LECTURE 13- CROP DISCRIMINATION AND YIELD MONITORING USING REMOTE SENSING
Identifying, discriminating and mapping crops is important for a number of reasons. Maps of different
crop types are necessary to prepare an inventory of what was grown in which area and when. This serves
the purpose of forecasting grain supplies (yield prediction), collecting crop production statistics,
assessment of crop damage due to storms and drought and monitoring farming activity
The two most important data requirement activities related to agriculture are crops grown with their extent
and yield monitoring. Traditional methods of obtaining these information is to take ground survey,
which consumes more time and labour. Also, the accuracy of these data is also under close scrutiny.
Why remote sensing?
Remote sensing offers an efficient and reliable means of collecting information required
for crop mapping and area estimation. Remote sensing enables collection of data in
inaccessible areas and over large areas in short time. Besides providing a synoptic view,
remote sensing can provide additional information on the status, stage and health of the
crop under consideration.
1 . Crop mapping / discrimination
Crop mapping or Crop discrimination is the preparation of maps with crops grown in the area of interest
using spectral differences of multi data remote sensing data.
The data (historical or in-season data) used for crop mapping depends on the scale of the
project. Generally, for crop mapping projects, multi temporal (multi date) data will be
used to identify the different stages of crops and thereby identifying the crop itself.
Ancillary data (rainfall data, collection in addition to remote sensing data is also of
paramount importance
Crop mapping and acreage estimation through remote sensing (Figure 13.1) broadly
consists of identifying representative sites (sampling plan) of various crops / land cover
classes on the image based on the ground truth collected. Ground truth planning and
collection is another important step. While collecting ground truth points, care should be
taken so that no classes / crops are left out. Also, number of samples to be collected are
to be estimated in prior before going for ground truth collection. Ground truth collection
should be done during different stages of crop growth. Once ground truth collection is
completed, generation of signatures for different training sites and classifying the image
using training statistics has to be done. Finally, accuracy of classification is to be estimated
and the user’s, producer’s and overall accuracy is to be estimated. Based on the crop
concentration statistics, agrophysical and / or agroclimatic conditions, the study area is
divided into homogenous strata and sample segments from each stratum are analysed.
Historical / In-season remote
sensing and Ancillary data
Sampling Ground truth
Plan Planning
In-season
remote sensing Ground truth
data collection
Sample segment Training
extraction signature
Image classification
Classification accuracy
Mapping / Acreage
aggregation
Figure 13.1 Methodology for Crop Mapping and Acreage estimation
During the Kharif season, the availability of cloud free data of optical sensors are difficult and thereby the
potential of microwave sensor operated in C-band is utilized for acreage estimation and crop monitoring.
2.Yield monitoring
Yield is influenced by a large number of factors such as crop genotype, soil characteristics, cultural practices
adopted (e.g. irrigation, fertilizer), weather conditions, and biotic influences, such as weeds, diseases,
pests, etc.
The procedure for yield mapping is explained in the Figure 13.2. The classified data with
different crops is the input required. A crop mask highlighting specific crop can be
prepared using any GIS software. Remote Sensing data is used to estimate some of the
biometric parameters, which in turn are input parameters to a yield model. Spectral index
of the crop canopy (NIR / Red, Greenness, NDVI) at any given point of time reveals the
crop growth and its decay as affected by various factors in the time domain. These inputs
are provided to any crop model and the output is overlaid on the crop map to arrive at
the yield map.
The crop health/condition is affected by factors such as supply of water and nutrients, insect/pest attack,
disease out-break and weather conditions. These stresses cause physiological changes which may alter
the optical properties of leaves and bring about changes in crop geometry. The regular monitoring of crop
health helps in differentiation of stressed crops from the normal crop at a given time, quantification of
extent and severity of stress and ultimately assessment of production loss, if any.
Figure 13.2 Methodology for Yield mapping
Crop Simulation Model Satellite Data Products
DSSAT 4.5 Sentinel 1A
SAR Data
CERES - Maize
Pre–Processing
MAPscape 5.4
Simulated LAI
dB Map
Integrating with DSSAT Simulated LAI and dB
LAI
Integrating with DSSAT Simulated Yield Data
LAI based Yield
Validation
Map
Methodology for integrating remote sensing data with DSSAT model for yield estimation
