pe
incé the
est crop
ction by
inalysed
ification
i fallow
es were
at India,
f 40.18,
nations
s in GIS
ity in a
mation
yduction
veloped
timation
canning
ites IRS
district,
second
/1D on-
| (PAN),
ry useful
AN data
0 to 750
1 micro-
LISS-III
:d bands
vering a
seful for
c. grown
ation of
(Krishna
ind near
re-visit,
deriving
of the
it period
ngan et
deriving
'ari delta
ivated in
the post kharif rice fallow fields of Krishna delta of Krishna
and Guntur districts of Andhra Pradesh (Navalgund, et al.
1996).
The procedures are also developed and are operational to
estimate rice cropped area using microwave data from
RADARSAT (Chakraborty et al, 1997) to overcome the
problem of non-availability of cloud free optical satellite data
during the kharif season. National level wheat production
forecast using multi date WiFS data are operational under the
Forecasting Agricultural Output using Space, Agrometeoro-
logy and Land based Observations (FASAL) project (SAC,
1999). Multi temporal IRS-WiFS data of kharif (October, 1999)
and rabi (January, 2000) were used in the study with the
following objective.
2. OBJECTIVE
To identify, delineate, estimate acreages and provide maps
showing spatial distribution of kharif rice crop and the post
kharif rice fallow lands in India, Pakistan, Nepal and
Bangladesh countries of the South Asia using temporal WiFS
data
3. STUDY AREA
All the major kharif rice cultivating States / Provinces of the
four countries viz., India, Nepal, Pakistan and Bangladesh
constituted the study area. Eighteen major rice growing States
of India were selected, contributing to about 98% of kharif rice
acreage of the country.
4. METHODOLOGY
The methodology essentially consisted of the selection of the
datasets, processing of the satellite data, incorporation of
ground information, analysis of the satellite data and generation
of the output products in hard and soft formats.
4.1 Selection of the satellite datasets
After the harvest of kharif rice, the land will be either left
fallow or cultivated with a suitable crop in the following rabi
season. The time gap between the harvest of the kharif rice and
the cultivation of the rabi crop depends upon the suitability of
the prevailing weather, availability of water etc. Satellite data
of the period immediately after the harvest of kharif rice crop
will depict large area under fallows though these lands are sown
with rabi crop because of poor manifestation on the image
leading to in over estimation of the fallow lands. In order to
properly estimate the post kharif rice fallows, satellite data of
rabi period was selected based upon the prevailing cropping
pattern of the region coinciding with the maximum vegetative
stage of the dominant crop and manifested crop is clearly
discernable on the satellite data.
IRS-WiFS data corresponding to the peak vegetative stage of
rice of the kharif 1999 season ( October) and the other data of
the rabi 2000 (January to March), post kharif period were
selected after extensive data browsing. The state specific crop
calendar information on the staggering in the transplantation
operations of the paddy crop was utilised in the selection of the
appropriate satellite data. Information on variations in crop
calendar across the states within the kharif rice season was also
IAPRS & SIS, Vol.34, Part 7, “Resource and Environmental Monitoring”, Hyderabad, India, 2002
245
essential, as the period of post kharif season would vary as
influenced by the actual rice-growing period during the kharif
season.
Forty three IRS-WiFS scenes were analysed in ths study.
Efforts were also made to have a uniform database across the
study area to obtain the standardised estimates as well as to
optimise the number of satellite datasets required for the
investigation. In case of States wider than the 800 km swath of
the WiFS sensor, multiple (2 to 3) satellite datasets were used.
Subsets of the satellite data, to the extent of the State coverage,
were extracted and the analysis was carried out at individual
State level. The corresponding data sets were mosaicked and
the State administrative boundaries were overlaid.
4.2 Geo-referencing of the datasets
Survey of India generated toposheets on 1: 1000000 scale were
used in the study, which were geo-referenced to polyconic
everest projection and subsequently used for geometric
correction of the satellite datasets. The first date satellite data,
of kharif or rabi seasons, were registered with these projected
toposheets using affine transformation technique followed by
re-sampling to obtain the geo-referenced image, which was
considered as the master image. The satellite data of the
subsequent dates were co-registered with this master image.
ERDAS - IMAGINE 8.0v was used in geo-referencing of the
data. Polyconic projection was applied uniformly to all the
datasets with graticular (latitude / longitude) information in
degrees. Common central meridian points and origins were
used in developing models for geo-referencing wider States,
which were not covered in a single swath of WiFS, to enable
efficient georeferencing while mosaicking.
4.3 Overlaying of the administrative boundaries and forest
masks
The administrative boundaries of the countries viz., Pakistan,
Nepal and Bangladesh obtained from ICRISAT were overlaid
on the satellite data to generate country level information and to
develop the output products. In case of India, statewise
boundaries were overlaid. Forest masks were also generated
and were excluded in the classification of satellite data.
4.4 Ground truth information
This information was highly essential right from the selection
of the data to the verification of the results at various
intermediate steps such as defining the training areas,
generation of spectral signatures, testing the separability and
classification of the satellite data. Since the analysis was carried
out post season, real time ground truth was not used in this
project. Hence, ground information as available in the form of
published literature was used to derive the information on the
growing period and distribution of rice crop in India. Ground
information corresponding to Pakistan and Bangladesh
countries provided by the ICRISAT were used in the analysis.
In case of Nepal, the ground information of Uttar Pradesh State
of India was extrapolated because of the spatial contiguity of
these two regions.
4.5 Analysis of the satellite data
Digital data analysis techniques viz., delineation of the training
areas, generation of spectral signatures, checking for the
spectral separability of different agricultural land covers,
classification of the datasets, were applied in the analysis and