Lecture 4

Statistical, weather analysis and crop

simulation models
(…contd)

By
Dr. Sabah Parvaze
Div. of IDE
CoAE&T, SKUAST-Kashmir
 Types of crop-weather models

1. Empirical-statistical models or Simple statistical model,

2. Crop-weather analysis models or Parameterization models, and

3. Crop growth simulation models or Analog-physical model.

 Empirical-statistical models or Simple statistical model

Empirical statistical model are developed on the basis of long term relationship between
crop yield and several variables (representing weather, soil characteristic, technology
trend, etc.,).

There are two types of Statistical models:

a) Trends model, and

b) Agrometeorological Model

a) Trends model
Trendmodels are those model in which yield or other crop parameters are related with
time. These model represent the technological advancement and can be expressed as:
Y = a + bX,
Where Y is yield or any other plant entity,
“a” is constant (intercept),
“b” is slope,
X is independent variable such as time
b) Agrometeorological Model
Agrometeorological model involve meteorological parameters which are
important for growth and development of plants. These model can be further
divided in two categories:
I) Linear models and
II) Non linear models

I) Linear models
Linear model show straight relationship between crop entity and independent
variables and can be of two type A) Single variable type model and B) Multiple
variable regression model.
50 y = 2.7033x + 8.7253
45 R2 = 0.8843
40
35
Slope = y/x
30
25
20
15
10
5
Intercept
0

0 2 4 6 8 10 12 14 16
A) Single variable type model: These type of model uses only one independent
weather variable and can be represented by following expression:
Y = a + bX, where X is weather variable
B) Multiple variable regression model: These type of model uses two or more
independent weather variables and can be represented by following expression:
Y = a + bX1 + cX2……..nXn ,
where where b, c…n are slope coefficients and X1, X2….Xn are independent
variables

II) Non linear models

These models represent the nonlinear relationship between crop yield and weather
variables.These kind of relationship is very common weather and crop yield or other
entity. Mainly there are two type of non-linear models A) Quadratic models, and B)
exponential models.
A) Quadratic models 30
y = -0.5405x2 + 8.3188x- 9.0604
R2 = 0.9361
Quadratic model uses only one
25
independent variable. The
20
dependent variables first increases
with increasing independent 15

variables and after certain point it 10

decreases.
5
Y = a + bX - cX2
0
0 2 4 6 8 10 12 14 16

B) exponential models 100

in exponential model the dependent 90 y = 6.6962e0.1826x
80 R2 = 0.9845
variables increases slowly during first 70
stage and then tends to be increase 60
sharply in later stage when 50
40
relationship is positive, while it show
30
reverse trend when relationship is 20
negative. 10
0
Y = a * Exp(bx)
0 2 4 6 8 10 12 14 16
 Crop Weather Analysis Models

The crop-weather analysis models are defined as the product of two or more
factors, each representing the simplified functional relationship between a
particular plant response (e.g., yield) and the variations in selected variables at
different plant development phases (Baier, 1977). Time interval of one day or
more is used.

m
Y =∑ V1 * V2 * V3
t=1

Where, V1 = Maximum temperature,

V2 = Minimum temperature
V3 = Ratio of actual to potential evapotranspiration
 Crop Growth Simulation Models
Crop growth simulation models are simplified mathematical representation of the
complex physical, chemical and physiological mechanisms underlying plant growth and
its response to environment. They are dynamic models, which provide prediction as
well as explanation of the integrated behaviour from more detailed knowledge of the
underlying physiological processes. In these models the crop is described by a set of
state variables (e.g., weight of various organs) that are updated at each iteration of the
variable by rate variables (e.g., flow of carbon in photosynthesis and respiration)
defining changes in the state variables. The rate variables are assumed to be constant
during the iteration interval, which is generally one hour and sometimes even less.
Crop growth models can be categorized as Qualitative model and Quantitative model.
Qualitative models represent the conceptual relationships and feedback among the
various components of the system. While Quantitative models, are the translations of a
qualitative model into a computer program.
 Reduced Oxidised Floodwater
soil layers soil zone
A complicated system
 Crop Simulation Models

