IAPRS & SIS, Vol.34, Part 7, “Resource and Environmental Monitoring", Hyderabad, India, 2002
CROP PRODUCTION FORECASTING USING REMOTE SENSING DATA: INDIAN
EXPERIENCE
J.S. Parihar and V.K. Dadhwal
Space Applications Centre (ISRO), Ahmedabad — 380 015, INDIA
KEY WORDS: Crop production, Crop Inventory, Spectral Yield Models, Sample Design, Stratification, WiFS, RADARSAT
ABSTRACT:
The Indian experience, in use of satellite remote sensing (RS) data for estimation of crop area and forecast of likely yield, gained
over past more than 15 years of implementation of ‘Crop Acreage and Production Estimation’ (CAPE) project is described. This
project has been sponsored by Ministry of Agriculture (MOA) and uses single-date RS acquisition and sample segment-based crop
inventory and yield forecast models for making production forecast for 7 major growing areas of 7 crops (wheat, rice, cotton,
sugarcane, mustard, groundnut and sorghum). The methodology has been continuously improved to benefit from improved sensor
capabilities. The approach for district-level estimates has now incorporated in-season digital stratification and smaller segment size
to attain lower sampling error. Use of regional sampling strategy using land cover, crop distribution and administrative boundary
information for large area crop inventory has been demonstrated. National-scale forecasts for wheat and rice have been demonstrated
using multiple-date WiFS and RADARSAT data, respectively. The use of RS data for yield forecasting, with high and coarse
resolution and single or multiple acquisitions has generally followed empirical regression approach. The information on soil and
weather has been integrated for yield modelling through use of crop simulation models. A successor to CAPE project, i.e., FASAL
(Forecasting Agricultural output using Space Agrometeorology and Land-based observations), which aims at multiple national-scale
forecasts with synergistic use of all information sources is expected to be soon implemented by MOA.
1.0 INTRODUCTION
The modern crop statistics in India form an uninterrupted series
since the Government of India made a wheat assessment/
forecast as early as 1884. The acreages are derived from
complete enumeration based on land revenue system and yield
on sample crop cutting experiments. This system is not able to
meet the much-desired requirement of pre-harvest forecasts for
policy and planning decisions. Since the successful
demonstration of capability of RS technology for worldwide
operational wheat production forecast during LACIE in mid-
seventies (MacDonald and Hall, 1980), RS has emerged as an
important data source for crop forecasting.
A systematic study on crop inventory using CIR aerial data was
carried out in the joint ISRO-ICAR Agricultural Resource
Inventory and Survey Experiment (ARISE) project, which was
conducted in Anantapur District (Andhra Pradesh) and Patiala
District (Punjab) in 1974-75. An interesting finding was under
reporting of acreage of paddy, a levy crop by government
agencies. These studies were extended to cotton, including
vigour identification in Karjan (Gujarat), and rice and
sugarcane in Mandya (Karnataka), where multi-temporal data
allowed crop identification and condition assessment.
Use of satellite data for crop inventory in India faces tougher
technical challenges (Sahai, 1985) due to (a) small field sizes,
(b) a large diversity of crops sown in an area, (c) large field-to-
field variability in sowing and harvesting dates, cultural
practices and crop management, (d) large areas under
rainfed/dryland agriculture with poor crop canopies, (€) practice
of inter-cropping and mixed cropping, and (f) extensive cloud
cover during kharif crop season. However, significant success
has been achieved due to intensive efforts in this direction.
The Indian experience on use of satellite data for crop
production forecasting mainly comes from studies conducted at
Space Applications Centre in collaboration with a number of
other agencies, first under a project on ‘Crop Production
Forecasting’ (CPF) Project that was initiated in 1983. The
studies were conducted in selected districts for wheat, rice and
groundnut. The first study on wheat acreage estimation for
Karnal was carried out in 1983-84 season (Dadhwal and
Parihar, 1985). The promising results led to an attempt to
estimate state-level wheat acreage using Landsat MSS data for
Haryana and Punjab in 1985-86 (Dadhwal, 1986). The results
were considered encouraging and a project called Large Area
Crop Acreage (LACA) was initiated. It became a sponsored
project by Ministry of Agriculture and yield forecasting was
included. Since 1988 it is known as 'Crop Acreage and
Production Estimation’ (CAPE) project.
A large body of experience has been gained in CAPE project on
efficient sample design, factors affecting crop discrimination,
spectral-yield relationships and realization of timeliness and
accuracy for pre-harvest crop forecasts. Understanding of user
requirements and various limitations in CAPE has led to
formulation of a FASAL (Forecasting Agricultural Output
Using Space, Agrometeorology and Land-based Observations)
proposal to meet the stringent requirements of multiple, nation-
wide and multi-crop forecasts. The major studies and results
from these projects are briefly described here.
It would not be possible to cite all the contributions in this area
and reader may refer to reviews by Navalgund and Sahai
(1985), Sahai (1985), Sahai and Dadhwal (1990), Navalgund et
al., (1991), Dadhwal (1999) and Dadhwal et al. (2002).
2.0 CROP ACREAGE & PRODUCTION ESTIMATION
(CAPE) PROJECT
CAPE aims at regional / group of district-level production
forecasting by combining RS-based crop inventory with yield
forecasts based on weather or spectral index yield empirical
models. It covered wheat, rice, groundnut and rabi sorghum in
