int
ge
nd
se
he
ne
re
be
tal
ve
ge
ta
on
en
S-
MONITORING OF LAND COVER CONDITIONS IN PADDY FIELDS
USING MULTITEMPORAL SAR DATA
Shoji Takeuchi*, Mitsunori Yoshimura* and Rasamee Suwanwerakamtorn**
* Remote Sensing Technology Center of Japan (RESTEC)
** National Research Council of Thailand (NRCT)
ABSTRACT: The synthetic aperture radar (SAR) data by the Japanese Earth Resources Satellite 1, JERS-1, are
expected to be used effectively for the purpose of monitoring of land cover conditions in tropical regions since the SAR
The authors conducted case studies for monitoring of land
cover conditions in the paddy fields in Thailand using multitemporal JERS-1 SAR data. Two case studies concerning to
the change of land cover by seasonal variation and flooding proved that these changes can be detected as the change of
L-band SAR backscatter. The result of these case studies indicates the effectiveness of L-band SAR data by JERS-1 for
the monitoring of several types of changes of land cover conditions in tropical regions, where data acquisition is
data can be obtained repetitively in any weather condition.
necessary even in bad weather conditions of the rainy season.
KEY WORDS: SAR, Multitemporal Data, Backscatter, Soil Moisture, Vegetation Coverage, SWI, PVI, Flood.
1.INTRODUCTION
Land cover change monitoring using multitemporal data
is one of the important and practical applications in
remote sensing fields. However, the observations by
conventional optical sensors are much affected by
weather conditions. Especially in tropical regions,
observation in rainy season has been almost impossible
using optical sensors. On the other hand, the Synthetic
Aperture Radar (SAR) can observe in any weather
condition and the SAR data can be used practically to
monitor land cover changes which occur in the season of
bad weather conditions like the rainy season. The
authors studied the applicability of the SAR data by the
Japanese Earth Resources Satellite 1, JERS-1, for the
monitoring of the land cover change in paddy fields at a
test site of the Central Plain of Thailand.
2.RELATION BETWEEN SAR BACKSCATTER
CHANGE AND LAND COVER CONDITIONS
The authors have conducted the preliminary analysis of
the multitemporal JERS-1 SAR data taken in 1993 at the
test site of Thailand and the result suggests that those
data contain much information about the seasonal
variations in SAR backscatter for various land cover types.
Three images in Fig.1 show seasonal changes of L-band
SAR backscatter among the middie of dry season (March
16th) , the end of dry season (April 29th) and rainy
Season (September 8th). The change of SAR
backscatter is represented by the Normalized Radar
Cross Section (NRCS) which can be derived from the
SAR level 2.1 product using the calibration coefficient
given by NASDA.
683
For the purpose to investigate the physical meanings of
the SAR backscatter change, regression analysis was
performed between SAR backscattering intensity and two
parameters for representing land cover conditions, Soil
Wetness Index (SWI) and Perpendicular Vegetation Index
(PVI) using JERS-1 SAR and Landsat TM images
acquired on the same day (April 29th, 1993). These
parameters can be estimated in the two dimensional
plane composed by the intensity of TM band 3 (visible-
red) and band 4 (near infrared) as shown in Fig.2 . The
SWI is defined as the regression line derived from points
regression of bare soils and indicates the degree of soil
moisture content. The PVI is defined as the line
perpendicular to the SWI line and indicates the degree of
vegetation coverage. The maximum value of PVI
corresponds to the averaged value of PVI for the areas
fully covered by vegetation.
The two images were co-registered each other and twenty
pairs of the averaged values in 5 x 5 window areas for
SAR and TM images which were sampled from the bare
soil areas and the vegetated areas were used for the
regression respectively. Fig.3 and 4 show the results
from the regression analysis between SAR-NRCS and
SWI and between SAR-NRCS and PVI respectively.
The results indicate positive correlation for both of the
relations between SAR backscatter and soil moisture and
between that and vegetation coverage, although the
correlation coefficients are moderate values. There are
seen relatively large variations of NRCS for large PVI
samples and for dry bare soil samples, which are the
factors decreasing correlation coefficients. The variation
for PVI is considered to be caused by the differences of
vegetation types, that is, the SAR backscatter for grasses
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B7. Vienna 1996