• Crop simulation models integrate the

current state-of-the art scientific
knowledge from many different
disciplines, including crop physiology,
plant breeding, agronomy,
agrometeorology, soil physics, soil
chemistry, soil fertility, plant pathology,
entomology, economics and many others.
 Agricultural Models
• Crop simulation models in general
calculate or predict crop growth and yield
as a function of:
– Genetics
– Weather conditions
– Soil conditions
– Crop management
 Soil Conditions Weather data

Model

Crop Management Genetics

Simulation

Growth Development

Yield
 Soil Conditions Weather data

Model

Crop Management Genetics

Simulation

Growth Development

Yield

Pollution Net Income Resource Use

 Crop Simulation Models
Four levels or phases (School of De Wit)
LEVEL 1
• Potential Production
– Solar radiation and temperature as input
– Simulate growth and development
– Plant carbon balance (photosynthesis,
respiration, partitioning)
 Level 2
Water-Limited Production
– Potential production +
– Precipitation and irrigation as input
– Soil profile water holding characteristics
– Plant water balance (transpiration, water
uptake)
– Soil water balance (evaporation, infiltration,
runoff, flow, drainage)
 Level 3
Nitrogen-Limited Production
– Water-limited production +
– Nitrogen fertilizer applications as input
– Soil nitrogen conditions
– Plant nitrogen balance (uptake, fixation,
mobilization)
– Soil nitrogen balance (mineralization,
immobilization, nitrification, denitrification)
 Level 4
Nutrient-Limited Production
– Nitrogen-limited production +
– Fertilizer applications as input
– Soil nutrient conditions
– Plant nutrient balance (uptake, mobilization)
– Soil nutrient balance
• Phosphorus, potassium, other minerals
 Level 4
Pest-Limited Production
– Nitrogen-limited production +
– Pest inputs - scouting report
– Dynamic pest simulation
• Insects, diseases, weeds
 Complexities of Agricultural
Production
Model
• Potential production
• Water-limited production
•

Complexity
Nitrogen-limited production
• Nutrient-limited production
• Pest-limited production
• Other factors
• Extreme weather events
• Salinity

Real World
 Production Crop Model Concepts
situation defining factors: CO2
Radiation
1 potential
Temperature
Crop characteristics
-physiology, phenology
-canopy architecture
limiting factors: a: Water
2 attainable b: Nutrients
- nitrogen
Yieldincreasing - phosphorous
measures

reducing factors: Weeds

3 actual Pests
Diseases
Yield protecting measures Pollutants

1500 5000 10,000 20,000 Production level (kg ha-1)

Source: World Food Production: Biophysical Factors of Agricultural Production, 1992.
 Crop Simulation Models
• Require information (Inputs)
– Field and soil characteristics
– Weather (daily)
– Cultivar characteristics
– Management
• Model calibration for local variety
• Model evaluation with independent data set
• Can be used to perform “what-if” experiments
 DSSAT
Decision Support System for
Agrotechnology Transfer
DSSAT v2.1 in 1989 → DSSAT v4.5 2010
About 2000 users in over 90 countries
 Components of DSSAT
DATABASES MODELS APPLICATIONS
Weather Validation /
Crop Models Sensitivity
Analysis
Soil
Seasonal Strategy
SUPPORT SOFTWARE Analysis
Genetics DSSAT User Interface
Graphics
Crop Rotation /
Pests Sequence
Weather Analysis

Experiments Spatial Analysis /

Soil GIS Linkage

Economics
Experiments

Pests

Genetics

Economics
DSSAT v4.5
• Windows-based
• Incorporates DSSAT CSM (+ Legacy Models)
• Field scale

• Data management tools

• XBuild: Input crop management information in standard format
• SBuild: Create and edit soil profiles
• GBuild: Display graphs of simulated and observed data, compute
statistics
• ATCreate: Create and edit observations from experiments,
formatted correctly
• WeatherMan:Assist users in cleaning, formating, generating
weather data
• ICSim – Introductory tool to demonstrate potential yield
concepts
DSSAT v4.5
Several different analytical capabilities

