IAPRS & SIS, Vol.34, Part 7, "Resource and Environmental Monitoring", Hyderabad, India,2002
intermediate and deep in South 24 Paragana district. Plant
maturity and life span of the crop is variety dependent and
harvesting period is November -December.
3. DATA USED
RADARSAT ScanSAR Narrow B (SCNB) multidate dataset
with C-band, HH-polarization and 31-46°(degree) incidence
angle have been used as data source. The SCNB data generated
by combining the beams W2, S5 and S6 has been selected for
its large area coverage (300-km swath), acceptable pixel
spacing (25 m), high temporal resolution (2/3 side-lap pass
within 7 days) and lesser volume of data. The methodology
developed using this data set will be used for district-wise
National Kharif Rice Monitoring programme. The C- band and
shallow incidence angle have been selected for its suitability for
identification of crops. The shallow incidence angle increases
the path length through the vegetation maximising crop
response and hence more volume/multiple scattering
contributions to radar backscatter than surface scattering.
The sensitivity of water surface roughness created by wind
induced ripple to SAR backscatter is reduced by using HH
polarization and large incidence angle. The first date data was
acquired coinciding with the puddling to transplanting stage
and the other two dates were acquired on 24 days repitivity. In
case of Orissa (Baleswar and Bhadrak districts), the multi-date
are 7^ July, 31* July and 24” August 1999. In case of West
Bengal (South 24 Paragana district), two year's data were used
for analysis viz.14th July, 7^ August and 31* August for 2000
and 9th July, 2™ August, 27" August for 2001. The three-date
data were co-registered in to single multitemporal FCC data set,
which reflects well the variation in rice planting of the study
area. The multi-date FCC image of South 24 Paragana district
(West Bengal) as well as Baleswar and Bhadrak districts
(Orissa) are shown in Figure la and 1b respectively. The
ground truth data was collected coinciding with the satellite
pass by the scientists of Institute of Wetland Management and
Ecological Design (IWMED) for West Bengal and the scientists
of ORSAC for Orissa.
Figure la. Multi-date georeferenced FCC image of South 24
Paragana district of West Bengal.
Figure 1 b. Multi-date georeferenced FCC image of Baleswar
and Bhadrak districts of Orissa.
4. METHODOLOGY
An empirical quadratic fit approach has been adopted to relate
the plant height and the backscatter co-efficient and is given by,
0°= c+bh+ah’ 0
where 0° = backscatter co-efficient in dB
h = height of the crop in cm
a,b,c = constants of the model
By regressing the backscatter with plant height, the constants a,
b and c were determined as,
a- -0.002100; b = 0.335082 and c= -18.9333.
This empirical relationship has been developed using
RADARSAT Standard beam 5 over the Burdhwan districts of
West Bengal and with intensive ground truth study over the
entire growing season. The same model has been found
applicable for ScanSAR Narrow B, as the incidence angle (38)
of Standard beam 5 matches with that of the Scan SAR Narrow
B (SCNB) data.
The radar backscatter has been analysed as a function of plant
height as depicted in Figure 2.The empirical relationship has
been inverted to estimate the plant height using the inverse
formula,
h = -b/2a - V{(b/2a)"-(c-0°/a) } (ii)
In this study, the height is considered to be an important
parameter as it can be measured in the field. À standard of 5-cm
plant height is considered fixed, when the rice plant is ready for
transplantation. Using the ground truth information, the
predicted height is related to the days since transplantation. The
relationship between the predicted height, days since
transplantation and backscatter co-efficient are shown in
Table 1.
A model is coded in EASI-plus macro language of EASI/PACE
software. The model first classifies the rice subclasses (normal,
late and very late-planted rice crop) as well as nonrice areas