
Uses Limitations
• Get a better understanding of • Crop models are crude
the system (observations on images of reality
daily basis) • Models need to be
• Save resources and time by calibrated and validated to
reducing field experimentation represent reality
• Experiment in new situations • Models need data and
and environment technical expertise
• Compress or expand the time • Model developed for one
• Number of treatments are objective, failure for other
possible • Models alone do not provide
• Study the effect of policy an answer, stakeholder
changes interaction is essential
46
 Conclusion
• The models help to compare experimental research findings
across sites, extrapolate experimental field data to wider
environments, save resources and time by reducing field
experimentation, develop management recommendations and
decision support systems, explore effects of climate change and
make yield predictions.
• DSSAT-CSM-CERES-Wheat model satisfactorily simulates
phenology, growth and yield of wheat over time and space.
• CropSyst model performs reasonably good to simulate water
and nitrogen bugdets.

47
Thank
s
48