• Sensitivity Analysis: vary soil, weather, management or variety

characteristics for insight

• Seasonal Analysis: multiple-year simulations to evaluate

uncertainty in biophysical and economic responses

• Rotation/Sequence Analysis: long-term simulations to analyze

changes in productivity and soil conditions associated with
cropping systems

• Spatial Analysis: define spatially variable soil, weather,

management characteristics across a field or region for analysis
Main window in DSSAT v4.5
Selection of maize
experiment, all treatments
selected for simulation.
Circle shows button for
running the model and
for graphing results.
DSSAT4.5 graphics screens
Simulated and observed biomass accretion (kg DM/ha) for cowpea
cultivar TVU 3046 grown in Griffin, Georgia, in 1998

canopy

stem

leaf

Hoogenboom et al., 2000

 Input-Output files
Input Files Output Files
Mandatory files Plant Growth (at chosen time step)
-Management File Plant Nutrient (mostly Nitrogen)
-Genotypic file (cultivar Soil Nitrogen Balance
characteristics)
-Soil File (soil horizon) Soil Water Balance
-Weather File Soil Temperature
Optional Files Evapotranspiration
- Pests and disease Overview
-Economics Summary
-Plant out put (summary- for Errors and Warning
comparison with model output)
-Plant Growth (for graphical
comparison with model output)
 Calibration, Validation & Verification of Model
• Calibration refers to the estimation and adjustment of model
parameters to improve the agreement between model output and a
data set. It is done on independent dataset.

• Validation is the process of determining the degree to which a

model and its associated data are an accurate representation of the
real world from the perspective of the intended uses of the model.
Validation is done on independent data set. A certain degree of error
is acceptable depending upon application.

• Verification of a model is the process of confirming that it is

correctly implemented with respect to the conceptual. Verification is
done on the same dataset which has been used to build the model to
find and fix errors in the implementation of the model. The objective
of model verification is to ensure that the implementation of the
model is correct.
Sensitivity and Uncertainity Analysis
of Model
The aim of sensitivity analysis is to determine how sensitive the
output of a crop model is, with respect to the elements of the model
which are subject to variability. A sensitivity analysis consists of
estimating the influence of parameters on the state variables or on
the model outputs.

Uncertainty analysis consists of evaluating quantitatively the

uncertainty or variability in the model components (parameters,
input variables, equations) for a given situation, and deducing an
uncertainty distribution for each output variable rather than a
misleading single value. It provides methods to assess the
probability of a response to exceed some threshold. This makes
uncertainty analysis a key component of risk analysis.
 Limitations of Model
1. Crop simulation model assumes that crop field is homogenous in
respect to soil, microclimatic conditions, and crop conditions, while
large variability is actually found in field.
2. Crop models are not able to give accurate projections because
of inadequate understanding of natural processes and computer
power limitation.
3. Measured parameters also vary due to inherent soil heterogeneity
over relatively small distances.
4. Model performance is limited to the quality of input data.
5. It is common in cropping systems to frequently measure above-
ground crop growth and development, but data relating to root
growth and soil characteristics are generally not as extensive.
6. Simulation models require meteorological data, which is not
always reliable and complete.
7. At times, parameters that were not routinely measured may turn out
to be important and they are then arbitrarily estimated.
8. Sampling errors also contribute to inaccuracies in the observed data.
 Model Application
As a Research Tool
-Synthesize Research Understanding
-Integrate Knowledge Across Disciplines
-Experiment Documentation
-Assist in Genetic Improvement
-Yield Gap Analysis
-Separating Weather trend and technology trend
Crop System Management
-Assist in cultural management (planting time,
row spacing, fertilizer application, irrigation,
chemical application, harvesting etc.
 Model Application
-In season Decision- Contingent operations
-Site Specific or Precision Farming
Policy Analysis Tool
-Best Management Decisions to reduce Fertilizer
and Pesticide Leaching
-Option for reducing Soil Erosion
-Yield Forecasting
-Climate Change : Impact and resource
optimization
-Introduction of new crop / cultivar/cropping syst